Post on 05-Jan-2017
I
Caractérisation par spectroscopie et analyse compositionnelle des formes du phosphore dans des
sols agricoles canadiens
Thèse
Dalel Abdi
Doctorat en sols et environnement
Philosophiae doctor (Ph. D.)
Québec, Canada
© Dalel Abdi, 2014
II
III
RÉSUMÉ
Une meilleure compréhension de la dynamique et des changements des formes du
phosphore (P) dans les sols agricoles est indispensable pour le maintien de leur productivité
et de la qualité des eaux de surface. L’objectif de cette thèse a été de développer et
d’utiliser des méthodes innovatrices pour mesurer les différentes formes de P et caractériser
leurs changements dans des sols soumis à différentes pratiques culturales. Nous avons
évalué le potentiel de la spectroscopie dans le proche infrarouge (SPIR) à prédire le P total
(PT), le P chimiquement extrait à la solution Mehlich-3 (PM3) et à l’eau (Cp) et le P
organique (Po) dans deux sites de teneurs variables en P situés au Québec et en
Saskatchewan. Les résultats obtenus ont démontré que la prédiction du PT et du PM3 dans
le site du Québec est modérément utile et non acceptable, respectivement. Cependant, des
résultats inverses ont été trouvés dans le site du Saskatchewan. La prédiction du Po est de
modérément utile à modérément réussie dans le site du Saskatchewan. Le potentiel de la
prédiction de ces formes du P par la SPIR dépend de la texture du sol, de leur variation
dans le sol et de leur lien à la matière organique. En outre, nous avons démontré que les
résultats de l’analyse de variance et de la corrélation des espèces moléculaires de P, brutes
ou ordinairement transformées, varient en fonction de leur unité de mesure. L’utilisation de
l’analyse compositionnelle avec les transformations du log ratio centré ou du log ratio
isométrique permet d’éviter ce biais et d’avoir des interprétations cohérentes des résultats.
Finalement, l’analyse par résonance magnétique nucléaire des sols sous rotation de maïs et
de soya de longue durée au Québec a démontré que l’accumulation de PT (1326 mg kg-1)
dans la couche superficielle du sol (0-5 cm) soumis à la fertilisation phosphatée et au semis
direct était principalement due aux ions orthophosphates (49,7% du PT). Cependant, les
formes organiques s’accumulaient en profondeur sous forme d’inositols monoesters et de
nucléotides qui sont donc susceptibles d’atteindre les cours d’eaux adjacents par drainage.
Ce projet de recherche nous permet de mieux caractériser et gérer les formes de P dans les
écosystèmes cultivés en adoptant les pratiques culturales adéquates.
IV
V
ABSTRACT
Understanding of phosphorus (P) forms dynamics and changes in agro-ecosystems
is essential for the development of best management practices to maintain soil productivity
and surface water quality. The objective of this thesis was to develop and use innovative
methods to characterize soil P forms and their changes under different management
practices. We examined the potential of near infrared spectroscopy (NIRS) to predict soil
total P (TP), soil P extracted by Mehlich-3 solution (M3P) and by water (Cp), and soil
organic P (OP) for soil samples taken from two sites with different P content located at
Quebec and Saskatchewan. The results showed that the prediction of TP and M3P in the
site of Quebec were moderately useful and not acceptable, respectively. However, the
opposite was found in the site of Saskatchewan. The prediction of OP was moderately
useful to moderately successful in experimental site of Saskatchewan. The potential of
NIRS to predict P depends to the soil texture, to P soil content variation and to the relation
of P to organic matter. Furthermore, contradictory results of variance and correlations
analyses were found for the raw and ordinary log transformed molecular P species
expressed as proportions or concentrations, indicating spurious correlations. Using
compositional analysis with centred log ratio or isometric log ratio transformations avoid
the methodological biases and allow coherent interpretation. Finally, phosphorus-31
nuclear magnetic resonance spectroscopy was used to characterize P species for soil
samples collected from a long-term corn-soybean rotation experiment in Quebec. Results
showed an accumulation of TP (1326 mg kg-1) on the top 5 cm of P fertilized soil under no-
till primarily due to orthophosphate ions accumulation (49.7% of TP). However, the
organic P forms of inositol monoesters and nucleotides had accumulated in the deep layer;
indicating a potential loss through different hydrological pathways. Overall, these studies
allow us to improve our understanding of P forms and to better monitor them under
different agro-ecosystems using the best management practices.
VI
VII
TABLE DES MATIÈRES
RÉSUMÉ .............................................................................................................................. III
ABSTRACT ........................................................................................................................... V
TABLE DES MATIÈRES .................................................................................................. VII
LISTE DES TABLEAUX ................................................................................................... XI
LISTE DES FIGURES ...................................................................................................... XIII
ABRÉVIATIONS ET DÉFINITIONS ............................................................................... XV
DÉDICACE ..................................................................................................................... XVII
REMERCIEMENTS .......................................................................................................... XIX
AVANT-PROPOS ............................................................................................................. XXI
CHAPITRE I: INTRODUCTION .......................................................................................... 1
CHAPITRE II: REVUE DE LA LITTÉRATURE ................................................................. 5
2.1 Cycle biogéochimique du phosphore dans le sol .......................................................... 5
2.2 Les formes du phosphore dans les sols ......................................................................... 6
2.2.1 Phosphore inorganique .......................................................................................... 6
2.2.2 Phosphore organique .............................................................................................. 7
2.3 Mesures du phosphore du sol ...................................................................................... 10
2.3.1 Méthodes conventionnelles .................................................................................. 10
2.3.1.1 Phosphore total ............................................................................................. 10
2.3.1.2 Phosphore disponible aux plantes ................................................................. 10
2.3.1.3 Phosphore organique ..................................................................................... 11
2.3.1.4 Pools du P ..................................................................................................... 12
2.3.2 Méthodes spectroscopiques ................................................................................. 14
2.3.2.1 Spectroscopie dans le proche infrarouge ...................................................... 14
2.3.2.2 Spectroscopie de résonance magnétique nucléaire du 31P ............................ 15
2.4 Changements des formes du phosphore selon les pratiques culturales....................... 16
2.4.1 Changements des pools du P ............................................................................... 16
2.4.2 Changements des espèces de P ............................................................................ 17
2.5 Concept d’analyse des données compositionnelles .................................................... 18
2.6 Hypothèses .................................................................................................................. 21
2.7 Objectifs ...................................................................................................................... 22
2.8 BIBLIOGRAPHIE ...................................................................................................... 23
CHAPITRE III: PREDICTING SOIL PHOSPHORUS-RELATED PROPERTIES USING
NEAR-INRARED REFLECTANCE SPECTROSCOPY .................................................... 31
3.1 RÉSUMÉ .................................................................................................................... 32
3.2 ABSTRACT ................................................................................................................ 33
3.3 INTRODUCTION ...................................................................................................... 34
3.4 MATERIALS AND METHODS ................................................................................ 35
3.4.1 Experimental site description ............................................................................... 35
3.4.2 Soil and plant analyses ......................................................................................... 36
3.4.3 Near-infrared reflectance spectroscopy spectrum acquisition ............................. 37
3.4.4 Pretreatment, calibration, and cross-validation .................................................... 38
3.4.5 Validation ............................................................................................................. 39
3.5 RESULTS AND DISCUSSION ................................................................................. 39
VIII
3.5.1 Reference data ..................................................................................................... 39
3.5.2 Spectral pretreatments ......................................................................................... 40
3.5.3 Near-infrared reflectance spectroscopy prediction of soil and crop P properties 40
3.5.4 Near-infrared reflectance spectroscopy prediction of other soil properties ........ 42
3.6 CONCLUSIONS ........................................................................................................ 43
3.7 ACKNOWLEDGEMENTS ....................................................................................... 43
3.8 REFERENCES ........................................................................................................... 44
CHAPITRE IV: PREDICTING SOIL ORGANIC PHOSPHORUS USING NEAR-
INFRARED REFLECTANCE SPECTROSCOPY ............................................................. 55
4.1 RÉSUMÉ .................................................................................................................... 56
4.2 ABSTRACT ............................................................................................................... 57
4.3 INTRODUCTION ...................................................................................................... 58
4.4 MATERIALS AND METHODS ............................................................................... 59
4.4.1 Experimental site description .............................................................................. 59
4.4.2 Soil sampling and analysis .................................................................................. 59
4.4.3 Near-infrared reflectance spectroscopy spectrum acquisition ............................ 60
4.4.4 Spectral pre-treatment ......................................................................................... 60
4.4.5 Calibration, cross-validation and validation ........................................................ 60
4.5 RESULTS AND DISCUSSION ................................................................................ 61
4.5.1 Soil reference data ............................................................................................... 61
4.5.3 Spectral pre-treatment, calibration, and prediction of soil organic P .................. 62
4.5.4 Spectral pre-treatment, calibration, and prediction of soil total and Mehlich-3 P
...................................................................................................................................... 62
4.5.5 Spectral pre-treatment, calibration, and prediction of soil organic matter and
Mehlich-3 nutrients ...................................................................................................... 63
4.6 CONCLUSION .......................................................................................................... 64
4.7 REFERENCES ........................................................................................................... 65
CHAPITRE V: UNBIASED STATISTICAL ANALYSIS OF SOIL 31P-NMR ................. 73
5.1 RÉSUMÉ .................................................................................................................... 74
5.2 ABSTRACT ............................................................................................................... 75
5.3 INTRODUCTION ...................................................................................................... 76
5.4 MATERIALS AND METHODS ............................................................................... 77
5.4.1 Datasets ............................................................................................................... 77
5.4.1.1 Compositional data transformations ................................................................. 78
5.4.1.1.1 Centred log-ratio transformation ............................................................... 78
5.4.1.1.2 Isometric log-ratio transformation ............................................................ 78
5.4.1.1.3 Choice of SBP ........................................................................................... 79
5.4.1.1.4 Ordinary logarithmic transformation ........................................................ 80
5.4.1.2 Statistical analysis ............................................................................................ 81
5.5 RESULTS AND DISCUSSION ................................................................................ 81
5.5.1 Biased analysis of variance ................................................................................. 81
5.5.2 Spurious correlations ........................................................................................... 83
5.6 CONCLUSIONS ........................................................................................................ 84
5.7 REFERENCES ........................................................................................................... 85
CHAPITRE VI: LONG-TERM IMPACT OF TILLAGE PRACTICES AND P
FERTILIZATION ON SOIL P FORMS AS DETERMINED BY 31P NUCLEAR
MAGNETIC RESONANCE SPECTROSCOPY ................................................................ 99
IX
6.1 RÉSUMÉ .................................................................................................................. 100
6.2 ABSTRACT .............................................................................................................. 101
6.3 INTRODUCTION .................................................................................................... 102
6.4 MATERIELS AND METHODS .............................................................................. 104
6.4.1 Experimental site ............................................................................................... 104
6.4.2 Soil sampling and chemical analysis ................................................................. 105
6.4.3 Solution 31P-NMR spectroscopy ........................................................................ 105
6.4.4 Compositional data analysis .............................................................................. 106
6.4.5 Statistical analysis .............................................................................................. 107
6.5 RESULTS ................................................................................................................. 107
6.5.1 Chemical soil properties .................................................................................... 107
6.5.2 Identification of phosphorus forms by 31P nuclear magnetic resonance
spectroscopy ................................................................................................................ 108
6.5.3 Distribution of 31P nuclear magnetic resonance phosphorus forms ................... 109
6.6 DISCUSSION ........................................................................................................... 110
6.7 CONCLUSIONS ...................................................................................................... 113
6.8 ACKNOWLEDGMENTS ........................................................................................ 113
6.9 REFERENCES ......................................................................................................... 114
CHAPITRE VII: CONCLUSIONS ET RECOMMANDATIONS .................................... 127
CHAPITRE VIII: ANNEXE .............................................................................................. 131
COMPOSITIONAL ANALYSIS OF POOLS IN CANADIAN MOLLISOLS ................ 131
8.1 RÉSUMÉ .................................................................................................................. 132
8.2 ABSTRACT .............................................................................................................. 133
8.3 INTRODUCTION .................................................................................................... 134
8.4 MATERIALS AND METHODS .............................................................................. 135
8.4.1 Isometric log ratio transformation and the Aitchison distance .......................... 136
8.4.2 The Mackenzie et al. (1992) dataset .................................................................. 137
8.4.3 The Tiessen et al. (1983) dataset ....................................................................... 138
8.5 CONCLUSION ......................................................................................................... 138
8.6 REFERENCE ............................................................................................................ 140
X
XI
LISTE DES TABLEAUX
Tableau 2-1 Composés organiques phosphatés communs du sol (Turner et al., 2005). ........ 9
Tableau 3-1 Descriptive statistics† for the soil P-related and other properties analyzed
using reference methods. .............................................................................................. 48
Tableau 3-2 NIRS spectral pre-treatments and statistics† of calibration, cross-validation,
and validation for the P-related soil properties. ............................................................ 49
Tableau 3-3 NIRS spectral pre-treatment and statistics† of calibration, cross-validation,
and validation for the other soil properties. .................................................................. 50
Tableau 4-1 Descriptive statistics for the soil organic (OP) and total (TP) P analysed for
long- and short-term no-till (NT) treatments. ............................................................... 67
Tableau 4-2 Descriptive statistics for the soil Mehlich-3 extracted nutrients and organic
matter for long- and short-term no-till (NT) treatments. .............................................. 67
Tableau 4-3 Statistics of near-infrared reflectance spectroscopy calibration, cross-
validation, and validation for soil (OP) and total (TP) P analysed for long- and short-
term no-till (NT) treatments. ......................................................................................... 68
Tableau 4-4 Statistics of near-infrared reflectance spectroscopy calibration, cross-
validation, and validation for soil Mehlich-3 extracted nutrients and organic matter for
long- and short-term no-till treatments. ....................................................................... 69
Table 5-1 Sequential binary partition of soil 31P-NMR P species analyzed by Abdi et al.
(2014). ........................................................................................................................... 88
Table 5-2 Sequential binary partition of soil 31P-NMR P species analyzed by Cade-Menun
et al. (2010). .................................................................................................................. 89
Table 5-3 ANOVA of the effect of tillage (T), P fertilization (P) and soil depth (D) on log-
and clr-transformed P compositions (Abdi et al., 2014). P species defined in Table 1.
...................................................................................................................................... 90
Table 5-4 ANOVA of the effect of tillage and soil depth on log- and clr-transformed P
compositions (Cade-Menun et al., 2010). P species defined in Table 2. ...................... 92
Table 6-1 Analysis of variance for the effects of tillage, P fertilization and depth on clr
transformed concentrations of soil total P (TP), Mehlich-3 extractable P (PM3),
aluminium (Al), iron (Fe), calcium (Ca), magnesium (Mg), total carbon (TC) and total
nitrogen (TN), and pH. ............................................................................................... 118
Table 6-2 Chemical shift of P forms detected in the 31P-NMR spectrum of the soil as
affected by tillage and P fertilization management and depth. ................................... 119
Table 6-3 Analysis of variance for the effects of tillage, P fertilization, and depth on
centered log ratio–transformed soil P forms determined by 31P nuclear magnetic
resonance spectroscopy. .............................................................................................. 120
Table 6-4 Back-centered log ratio–transformed soil P forms determined by 31P nuclear
magnetic resonance spectroscopy as affected by tillage, P fertilization, and depth. .. 121
XII
Table 8-1 Sequential binary partitions of soil P fractions (r is number of P fractions with
plus sign and s is number of P fractions with minus sign). ........................................ 141
Table 8-2 Mollisol P fractions following crop sequence and NP fertilization (data from
McKenzie et al., 1992). .............................................................................................. 142
Table 8-3 Ilr coordinates of P pools following crop sequence and fertilization (data from
McKenzie et al., 1992). .............................................................................................. 143
Table 8-4 Ilr differences in P pools between treatments and uncultivated check (data from
McKenzie et al., 1992). .............................................................................................. 144
Table 8-5 Effect of time on P pools in a Mollisol (data from Tiessen et al., 1984). ......... 145
XIII
LISTE DES FIGURES
Figure 1-1. Schéma général du projet de recherche ............................................................... 4
Figure 2-1 Cycle du P dans le sol (adapté du Pierzynski et al., 2005). .................................. 5
Figure 2-2 Effet du pH sur la forme des ions orthophosphates de la solution du sol (Holtan
et al., 1988). .................................................................................................................... 7
Figure 2-3 Méthode d’extraction séquentielle du P du sol selon la méthode de Hedley et
al., 1982 (tirée de Cross et Schlesinger, 1995). ............................................................ 13
Figure 2-4 Spectre obtenu par la spectroscopie magnétique nucléaire du 31P montrant les
composés phosphatés détectés dans la couche superficielle (5-10 cm) d’un sol cultivé
non labouré (Cade-Menun et al., 2010). ....................................................................... 16
Figure 3-1 NIRS predicted values against measured values of (a) soil P content extracted
using the Mehlich 3 method and analysed by colorimetry (M3P_Col); (b) soil P
content extracted using the Mehlich 3 method and analysed by ICP (M3P_ICP); (c)
soil P content extracted with water and analysed by colorimetry (Cp); (d) total soil P;
(e) annual timothy crop P-uptake, and; (f) annual P-budget. Based on validation
statistics reported here and in Table 3.2, NIRS predictions were considered moderately
useful (MU) when 0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25, and less reliable (LR)
when Rv2 < 0.70 and RPD < 1.75 (Malley et al., 2004). ............................................... 53
Figure 4-1 Near-infrared reflectance spectroscopy (NIRS) predicted vs. measured values
of soil organic P analysed for (a) long- and short term no-till, (b) long-term no-till, and
(c) short-term no-till treatments. ................................................................................... 70
Figure 5-1 Relationship between Mahalanobis distance from ilr with (a) ordinary log
transformed 31P NMR-P species concentration, and (b) raw of 31P NMR-P species
concentrations (data from Abdi et al., 2014). ............................................................... 97
Figure 6-1 Distribution of total P (TP), Mehlich-3 extractable P (PM3) and orthophosphate
concentrations at various soil depths under (a, c, e) mouldboard plow (MP) and (b, d,
f) no-till (NT) treatments. P0 and P35 represent additions of 0 and 35 kg P ha−1,
respectively. Values are means of three replications. For each treatment, different
letters indicate significantly different means among soil depth according to LSD
(0.05). † For each treatment, different letters indicate significantly different means
among depth according to LSD (0.1). ......................................................................... 122
Figure 6-2 Distribution of (a) total carbon (TC) and (b) total nitrogen (TN) content, and (c)
Al Mehlich-3 and (d) Mg Mehlich-3 at various soil depths under mouldboard plow
(MP) and no-till (NT) treatments. Values are means of three replicates. For each
XIV
treatment, different letters indicate significantly different means among soil depth
according to LSD (0.05). ............................................................................................ 123
Figure 6-3 Phosphorus-31 nuclear magnetic resonance spectroscopy spectrum showing the
range of P compounds detected at the 0 to 5 cm depth of the mouldboard plow
fertilized treatment (Oth.D1, other diester 1; Oth.D2, other diester 2). ..................... 124
Figure 6-4 Phosphorus-31 nuclear magnetic resonance spectroscopy spectrum showing the
P compounds detected in the monoester region at the 0 to 5 cm depth of mouldboard
plow fertilized treatment. (A) neo-IP6; (B) orthophosphate; (C) myo-IP6; (D) glucose-
6P; (E) unknown; (F) α-glycerophosphate; (G) β-glycerophosphate; (H) nucleotides;
(I) choline-P; (J) scyllo-IP6; (M1) monoester 1; (M2) monoester 2. .......................... 125
Figure 6-5 Distributions of (a) pyrophosphate, (b) scyllo-IP6, (c) DNA and (d) nucleotides
concentrations at various soil depths under mouldboard plow (MP) and no-till (NT)
treatments. Values are means of three replicates. For each treatment, different letters
indicate significantly different means among depths according to LSD (0.05). ........ 126
Figure 8-1 Conceptual relational model between P pools in Mollisols (modified from
Tiessen et al., 1984). ................................................................................................... 146
Figure 8-2 Time change in P balance distances from initial conditions in a Blaine lake soil
(data from Tiessen et al., 1983)……………………………………………….……147
XV
ABRÉVIATIONS ET DÉFINITIONS
clr: Log ratio centré (centred log ratio)
Cp: Concentration des ions phosphates dans la solution du sol estimée par une
extraction à l’eau
CV: Coefficient de variation (coefficient of variation)
ADN : Acide désoxyribonucléique (DNA: Deoxyribonucleic acid)
Espèce de P: Composé ionique ou moléculaire de P
Forme de P: Désigne la nature chimique du P (Pi, Po)
ilr: Log ratio isométrique (isometric log ratio)
IP6: Inositol Hexakisphoaphate
MP: Labour conventionnel (Mouldboard plow)
NT: Semis direct (No-till)
P: Phosphore
Pi: Phosphore inorganique
PM3: Phosphore disponible aux plantes estimé avec la méthode Mehlich-3
Po: Phosphore organique
Pools de P: Fractions du P total déterminées par une extraction séquentielle
PT: Phosphore total
R2: Coefficient de détermination (Coefficient of determination)
RMN-31P: Spectroscopie de resonance magnétique nucléaire du 31P (31P-NMR :
Phosphorus-31 nuclear magnetic resonance)
RPD: Rapport de déviation de la performance de prédiction (Ratio of standard error of
prediction to standard deviation)
SPIR: Spectroscopie dans le proche infrarouge (NIRS: Near-infrared reflectance
spectroscopy)
XVI
XVII
DÉDICACE
À ma famille et à ma Tunisie !
XVIII
XIX
REMERCIEMENTS
Je tiens à remercier vivement ma co-directrice docteur Noura Ziadi, chercheure au
Centre de recherche et de développement sur les sols et les grandes cultures d’Agriculture
et Agroalimentaire Canada, pour la confiance qu’elle m’a accordée en acceptant de diriger
mes travaux de recherches, pour son support scientifique, sa disponibilité, ses conseils, sa
patience, et son encouragement continu sans lesquels cette thèse n’aurait pas été possible.
Je tiens à l’assurer de ma profonde gratitude pour les facilités et les opportunités qu’elle
m’a permis de bénéficier.
Mes remerciements les plus sincères vont également à mon directeur de thèse docteur
Léon-Étienne Parent pour m’avoir fait bénéficier de sa grande compétence et de sa rigeur
scientifique tout au long de mon programme de doctorat. Les discussions que nous avons
eues ensemble et son esprit critique et innovateur m’ont toujours motivée et m’ont permis
de progresser avec beaucoup de succès.
Je suis très honorée à remercier docteur Alfred Jaouich pour avoir accepté de faire la
prélecture de ma thèse. Ses commentaires et suggestions m’ont permis d’améliorer
beaucoup la qualité de cette thèse.
J’adresse mes remerciements aussi aux Dr. Judith Nyiraneza et Dr. Christian Morel
d’avoir accpeté de consacrer du temps pour l’évaluation de cette thèse. Leurs remarques et
suggestions ont permis l’amélioration de la version finale.
Je suis très reconnaisante au docteur Barbara J. Cade-Menun, chercheure au Centre de
recherche sur l’agriculture des Prairies semi-arides d’Agriculture et Agroalimentaire
Canada, pour son acceuil bienveillant durant mon stage et pour m’avoir fait bénéficier de sa
compétence distinguée dans l’étude du phosphore organique, et pour m’avoir formée au
traitement des spectres de la résonance magnétique nucléaire.
Je voudrais aussi adresser ma gratitude aux docteurs Gaëtan F. Tremblay et Gilles
Bélanger, chercheurs au Centre de recherche et de développement sur les sols et les grandes
cultures d’Agriculture et Agroalimentaire Canada, pour leurs remarques et orientations
XX
dans les études faites par la spectroscopie dans le proche infrarouge. Nos nombreuses
discussions étaient à la fois enrichissantes et intéressantes.
Je souhaiterais remercier tout particulièrement Sylvie Michaud, Sylvie Côté, Mario
Laterrière, Bernard Gagnon et Claude Levesque des laboratoires du Centre de recherche et
de développement sur les sols et les grandes cultures d’Agriculture et Agroalimentaire
Canada, pour leur support technique et leur gentillesse.
Je tiens à remercier également Dr. Serge-Étienne Parent, professionnel de recherche en
agrostatistiques à l’Université Laval, pour sa précieuse aide aux analyses mathématiques et
statistiques.
Je remercie le Ministère de l’Enseignement Supérieur et de la Recherche Scientifique
de Tunisie pour le soutient financier de cette thèse. Je tiens à remercier aussi les organismes
canadiens qui m’ont accordée des bourses pour présenter mes résultats de recherches dans
des congrès scientifiques, notamment le centre SÈVE, l’Association Québécoise des
Spécialiste en Sciences du sol et le Bureau des bourses et de l’aide financière de
l’Université Laval.
Je remercie chaleureusement mes amis Aimé Messiga, Yichao Shi, et Mervin St. Luce
pour les discussions qu’on a eues ensemble et pour l’ambiance du travail très agréable.
Enfin, j’adresse mes vifs remerciements à mes parents et à tous les membres de ma
famille pour leur amour, leur confiance et leur encouragement. Un grand merci pour
m’avoir fait de moi ce que je suis aujourd’hui !
XXI
AVANT-PROPOS
Cette thèse est le fruit d’un travail de recherche doctoral qui s’inscrit dans le cadre
d’une coopération entre l’Université Laval et le Centre de recherche et de développement
sur les sols et les grandes cultures d’Agriculture et Agroalimentaire à Québec, Canada.
Dans cette thèse, nous présentons huit chapitres dont le premier et le deuxième
consistent à une introduction générale et à une revue de littérature sur l’ensemble des
thèmes de recherches à l’étude. Les quatre chapitres qui suivent ont fait l’objet de : deux
articles scientifiques publiés dans « Soil Science Society of America Journal » (chapitre 3)
et « Journal of Environmental Quality » (chapitre 6), et un article soumis à Geoderma
(chapitre 5). Le chapitre 4 sera aussi soumis à Geoderma. Le chapitre 7 présente la
conclusion générale pour l’ensemble des études et les recommandations, et le dernier
chapitre présente l’annexe déjà publié suite à un congrès qui a lieu en Espagne.
Tous ces articles ont été rédigés en anglais et sont insérés dans cette thèse comme
publiés ou soumis. Un court résumé en français (envrion 150 mots selon les exigences de
rédaction de thèse de l’Université Laval) précède chacun d’eux. Je suis l’auteure principale
de chaque article et la responsable des analyses de laboratoire, de traitement de données et
d’interprétation des résultats. Les co-auteurs sont le docteur Léon-Étienne Parent de
l’Université Laval et d’autres chercheurs des centres de recherche d’Agriculture et
Agroalimentaire Canada situés à Québec, Qc, et à Swift Current, SK.
Les résultats de ces études ont été présentés dans des congrès nationaux (Québec) et
internationaux en Espagne, Italie, Suède, Panama, France et en Corée du Sud pour un total
de sept communications orales et neuf affiches. Un résumé du chapitre 6 a été publié
également dans « Crops Soils Agronomy News » de la revue américaine « American
Society of Agronomy » en Juillet, 2014.
1
CHAPITRE I: INTRODUCTION
Le phosphore (P) est un élément naturel qui se trouve dans tout organisme vivant. Il
est impliqué dans plusieurs fonctions vitales telles que les métabolismes énergétiques
(AMP, ADP, ATP) et la constitution des acides nucléiques (ADN, ARN) et des membranes
cellulaires (phospholipides). Chez les plantes, le P joue un rôle important dans la
photosynthèse, la floraison et la production des fruits (Morel, 1989).
Dans la nature, le P est peu abondant; il constitue environ 0,12% des éléments de
l’écorce terrestre (Cathcart, 1980). Il se trouve dans le sol, les sédiments et les eaux de
surface en provenance de la désagrégation des roches minérales phosphatées,
majoritairement de l’apatite (Stevenson, 1986). Les principales réserves naturelles en
phosphate se trouvent au Maroc-Sahara Occidental (74,6%), en Chine (5,5%), en Algérie
(3,2%), en Syrie (2,7%), en Afrique du sud (2,2%), en Jordanie (1,9%), en Russie (1,9%) et
aux États-Unis (1,6%) (U.S. Geological Survey, 2014). Bien qu’elles soient abondantes, ces
réserves en P sont en voie d’épuisement face à l’accroissement de l’industrie des engrais
phosphatés (Cordell et al., 2009).
La forme naturelle du P (apatite, variscite, strengite ou vivianite,..) est peu
disponible aux plantes étant donné qu’elle est peu soluble à l’eau, d’où la nécessité de la
fertilisation phosphatée. Cependant, seulement une faible proportion du P ajouté est
assimilée par la plante et le reste est rapidement fixé par les argiles ou les sesquioxydes de
fer et d’aluminium (Khiari et Parent, 2005). D’autre part, l’intensification des apports des
fertilisants phosphatés en agriculture est à l’origine des pertes en P qui atteignent les cours
d’eau et perturbent leur équilibre écologique (Liu et Chen, 2008). Le maintien de la
durabilité des écosystèmes agricoles et l’optimisation de leur productivité nécessitent une
meilleure compréhension du fonctionnement du cycle biogéochimique du P.
Dans les sols cultivés, le P se trouve sous des formes inorganiques et organiques. Le
P inorganique existe sous forme d’ions phosphates libres en solution, adsorbés ou précipités
dans les minéraux apatites. Les plantes prélèvent le P essentiellement dans la solution du
sol. En absence de tout apport de P sous forme d’engrais, le réapprovisionnement de la
solution du sol se fait par dissolution des minéraux apatites et/ou par minéralisation du P
2
organique. Beaucoup d’études ont révélé l’importance de la contribution du P organique
dans la nutrition des plantes (Firsching et Claassen, 1996; Oehl et al., 2001; Chen et al.,
2002). Le changement de P de la forme non disponible à court terme à la forme disponible
dépend de la nature chimique des formes du P dans le sol et de leur concentration.
La dynamique des formes de P et leur distribution dans le sol sont contrôlées par la
pédogenèse (Tiessen et al., 1984) et les pratiques agricoles (Ross et al., 1999, Negassa et
Leinweber, 2009). Par exemple, le P s’accumule sous la forme d’ions phosphates dans la
couche superficielle des sols sous semis direct et augmente le risque de transport de P par
ruissellement vers les eaux de surface (Sharpley et Smith, 1994). Dans les sols labourés, le
P associé aux particules du sol peut être facilement transporté par érosion vers les cours
d’eau. Selon la texture du sol et les conditions climatiques, le P peut aussi être transporté de
la surface vers les couches de profondeur par écoulement préférentiel (Simard et al., 2000).
L’étude de la dynamique des formes de P et leur distribution dans le sol peut aussi
nous permettre d’améliorer notre compréhension de la contribution du P du sol à la
nutrition minérale des plantes. En effet, il a été démontré dans des études faites au Québec
que l’apport de P sous forme d’engrais ne permettait pas d’augmenter le rendement en grain
du soja (Glycine max L.) ou du maïs (Zea mays L.) dans des sols sous soumis direct ou
labour conventionnel (Messiga et al., 2010) et le rendement de la fléole des prés (Phleum
pratense L.; Bélanger et Ziadi, 2008). D’autres études ont rapporté que les apports continus
des engrais phosphatés ou des amendements organiques augmentaient la fraction du P
organique disponible à long terme (Shi et al., 2013; Zhang et MacKenzie, 1997).
Cependant, la mise en culture des sols sans apports de P peut diminuer leurs teneurs en P
organique (Tiessen et al., 1982).
La dynamique des formes de P et leur distribution dans les sols ont jusqu’ici été
étudiées par la méthode d’extraction séquentielle de Hedley et al. (1982). L’une des
contraintes de cette méthode est la caractérisation opérationnelle des fractions de P suivant
leur solubilité à des extractifs chimiques. Une autre est son analyse statistique étant donné
la contrainte à 100% des fractions de P obtenues. La majorité des études utilisant cette
méthode présentait des quantités brutes des pools de P sans ne leur attribuer aucun sens
biologique, géochimique ou environnemental. L’utilisation du « pathways analysis » a
3
permis d’établir des relations entre les différents pools de P et notamment l’effet des
pratiques agricoles sur celles-ci (Tiessen et al., 1984; Zheng et al., 2002). Cependant, le
« pathways analysis » repose essentiellement sur des coefficients de corrélations de Pearson
qui peuvent varier en fonction du temps dans un système dynamique. De plus, de fausses
corrélations se produisent en raison de la redondance d’information dans un système clos à
100% et de la dépendance d’échelle de mesure (pourcentage par rapport au P total ou à la
matière sèche du sol) (Aitchison, 1986). Il devient alors difficile d’interpréter sans biais les
transformations du P dans les sols à diverses échelles de temps sans tenir compte de la
contrainte à 100% des données compositionnelles.
L’application de nouvelles méthodes d’analyses telle que l’analyse des données
compositionnelles utilisant les coordonnées du log ratio isométrique (ilr) calculées pour des
répartitions binaires des formes de P (Egozcue et al., 2003) peut complémenter les
connaissances obtenues sur les sols à l’aide du « pathways analysis » en décrivant les
interactions entre les différentes fractions du P. L’analyse des données compositionnelles
est un domaine récent des mathématiques appliquées consacré à l’analyse des données
strictement positives comprises entre zéro et une quelconque unité ou échelle de mesure.
Dans un essai de longue durée sur un loam sableux dans la province canadienne de
l’Ile du Prince-Édwouard, il a été démontré à l’aide de la résonance magnétique nucléaire
au 31P (RMN-31P) que la plupart des formes organiques de P ne contribuaient pas à
l’accumulation de P dans la couche superficielle sous le semis direct parce qu’elles sont
plus facilement emportées vers les couches de profondeur par lixiviation (Cade-Menun et
al., 2010). La RMN-31P a l’avantage d’identifier aussi bien les formes inorganiques
qu’organiques. L’utilisation de la RMN couplée aux analyses chimiques (Mehlich, 1984;
Hedley et al., 1982) peut permettre d’améliorer notre compréhension du cycle
biogéochimique de P dans les sols du Québec.
La quantification du P du sol et l’estimation de sa disponibilité aux plantes est un
important aspect de la réussite de systèmes de cultures à la fois productifs et sains pour
l’environnement. La majorité des méthodes d’analyse du P (digestion acide, ignition,
Olsen, Bray 1, Bray 2 et Mehlich-3) sont relativement lentes et utilisent des extractifs
chimiques. Le développement de nouvelles techniques ou outils d’analyse du sol visant à
4
maximiser la gestion des éléments n’a cessé de progresser, et ce, pour des raisons
économiques et environnementales. L’utilisation de la spectroscopie dans le proche
infrarouge (SPIR) pour prédire la composition chimique des sols est de plus en plus
considérée. Les récentes études effectuées au Québec sur la SPIR (Nduwamungu et al.,
2009a) ont démontré l’efficacité de cette technique pour, notamment, prédire la texture,
l’azote potentiellement minéralisable et le carbone du sol. Nduwamungu et al. (2009b) ont
aussi réussi à prédire les concentrations en Ca et Mg extraits au Mehlich-3 en utilisant la
SPIR.
Ce projet de recherche a eu pour objectif général de mesurer les formes de P et de
caractériser leurs changements dans les sols sous l’effet de différentes pratiques culturales
(rotation, fertilisation minérale phosphatée, travail du sol), en utilisant de nouvelles
méthodes spectroscopiques (SPIR, RMN-31P) et mathématique (analyse compositionnelle),
tel que présenté dans la Figure 1.1. Une brève revue de littérature sur les connaissances
actuelles sur les méthodes de mesure et les transformations de différentes formes du P dans
les sols agricoles sera présentée en premier lieu, suivie par les hypothèses et les objectifs de
cette étude.
Figure 1-1. Schéma général du projet de recherche
5
CHAPITRE II: REVUE DE LA LITTÉRATURE
2.1 Cycle biogéochimique du phosphore dans le sol
La dynamique du phosphore (P) dans les écosystèmes cultivés est représentée par
un cycle biogéochimique (Fig. 2.1) qui intègre les réserves et les mécanismes de
transformation du P dans le sol, et les différents flux impliqués dans le transfert du P vers
les compartiments de l’écosystème.
Le P peut être ajouté au sol sous forme d’engrais minéraux ou de ferme (fumier,
lisiers), d’amendements organiques (composts, boues d’épuration, biosolides papetiers), ou
de résidus de cultures. Dans le sol, le P existe sous différentes formes qui interagissent via
différents mécanismes physico-chimiques, biologiques et biochimiques impliquant des
réactions d’adsorption et de désorption, de précipitation et de dissolution, de minéralisation
et d’immobilisation (Fig. 2.1). Le cycle biogéochimique du P inclut aussi des flux de P sous
forme de prélèvements par les plantes, et des pertes par érosion, ruissellement de surface et
de profondeur, lessivage et drainage (Kleinman et al., 2009; Haygarth et al, 1998).
Figure 2-1 Cycle du P dans le sol (adapté du Pierzynski et al., 2005).
P adsorbé Argile, Oxydes d’Al, Fe Carbonates de Ca
Minéraux secondaires Phosphates de Fe, Al et Ca
Minéraux primaires Apatites
P dans la solution du sol (H 2 PO 4
- , HPO 4
2 - )
P organique Biomasse microbienne Matière organique P organique dissous
Résidus végétaux
Lessivage
Eaux de surface
Érosion, ruissellement
Dissolution
Immobilisation
Minéralisation
Adsorption
Désorption
Précipitation
Dissolution
Prélèvements par la plante
Engrais organiques Engrais minéraux
6
2.2 Les formes du phosphore dans les sols
Dans la couche superficielle des sols (0-15 cm), le P total (PT) varie entre 50 et 3000 mg
kg-1 dépendamment de la roche mère, du type de sol, de la nature de la végétation et de
l’aménagement du sol (Sims et Pierzynski, 2005). Dans les sols minéraux, 35 à 70% du PT
se trouve sous forme inorganique et le reste est sous forme organique. Le P organique peut
présenter jusqu’à 90% du PT dans les sols organiques (Harrison, 1987).
2.2.1 Phosphore inorganique
Le phosphore inorganique (Pi) existe en plus fortes proportions dans les sols cultivés que
dans les sols de prairies ou forestiers. À titre d’illustration, il constitue dans les sols de
grandes cultures environ 75% du P total (Morel, 2002). Dans la solution du sol, il est
présent sous forme d’ions orthophosphates libres dont la nature et la proportion dépendent
du pH du sol (Fig. 2.2). Pour un pH entre 4.0 et 9.0 pour la majorité des sols, les formes
dominantes sont les orthophosphates H2PO4- et HPO4
2- (Pierzynski et al., 2005).
Dans les sols agricoles, la concentration des ions du P en solution est de l’ordre de 0.01 à
3.0 mg P L-1 (Frossard et al., 2000). Les transferts de ces ions à l’interface sol-solution sont
régulés par différents mécanismes chimiques, physiques et biologiques dont l’adsorption-
désorption, la précipitation-dissolution, la minéralisation-immobilisation, la diffusion intra-
particulaire et plusieurs autres réactions (Holtan et al., 1988; Frossard et al., 2000;
McGechan et Lewis, 2002; Oberson et Joner, 2005). Ces divers processus qui conditionnent
la mobilité du P, sont responsables du maintien de l’équilibre de la concentration des ions
phosphates dans la solution du sol, et par la suite, de leur disponibilité aux plantes et aux
microorganismes (Hinsinger, 2001; Hinsinger et al., 2007). En effet, suite au prélèvement
des ions P par les plantes, la concentration du P soluble diminue dans la solution du sol et le
P adsorbé diffuse vers la solution. La dissolution du P de la phase solide et la minéralisation
du Po par les microorganismes constituent aussi des sources importantes de
réapprovisionnement de la solution du sol en ions P (Hinsinger, 2001). Cependant, les ions
orthophosphates peuvent être aussi immobilisés par la biomasse microbienne du sol qui
représente alors un compétiteur pour les racines de la plante.
7
À l’interface solide-solution du sol, les ions phosphates se trouvent adsorbés à la surface
des minéraux argileux, du fer et d’aluminium, des cations échangeables et du carbonate de
calcium (Holford, 1997). Dans les sols acides, les constituants les plus responsables de la
fixation du P sont les oxydes et les hydroxydes de fer et d’aluminium. Tandis que dans les
sols neutres à alcalin, les facteurs les plus impliqués sont le carbonate de calcium et le
magnésium de façon moins importante (Holford, 1997; Hinsinger, 2001). A l’état solide, le
Pi précipite avec les métaux du sol tels que l’Al, le Fe et le Ca en formant des minéraux
phosphatés plus ou moins cristallisés (Hinsinger, 2001). Il est à noter que seulement 10 à
20% du P ajouté sous forme d’engrais est prélevé par la plante (Richardson, 2001;
McLaughlin et al., 1988) puisqu’une grande partie de la dose appliquée évolue rapidement
vers des formes fixées (Khiari et Parent, 2005).
Figure 2-2 Effet du pH sur la forme des ions orthophosphates de la solution du sol (Holtan
et al., 1988).
2.2.2 Phosphore organique
Le phosphore organique (Po) est défini comme étant le P trouvé dans des composés
organiques en liaison avec le carbone. Dans les sols sous prairie et forêts, le P organique
représente 41 à 88% du P total (Bowman et Cole, 1978; Ross et al., 1999; Motavalli et
Miles, 2002; Chen et al., 2003). Dans les sols de grandes cultures, cultivés pendant
plusieurs décennies, il ne représente que 20% du PT en moyenne (Morel, 2002).
8
Le P organique se trouve dans les végétaux, les microorganismes et la matière
organique du sol (Harrison, 1982; Stewart et Tiessen, 1987). Il est synthétisé par les plantes
et les microorganismes via des processus biochimiques à partir des ions orthophosphates
absorbés de la solution du sol (Condron et al., 2005). Outre les résidus de cultures et les
amendements organiques de ferme (fumier, compost) et des industries (biosolides), le Po
peut être introduit aux sols cultivés sous forme synthétisée d’insecticides, fongicides ou
herbicides tel que le phosphonate glyphosate (Condron et al., 2005). Les stocks en Po dans
le sol constituent une source importante du P disponible aux plantes et aux
microorganismes via des processus biotiques (Quiquampoix et Mousain, 2005) et
abiotiques (Baldwin et al., 2005). Le taux de libération des ions phosphates à travers ces
mécanismes de dégradation dépend de la nature chimique des composés organiques
(Quiquampoix et Mousain, 2005).
Selon la nature des liaisons, les composés organiques phosphatés sont classés en
esters phosphate (monoester et diester), en phosphonates (composés avec liaison directe
carbone-phosphore), et en polyphosphates organiques (Mckelvie, 2005; Condron et al.,
2005; Turner et al., 2005). Quelques exemples de composés sont présentés dans le tableau
2.1. Les orthophosphates monoesters sont caractérisés par une seule liaison avec un radical
organique et représentent la forme la plus abondante du Po dans le sol (>90%, Turner et al.,
2003). Cependant, les orthophosphates diesters sont liés à deux radicaux et sont beaucoup
moins présents que les monoesters dans les sols agricoles (<10%, Condron et al., 2005).
Les principales formes de phosphate d’esters rencontrés dans les sols sont les inositols
phosphates (60%), les acides nucléiques (5-10%), et les phospholipides (1%) (Halstead et
McKercher, 1975). Malgré l’abondance de ces formes organiques dans les sols, leur
dynamique et leur devenir dans l’écosystème sont beaucoup moins étudiés par rapport au P
inorganique.
L’adsorption des composés organiques phosphatés aux minéraux du sol dépend de
leur nature et des propriétés édaphiques (Condron et al., 2005). Les inositols phosphates et
les phosphonates sont fortement retenus dans le sol et constituent, par conséquence, la
forme stable du Po. Par contre, les orthophosphates monoesters simples (avec un seul
groupe de phosphate), tels que les glucides phosphate et les mononucléotides, les
9
orthophosphates diesters et les polyphosphates organiques sont moins retenus et peuvent
être dégradés rapidement par les enzymes de type phosphatase (Celi et al., 1999).
Cependant, l’acide désoxyribonucléique (ADN) peut persister à la biodégradation lorsqu’il
est lié à l’argile, au sable ou aux acides humiques (Khanna et al., 1998). La rétention de
l’ADN est inversement proportionnelle au pH du sol; elle augmente particulièrement à un
pH inférieur à 5, qui correspond à son point isoélectrique (Khanna et al., 1998; Condron et
al., 2005).
Tableau 2-1 Composés organiques phosphatés communs du sol (Turner et al., 2005).
10
2.3 Mesures du phosphore du sol
2.3.1 Méthodes conventionnelles
2.3.1.1 Phosphore total
La caractérisation du phosphore total nécessite la solubilisation de toutes les formes
inorganiques et organiques du P dans le sol. Deux anciennes méthodes ont été largement
appliquées; soient, la fusion avec le carbonate de sodium (NaCO3), et la digestion avec
l’acide perchlorique (HClO4) (Olsen et Sommers, 1982). Cependant, ces méthodes ne sont
plus utilisées étant donné que la première était fastidieuse et inappropriée pour un grand
nombre d’échantillons de sol, et la deuxième présentait un risque potentiel d’explosion dû à
la réaction de l’acide perchlorique avec les composés organiques (O’Halloran et Cade-
Menun, 2007). Actuellement, trois autres méthodes sont adoptées par de nombreux
laboratoires. La première consiste à l’oxydation de l’échantillon du sol avec l’hypobromite
de sodium (NaOBr) et l’hydroxyde de sodium (NaOH) (Dick et Tabatabai, 1977). Les deux
autres méthodes consistent à une digestion acide humide avec une solution d’acide
sulfurique (H2SO4), de peroxyde d’hydrogène (H2O2), et (1) d’acide fluoridrique (HF)
(Bowman, 1988); ou (2) de sulfate de lithium (Li2SO4) et de sélénium (Se) (Parkinson et
Allen, 1975). La détermination du PT est alors jusqu’à date relativement lente et implique
l’utilisation de plusieurs produits chimiques.
2.3.1.2 Phosphore disponible aux plantes
La forme du P inorganique disponible aux plantes peut être mesurée par de
nombreuses méthodes dont les plus utilisées pour les analyses routinières sont les méthodes
d’extractions chimiques et à l’eau (Tableau 2.2). Le type de l’extractif utilisé dépend
fortement de la nature du sol. Pour les sols acides, les méthodes recommandées sont celles
qui reposent sur des solutions acides telles que les méthodes de Bray 1 et Bray 2 qui
permettent d’extraire principalement le P lié à l’aluminium et au fer (Bray et Kurtz, 1945).
Pour les sols calcaires, les solutions alcalines sont les mieux appropriées comme la
méthode Olsen utilisant le bicarbonate de sodium (Olsen et al., 1954). La méthode de
Mehlich 3 (Mehlich, 1984), est bien adaptée pour une large gamme de sols acides, neutres
et légèrement calcaires. C’est la méthode de référence au Québec. L’extraction à l’eau est
11
aussi utilisée comme une méthode commune aux sols minéraux et organiques avec des
rapports sol/eau allant de 1/5 à 1/10 et 1/60 (Morel, 2002; Koopmans et al., 2001; Fardeau,
1996; Sissingh, 1971). Le P extrait par ces différentes méthodes est dosé par la suite soit
par colorimétrie, par spectrophotométrie d’émission au plasma ou par spectrométrie au
plasma à couplage inductif (Kuo, 1996).
Tableau 2-2 Méthodes d’extraction du P (tiré et adapté du Ziadi et al., 2013).
Méthode Extractif
pH de la
solution
extractive
Volume :
masse de
sol
pH du sol Durée
d’extraction Référence
Bray 1 0.5N HCl +
1N NH4F 3.0 7:1
<6.0; 6.0
à 7.2 1 mn
Bray et
Kurtz, 1945
Mehlich 3
0.015N NH4F
+ 0.025N
NH4NO3 +
0.2N
CH3COOH +
0.013N HNO3
+ 0.001N
EDTA
2.3 10:1 <6.0; 6.0
à 7.2 5 mn
Mehlich,
1984
Olsen 0.5M NaHCO3 8.5 20:1
<6.0; 6.0
à 7.2;
>7.2
30 mn Olsen et al.,
1954
Eau 7
10 :1
60 :1
<6.0; 6.0
à 7.2 24 h
Morel, 2002;
Sissingh,
1971
2.3.1.3 Phosphore organique
La détermination du Po total du sol est indirecte. Elle consiste à mesurer
l’augmentation du Pi dans l’extrait d’un échantillon de sol soumis à une combustion ou une
digestion par rapport à un échantillon témoin. La méthode d’ignition se fait par une
oxydation du Po en Pi à des faibles températures (250°C, Legg et Black, 1955) ou à des
températures élevées (550°C, Saunders et Williams, 1955), suivie d’une extraction acide du
PT. La digestion suit une série d’extractions acides et/ou basiques (Bowman et Moir, 1993;
1980; Stewart and Oades, 1972; Mehta et al., 1954) du Po, et permet de déterminer le PT.
Le Pi est mesuré dans l’extrait du sol en utilisant l’acide sulfurique (H2SO4; Anderson,
1960). La concentration du Po peut être surestimée par une solubilisation du Pi suite à la
12
combustion de l’échantillon du sol (Williams et al., 1970), ou sous-estimée suite à une
oxydation ou une extraction incomplète (O’Halloran et Cade-Menun, 2007). Par
conséquent, la méthode de la combustion est adoptée pour des traitements avec le même
type de sol, alors que les techniques d’extraction sont recommandées pour comparer les
niveaux du Po dans différents types de sol (Bowman, 1989).
2.3.1.4 Pools du P
Dans le sol, le P est réparti en différents pools selon leur disponibilité aux plantes
(Cross et Schlesinger, 1995). Ces pools peuvent être déterminés par une méthode
d’extraction séquentielle utilisant des solutions chimiques de plus en plus fortes (Pierzynski
et al., 2005). La méthode de Chang et Jakson (1957) fut la première développée dans ce
contexte selon la séquence suivante : NH4Cl, NH4F, NaOH, H2SO4 et NH4F ou NaOH,
pour extraire respectivement le Pi disponible, le Pi associé à l’aluminium, au fer, au
calcium et le P fortement retenu. La fraction du P résiduel est obtenue par différence entre
le P total déterminé par digestion et la somme de ces cinq fractions (Chang et Jakson, 1957;
Pierzynski et al., 2005). Des modifications ont été rapportées à cette méthode (Peterson et
Corey, 1966; Smith, 1965; Williams et al., 1967) dans le but de corriger son inefficacité
d’extraction du P lié au fer et de l’adapter aux sols calcaires ainsi qu’aux sédiments.
Afin de faire ressortir la contribution du phosphore organique à la dynamique du P
dans le sol, une méthode de fractionnement développée par Hedley et al. (1982; Fig. 2.3) a
été largement utilisée (Cross et Schlesinger, 1995; Zheng et al., 2003; Agbenin et Tiessen,
1994; Frossard et al., 1989). Le pool inorganique labile, correspondant au Pi adsorbé aux
surfaces des composés phosphatés plus cristallins : sesquioxydes ou carbonates (Mattingly,
1975), est extrait par la résine échangeuse d’anions et par le NaHCO3. Cependant, la
fraction du Pi la moins biodisponible associée aux oxyhydroxydes de fer et d’aluminium
amorphes et cristallins, est extraite avec le NaOH. Le Pi extractible au HCl est lié au
calcium (Tiessen et Moir, 2007). Le Po facilement minéralisable est extrait au NaHCO3 et
constitue le pool labile du Po, tandis que le Po extrait au NaOH est plus stable (Bowman et
Cole, 1978). Le P total résiduel de cette extraction séquentielle, déterminé par digestion à
l’aide de H2SO4 et H2O2, regroupe le Pi occlus aux minéraux de sol et le Po non extractible
(Tiessen et al., 1984). Toutefois, la fraction du Po active dans le cycle de transformation du
13
P dans le sol à court terme est sous-estimée par cette méthode, d’où la nécessité de
l’utilisation de l’extractif HCl concentré dans la nouvelle méthode de fractionnement de
Tiessen et Moir (2007) pour surmonter ce problème. Néanmoins, ces méthodes de
fractionnement sont opérationnelles et ne déterminent pas les formes spécifiques des pools
inorganique et organique du P (Condron et Newman, 2011).
Figure 2-3 Méthode d’extraction séquentielle du P du sol selon la méthode de Hedley
et al., 1982 (tirée de Cross et Schlesinger, 1995).
14
2.3.2 Méthodes spectroscopiques
2.3.2.1 Spectroscopie dans le proche infrarouge
La spectroscopie dans le proche infrarouge (SPIR) est une technique analytique
indirecte qui permet d’estimer la composition de la matière à partir de ses propriétés
d’absorption de la lumière (Stenberg et al., 2010). Cette technique consiste à soumettre
l’échantillon aux rayonnements dans la gamme du proche infrarouge allant de 800 à 2500
nm et à mesurer la reflectance (R) de la lumière par des détecteurs à chaque longueur
d’onde et la convertir par la suite en absorbance (log (1/R)).
Les rayonnements infrarouges sont absorbés par les liaisons chimiques entre les atomes
de la matière (C-H, O-H, N-H, C-O, S-H, CH2, et C-C), causant des mouvements de
torsion, de flexion ou d’étirement (Ludwig and Khanna, 2000). Les mesures de
l’absorbance sont utilisées pour calibrer la SPIR avec des mesures de référence obtenues au
laboratoire via différentes méthodes de régression telle que la régression linéaire multiple et
la régression de moindre carrée partielle. Le modèle de calibration obtenu est utilisé par la
suite pour prédire la composition de l’échantillon à partir de son spectre d’absorbance
(Martens et Naes, 2001). La performance de la prédiction de la SPIR est évaluée en
utilisant des paramètres statistiques tels que le coefficient de détermination et le rapport de
déviation de la performance (Nduwamungu et al., 2009a).
La SPIR est une technique rapide, peu coûteuse, non destructive et permet d’analyser
plusieurs propriétés de l’échantillon à partir du même spectre et de réduire les coûts
d’analyse de l’ordre d’au moins 50%. Elle trouve des applications dans plusieurs
domaines : la biologie, la chimie, la médecine et les industries pharmaceutique et
agroalimentaire. Son utilisation dans les sciences du sol est de plus en plus considérée pour
une évaluation plus rapide et plus précise de la qualité du sol (Guerrero et al., 2010). De
récentes études ont démontré que la SPIR est efficace pour, notamment, prédire la matière
organique, la texture, le carbone, l’azote, la capacité d’échange cationiqueet le pH du sol
(Cozzolino et Morón, 2006; Brunet et al., 2007; Nduwamungu et al., 2009a), de même que
les concentrations en Ca et Mg extraits au Mehlich-3 (Chang et al., 2001; Nduwamungu et
al., 2009b).
L’évaluation de l’efficacité du potentiel de la SPIR dans la prédiction de la teneur en P
des sols a fait aussi l’objet de plusieurs études dont les résultats trouvés divergent. À titre
15
d’illustration, Chang et al. (2001), Nduwamungu et al. (2009b), McCarty et Reeves (2006)
et Ludwig et al. (2002) ont montré que la SPIR ne pouvait prédire le P-Mehlich 3, le P-
Mehlich 1 et le P-Bray 2, respectivement. Cependant, Maleki et al. (2006) et van Groenigen
et al. (2003) ont trouvé des résultats de prédiction acceptables pour le P-Olsen. Cette
différence de résultats peut être expliquée par la nature des extractifs chimiques utilisés
(Guerrero et al., 2010). D’après Chang et al. (2001), le P disponible peut être prédit par la
SPIR s’il est relié aux propriétés primaires du sol telles que la matière organique et la
texture. La performance de la SPIR peut être aussi affectée par l’hétérogenéité de
l’échantillon du sol, de la texture du sol et par la variation dans les méthodes de préparation
de l’échantillon (Nduwamungu et al., 2009a).
2.3.2.2 Spectroscopie de résonance magnétique nucléaire du 31P
La spectroscopie de résonance magnétique nucléaire du 31P (RMN-31P) est une
technique qui utilise la résonance magnétique du noyau du P dans l’échantillon pour
identifier et quantifier sa forme chimique (Cade-Menun et al., 2005). En effet, le noyau du
P émet de l’énergie après avoir été soumis à des impulsions de radiofréquences dans un
champ magnétique; les quelles sont détectées sous forme de signal dans un spectre et
transformée par la suite en pics relatifs à chaque espèce ionique ou moléculaire du P (Cade-
Menun et al., 2005). L’avantage de cette technique est qu’elle permet de caractériser
simultanément les espèces du P, extraites au préalable du sol par une une solution alcaline
(NaOH-EDTA; He et al., 2008).
Newman et Tate (1980) furent les premiers à utiliser la RMN pour caractériser le P dans
les couches superficielles des sols de prairies en Nouvelle-Zélande. Beaucoup de travaux de
recherche l’ont par la suite utilisée pour caractériser le Po dans les sols agricoles (Hawkes
et al. 1984, Leinweber et al. 1997; Smernik et Dougherty, 2007; Redel al., 2011 et Cade-
Menun et Liu, 2014), les écosystèmes aquatiques (Carman et al., 2000; Paytan et al., 2003)
ou les amendements organiques (Jing et al., 1992; Crousse et al., 2002). L’ensemble de ces
études ont permis d’améliorer les connaissances sur l’origine, la distribution, la
biodisponibilité et la dynamique des formes chimiques du P dans l’environnement, ce qui
permet de mieux gérer le P dans les écosystèmes cultivés. A titre d’illustration, la figure 2.4
16
montre des espèces inorganiques et organiques du P caractérisées par la RMN-31P dans la
couche superficielle d’un sol cultivé non labouré à l’Île du Prince Edward, Canada (Cade-
Menun et al., 2010).
Figure 2-4 Spectre obtenu par la spectroscopie magnétique nucléaire du 31P montrant
les composés phosphatés détectés dans la couche superficielle (5-10 cm) d’un sol
cultivé non labouré (Cade-Menun et al., 2010).
2.4 Changements des formes du phosphore selon les pratiques culturales
2.4.1 Changements des pools du P
L’effet des pratiques culturales sur les changements dans les pools de P a fait l’objet de
plusieurs travaux de recherche dans le but d’améliorer la gestion du P dans les systèmes
agricoles. A titre d’exemple, il a été démontré que l’apport des fertilisants phosphatés au
sol durant une longue durée engendrait une augmentation dans les fractions labiles du Pi
(Pi-résine, Pi-NaHCO3, Pi-NaOH) significativement plus importante en système de culture
continue qu’en rotation (McKenzie et al,. 1992a; McKenzie et al., 1992b). Ce P accumulé
risque d’atteindre les cours d’eau adjacents constituant ainsi un problème environnemental
17
majeur (Leinweber et al., 1999). Des résultats similaires conformant la forte corrélation
entre l’apport du P et l’augmentation des teneurs des pools de P labiles ont été observés par
Tran et N’Dayegamie (1995), Dobermann et al. (2002) et Redel et al. (2007) dans des
Inceptisols, des Oxisols et des Ultisols, respectivement. Zheng et al. (2002) ont montré que
l’apport d’une source organique de P en sol cultivé durant huit ans favorisait aussi
l’accumulation de la fraction disponible du Pi provenant de la minéralisation du Po ajouté.
En revanche, la non fertilisation phosphatée menait à l’épuisement des réservoirs en P
labile des sols (Agbenin et Goladi, 1998; McKenzie et al., 1992a). D’autre part,
l’application des sous-produits industriels tels que les biosolides papetiers en absence de la
fertilisation phosphatée a favorisé la mobilité et/ou la minéralisation du Po-NaOH et la
transformation du pool récalcitrant au cours du temps en forme labile de P (Fan et al.,
2010).
Les rotations culturales affectent aussi les teneurs et la distribution des pools de P à
différents degrés selon les cultures. En effet, Redel et al. (2007) ont observé une
accumulation du P total et de P non labile dans le sol après une culture de blé,
contrairement à la culture d’avoine qui a réduit la fraction du P non labile et a augmenté le
P relativement labile. De leur côté, Zheng et al. (2001) ont montré qu’une monoculture
d’orge combinée aux fertilisants minéraux réduisait les formes labiles du Po en faveur des
formes labiles du Pi. D’autre part, ils ont démontré que la rotation orge-fourrages en un
gleysol labouré et recevant du lisier conduisait à une augmentation en P labile plus
importante qu’avec les fertilisants minéraux. Magid (1993) a observé une augmentation des
fractions inorganiques labiles du P (Pi-résine et Pi-NaHCO3) sous une végétation d’hêtres,
alors que les formes organiques labiles étaient les plus favorisées sous un système cultural
sous prairie.
2.4.2 Changements des espèces de P
Le statut du Po dans le sol est fortement influencé par la matière organique. En effet,
Vincent et al. (2010) ont montré qu’un apport de litière à un sol minéral d’une forêt tropical
durant trois ans à une dose de 6 kg de P/ha/année a augmenté la teneur en Po de 16% dont
31% est sous forme d’ADN. Cependant, l’enlèvement de la litière durant 3 ans a réduit de
4,2 ± 1,6 kg ha-1 la concentration du Po à la surface du sol (0- 2 cm) incluant 0,84 ± 0,32 kg
18
ha-1 des phosphates monoesters et 1,26 ± 0.48 kg ha-1 d’ADN. À plus long terme, He et al.
(2008) ont trouvé que l’apport de litière de volaille à un sol argilo-loameux n’avait pas
changé les teneurs des fractions hydrolysables du Po (NaOH, HCl) alors qu’il avait
augmenté celles des pools labiles et stables du Pi; ce qui indique que les formes du P
organiques provenant de la litière se transformaient en Pi dans le sol. De leur côté, Cade-
Menun et al. (2010) ont observé que le labour du sol modifie la distribution des espèces du
P par rapport à un sol non labouré en augmentant les concentrations en ions
orthophosphates et en phytate dans la couche de 5 à 10 cm. D’autre part, Condron et al.
(1985) rapportent que la fertilisation phosphatée d’un sol sablo-loameux favorise à long
terme l’accumulation des orthophosphates monoesters par rapport aux autres espèces
jusqu’au 99% du Po.
2.5 Concept d’analyse des données compositionnelles
Les données compositionnelles sont définies comme étant des données strictement
positives d’un ensemble clos, de somme constante égale à 1 ou à 100%, ou à une unité de
mesure (e.g., mg kg-1, kg m-3, etc.), qui transmettent des informations relatives (Aitchison,
1986). Cet ensemble nommé simplexe, est défini comme suit avec un degré D représentant
le nombre de ses composantes (X) :
SD = [(X1,…, XD), X1 > 0,…, XD > 0, X1 + … + XD =1]
Les données compositionnelles sont multivariées et dépendantes les unes des autres; ce
qui signifie que toute augmentation d’une composante est accompagnée d’une réduction
d’au moins une autre composante indiquant une corrélation fausse et négative. Par
conséquent, chaque composante ne peut pas être interprétée indépendamment des autres
(Tolosana-Delgado et van den Boogart, 2011). Les données compositionnelles sont
caractérisées aussi par une distribution non-normale, une redondance de l’information au
sein du simplexe, et une dépendance de leurs résultats d’analyse à l’échelle de mesure
(Aitchison, 1986). Ces biais intrinsèques peuvent générer des résultats erronés menant à des
interprétations contradictoires suite à l’application des analyses statistiques classiques
(Pawlowsky-Glahn et Egozcue, 2006; Filzmoser et al., 2009). Pour corriger ces problèmes,
Aitchison (1986) a proposé d’utiliser le log ratio additif (alr, Éq. 1) et le log ratio centré
19
(clr, Éq. 2) pour projeter les données compositionnelles de leur espace fermé à l’espace
réel, et de créer de nouvelles variables indépendantes de l’échelle de mesure pour que
l’interprétation soit cohérente :
Éq. 1
Éq. 2
Où est une composante quelconque, est une composante sélectionnée comme
dénominateur commun et ) est la moyenne géométrique de toutes les composantes.
Egozcue et al. (2003), ont développé plus tard une transformation du log ratio
isometrique (ilr, Éq. 3) qui permet de corriger la redandance en générant D-1 variables
orthogonales (90°) à partir de D composantes sans perdre de l’information. Cette
transformation permet d’analyser les composantes du simplexe en termes de D-1 balances
entre deux composantes ou deux groupes de composantes avec des signes + et – comme
suit :
Éq. 3
Où r est le nombre de composantes dans le groupe +, s est le nombre de composantes
dans le groupe –, est la moyenne géométrique des composantes dans le groupe et
est la moyenne géométrique des composantes dans le groupe .Les balances sont
conçues selon une base théorique spécifique à chaque système à l’étude (Egozcue et al.,
2003; Egozcue et Pawlowsky-Glahn, 2005). La variation du système d’un état initial x à un
état final y, peut être mesurée par le concept de la distance d’Aitchison (A, Éq. 4; Aitchison
et Egozcue 2005) suivant:
et Éq. 4
L’analyse des données compositionnelles a été largement utilisée en géochimie
(Buccianti et Pawlowsky-Glahn, 2005; Borgheresi et al., 2013; Engle et Rowan, 2013; Zuo
et al., 2013; Buccianti et Grunsky, 2014) et en sciences des aliments (Korhoňová et al.,
2009; Hron et al., 2012; Veverka et al., 2012). En agronomie, des études récentes ont été
faites sur le diagnostic nutritif des tissus végétaux (Parent et al., 2012a; Parent et al., 2013a;
20
Parent et al., 2013b), la mesure de l’agrégation dans les sols (Parent et al., 2012b), la
décomposition des résidus organiques (Parent et al., 2011), et la dynamique des pesticides
dans le sol (Aslam et al., 2013).
L’analyse compositionnelle est tout à fait adaptée aux formes du P (fractions ou espèces
chimiques) puisque ce sont des données fermées à 100% du P total. Dans une étude récente,
Parent et al. (2014) ont démontré que l’analyse classique des fractions de P est biaisée par
la dépendance de l’échelle de mesure, et ont élaboré une hiérarchie de balances entre ces
fractions pour évaluer le risque de perte du P dans les écosystèmes agricoles.
21
2.6 Hypothèses
De cette revue de littérature, il découle que de nouvelles méthodes pourraient être
utilisées pour mesurer les formes de P et mieux caractériser leurs changements dans les sols
agricoles. D’où, les hypothèses suivantes ont été formulées :
1. La spectroscopie dans le proche infrarouge peut déterminer la concentration du P
total et du P disponible extrait à la solution Mehlich-3 et à l’eau dans un sol sablo-
loameux.
2. La spectroscopie dans le proche infrarouge est une technique efficace pour prédire le
P organique du sol.
3. L’analyse compositionnelle des espèces de P évite le biais lié à la dépendance de
l’échelle de mesure, contrairement aux analyses statistiques classiques.
4. Le système du travail du sol et la fertilisation phosphatée changent les concentrations
et la distribution des espèces chimiques du P dans un gleysol sous rotation maïs-soya.
22
2.7 Objectifs
L’objectif général de cette thèse a été de mesurer et de caractériser les changements des
formes opérationnelles et fonctionnelles du P dans les sols selon différentes pratiques
culturales en utilisant des nouvelles techniques spectroscopiques et mathématiques. Les
objectifs spécifiques sont les suivants:
1. Evaluer le potentiel de la spectroscopie dans le proche infrarouge à prédire le P total
et le P disponible extrait à la solution Mehlich-3 et à l’eau dans un sol sableux
loameux.
2. Evaluer le potentiel de la spectroscopie dans le proche infrarouge à prédire le P
organique dans des sols loameux et argileux-loameux.
3. Démontrer que les analyses statistiques conventionnelles des formes de P sont
biaisées par la dépendance de l’échelle et que l’utilisation de l’analyse
compositionnelle permet d’éviter ce biais et d’avoir des résultats cohérents et fiables.
4. Etudier l’effet du système du travail du sol et de la fertilisation phosphatée sur les
teneurs et la distribution des espèces de P à l’aide de la résonance magnétique nucléaire
du 31P et l’analyse compositionnelle.
23
2.8 BIBLIOGRAPHIE
Aitchison, J. 1986. The statistical analysis of compositional data. Cox, D.R., Hinkley, D.V.,
Rubin D. and Silverman, B.W. Eds. 416 p.
Agbenin, J.O., and Goladi, J.T. 1998. Dynamics of phosphorus fractions in a savanna
Afisol under continuous cultivation. Soil Use Manag. 14: 59-64.
Agbenin, J.O., and Tiessen, H. 1994. Phosphorus transformations in a toposequence of
Lithosols and Cambisols from semi-arid northeastern Brazil. Geoderma, 62: 345-362.
Anderson, G. 1960. Factors affecting the estimation of phosphate esters in soil. J. Sci. Food
Agric. 11: 497–503.
Aslam, S., Garnier, P., Rumpel, C., Parent, S-É., and Benoit, P. 2013. Adsorption and
desorption behavior of selected pesticides as influenced by decomposition of maize
mulch. Chemosphere 91: 1447–1455.
Baldwin, D.S., Howitt, J.A., and Beattie, J.K. 2005. Abiotic degradation of organic
phosphorus compounds in the environment. In: B.L. Turner, E. Frossard, and D.S.
Baldwin, Eds, Organic phosphorus in the environment. CABI Publishing. Oxfordshire,
U.K. pp.75-88.
Bélanger, G. and Ziadi, N. 2008. Phosphorus and nitrogen relationships during spring
growth of an aging timothy sward. Agron. J. 100:1757-1762.
Borgheresi, M., Buccianti, A., Benedetto, F., and Vaughan, D., 2013. Application of
Compositional Techniques in the Field of Crystal Chemistry: A Case Study of
Luzonite, a Sn-Bearing Mineral. Math. Geosci. 45: 183-206.
Bowman, R.A. 1988. A rapid method to determine total phosphorus in soils. Soil Sci. Soc.
Am. J. 52: 1301–1304.
Bowman, R.A. 1989. A sequential extraction procedure with concentrated sulphuric acid
and dilute base for soil organic phosphorus. Soil Sci. Soc. Am. J. 53: 362–366.
Bowman, R.A. and Cole, C.V. 1978. An exploratory method for fractionation of organic
phosphorus from grassland soils. Soil Sci. 125: 95-101.
Bowman, R.A. and Moir, J.O. 1993. Basic EDTA as an extractant for soil organic
phosphorus. Soil Sci. Soc. Am. J. 57: 1513–1518.
Bray, R.H., and Kurtz, L.T. 1945. Determination of total, organic, and available forms of
phosphorus in soils. Soil Sci. 59: 39- 45.
Brunet, D., Barthès, B.G., Chotte, J.-L. and Feller, C. 2007. Determination of carbon and
nitrogen contents in Alfisols, Oxisols and Ultisols from Africa and Brazil using NIRS
analysis: Effects of sample grinding and set heterogeneity. Geoderma, 139:106–117.
Buccianti, A., and Grunsky, E. 2014. Compositional data analysis in geochemistry : Are we
sure to see what really occurs during natural processes ? J. Geochem. Expl. 141: 1-5.
Buccianti, A., and Pawlowsky-Glahn, V., 2005. New perspectives on water chemistry and
compositional data analysis. Math. Geol. 37: 703-727.
Cade-Menun, B.J. 2005. Using phosphorus-31 nuclear magnetic resonance spectroscopy to
characterize phosphorus in environmental samples. In: B.L. Turner et al., editors,
Organic phosphorus in the environment. CABI Publ., Wallingford, UK. p. 21–44.
Cade-Menun, B.J., Carter, M.R., James, D.C., and Liu, C.W. 2010. Phosphorus forms and
chemistry in the soil profile under long-term conservation tillage: A phosphorus-31
nuclear magnetic resonance study. J. Environ. Qual. 39: 1647-1656.
24
Cade-Menun, B.J., and Liu, C.W. 2014. Solution phosphorus-31 nuclear magnetic
resonance spectroscopy of soils from 2005 to 2013: A review of sample preparation
and experimental parameters. Soil Sci. Soc. Am. J. 78:19–37.
Carman, R., Edlund, G. and Damberg, C. 2000. Distribution of organic and inorganic
phosphorus compounds in marine and lacustrine sediments: a 31P NMR study. Chem.
Geol. 163: 101–114.
Cathcart, J.B. 1980. World phosphate reserves and resources. In Khasawneh et al. (ed.).
The rôle of phosphorus in agriculture. American Society of Agronomy, Madison,
Wisconsin, USA, pp. 1–18.
Chang, C.W., Laird, D.A., Mausbach, M.J., and Hurburgh, C.R.Jr. 2001. Near-infrared
reflectance spectroscopy – Principal components regression analyses of soil properties.
Soil Sci. Soc. Am. J. 65:480–490.
Chang, S.C. and Jakson, M.L. 1957. Fractionation of soil phosphorus. Soil Sci. 84: 133-
144.
Chen, C.R., Condron, L.M., Davis M.R., and Sherlock, R.R.. 2002. Phosphorus dynamics
in the rhizosphere of perennial ryegrass (Lolium perenne L.) and radiata pine (Pinus
radiata D. Don). Soil Biol. Biochem. 34:487–499.
Chen, C.R., Condron, L.M., Davis, M.R., and Sherlock, R.R. 2003. Seasonal changes in
soil phosphorus and associated microbial properties under adjacent grassland and forest
in New Zealand. F. Ecol. Manag. 177: 539-557.
Celi, L., Lamacchia, S., Marsan, T.A. and Barberis, E.1999. Interaction of inositol
phosphate on clays: Adsorption and charging phenomena. Soil Sci. 164:574–585.
Condron, L.M., Goh, L.M., and Newman, R.H. 1985. Nature and distribution of soil
phosphorus as revealed by a sequential extraction method followed by 31P nuclear
magnetic resonance analysis. J. Soil Sci. 36: 199-207.
Condron, L.M., and Newman, S. 2011. Revisiting the fundamentals of phosphorus
fractionation of sediments and soils. J. Soils Sed. 11:830–840.
Condron, L.M., Turner, B.L., and Cade-Menun, B.J. 2005. Chemistry and dynamics of soil
organic phosphorus. Phosphorus: agriculture and the environment. Sims, J.T. and
Sharpley, A.N. Eds. pp: 87-121.
Cordell, D., Drangert, J. O. and White. S. 2009. The story of phosphorus: Global food
security and food for thought. G. Environ. Change, 19: 292–305.
Cozzolino, D., and Morón, A. 2006. Potential of near-infrared reflectance spectroscopy and
chemometrics to predict soil organic carbon fractions. Soil Tillage Res. 85:78–85.
Cross, A.F, and Sclesinger, W.H. 1995. A literature review and evaluation of the Hedley
fractionation: Applications to the biogeochemical cycle of soil phosphorus in natural
ecosystems. Geoderma, 64: 197- 214.
Crouse, D.A., Sierzputowska-Gracz, H., Mikkelson, R.L. and Wollum, A.G. 2002
Monitoring phosphorus mineralization from poultry manure using phosphatase assays
and phosphorus-31 nuclear magnetic resonance spectroscopy. Comm. in Soil Sci. Plant
Anal. 33: 1205–1217.
Dick, W.A. and Tabatai, M.A. 1977. An alkaline oxidation method for determination of
total phosphorus in soils. Soil Sci. Soc. Am. J. 41: 511–514.
Dobermann, A., George, T., and Thevs, N. 2002. Phosphorus fertilizer effects on soil
phosphorus pools in acid upland soils. Soil Sci. Soc. Am. J., 66: 652-660.
Egozcue, J. J., and Pawlowsky-Glahn, V. 2005. Groups of Parts and Their Balances in
Compositional Data Analysis. Math. Geol. 37: 795-828.
25
Egozcue, J. J., Pawlowsky-Glahn, V., Mateu-Figueras, G., and Barcel´o-Vidal, G. 2003.
Isometric logratio transformations for compositional data analysis. Math. Geol. 35: 279-
300.
Engle, M.A., and Rowan, E.L., 2013. Interpretation of Na-Cl-Br Systematics in
Sedimentary Basin Brines: Comparison of Concentration, Element Ratio, and Isometric
Log-ratio Approaches. Math. Geosci. 45: 87-101.
Fan, J.L., Ziadi, N., Gagnon, B., and Hu, Z.Y. 2010. Soil phosphorus fractions following
annual paper mill biosolids and liming materials application. Can. J. Soil Sci. 90: 467-
479.
Fardeau, J. C. 1996. Dynamics of phosphate in soils. An isotopic outlook. Fert. Res. 45: 91-
100.
Filzmoser, P., Hron, K., and Reimann, C. 2009. Univariate statistical analysis of
environmental (compositional) data: Problems and possibilities. Sci. Total Environ.
407: 6100-6108.
Firsching, B.M., and N Claassen. 1996. Root phosphatase activity and soil organic
phosphorus utilization by Norway spruce (Picea abies (L) Karst). Soil Biol. Biochem.
28: 1417–1424.
Frossard, E., Condron, L.M., Oberson, A., Sinaj, S., and Fradeau, J.C. 2000. Processes
governing phosphorus availability in temperate soils. J. Environ. Qual. 29: 15-23.
Frossard, E., Stewart, J.W.B., and Arnaud, R.J.ST. 1989. Distribution and mobility of
phosphorus in grassland and forest soils of Saskatchewan. Can. J. Soil Sci. 69: 401-
416.
Guerrero, C., Rossel, R.A.V., and Mouazen, A.M. 2010. Special issue ‘Diffuse reflectance
spectroscopy in soil science and land resource assessment’. Geoderma, 158: 1-2.
Halstead, R.L., and McKercher, R.B. 1975. Biochemistry and cycling of phosphorus. Soil
Biol. Biochem. 4: 31- 63.
Harrison, A.F. 1982. Labile organic phosphorus mineralization in relationship to soil
properties. Soil Biol. Biochem. 14: 343-351.
Harrison, A.F. 1987. Soil organic phosphorus-A review of world literature. CAB Int.,
Wallingford, Oxon, UK. 257 p.
Haygarth, P.M., L. Hepworth, and S.C. Jarvis. 1998. Forms of phosphorus transfer in
hydrological pathways from soil under grazed grassland. Eur. J. Soil Sci. 49:65-72.
Hawkes, G.E., Powlson, D.S., Randall, E.W., and Tate, K.R. 1984. A 31P nuclear magnetic
resonance study of the phosphorus species in alkali extracts of soils from long-term
field experiments. J. Soil Sci. 35: 35-45.
He, Z., Honeycutt, C. W., Cade-Menun, B.J. Senwo, Z. N., and Trazisong, I. A. 2008.
Phosphorus in poultry and soil: enzymatic and nuclear magnetic resonance
characterization. Soil Sci. Soc. Am. J. 72: 1426-1433.
Hedley, M.J., Stewart, J.W.B., and Chauhan, B.S. 1982. Changes in inorganic and organic
soil phosphorus fractions induced by cultivation practices and by laboratory
incubations. Soil Sci. Soc. Am. J. 46: 970-976.
Hron, K., Jelínková, M., Filzmoser, P., Kreuziger, R., Bednář, P., and Barták, P., 2012.
Statistical analysis of wines using a robust compositional biplot. Talanta, 90: 46-50.
Hinsinger, P. 2001. Bioavailability of soil inorganic P in the rhizosphere as affected by
root-induced chemical changes: a review. Plant Soil. 237: 173-195.
Hinsinger, P., Jaillard, B., Le Cadre. É., and Plassard, C. 2007. Spéciation et biodisponiblité
du phosphore dans la rhizosphère. Océanis, 33: 37- 50.
26
Holford, I.C.R. 1997. Soil phosphorus: its measurement, and its uptake by plants. Austr. J.
Soil Res. 35: 227-239.
Holtan, H., Kamp-Nielson, L., and Stuanes, A.O. 1988. Phosphorus in soil, water and
sediment: an overview. Hydrobiologia, 170: 19-34.
Jing, S.R., Benefield, L.D. and Hill, W.E. 1992. Observations relating to enhanced
phosphorus removal in biological systems. Water Res. 26: 213–223.
Khiari, L., and Parent, L.É. 2005. Phosphorus transformations in acid light-textured soils
treated with dry swine manure. Can. J. Soil Sci. 85: 75- 87.
Khanna, M., Yoder, M., Calamai, L. and Stotzky, G. 1998. X-ray diffractometry and
electron microscopy of DNA from Bacillus subtilis bound on clay minerals. Sci. Soils.
Sedim. 3: 1–8.
Kleinman, P. J. A, Sharpley, A.N., Saporito, L.S., Buda, A.R., and Bryant, R.B. 2009.
Application of manure to no-till soils: phosphorus losses by sub-surface and surface
pathways. Nutr. Cycl. Agroeco. 84: 215–227.
Koopmans, G. F.; van Der Zeeuw, M. E., Chardon,W. J., and Dolfing, J. 2001. Selective
extraction of labile phosphorus using dialysis membrane tubes filled with hydrous
iron hydroxide. Soil Sci. 166: 475-483.
Korhoňová, M., Hron, K., Klimčíková, D., Müller, L., Bednář, P., and Barták, P., 2009.
Coffee aroma—Statistical analysis of compositional data. Talanta 80: 710-715.
Kuo, S. 1996. Phosphorus. In methods of soil analysis. Part 3-Chemical methods. Ed. D.L.
Sparks. 1390 p.
Legg, J.O. and Black, C.A. 1955. Determination of organic phosphorus in soils. II. Ignition
method. Soil Sci. Soc. Am. Proc. 19: 139–143.
Leinweber, P., Haumaier, L., and Zech, W. 1997. Sequential extractions and 31P-NMR
spectroscopy of phosphorus forms in animal manures, whole soils and particle-size
separatesfrom a densely populated livestock area in northwest Germany. Biol. Fert.
Soils. 25: 89-94.
Leinweber, P., Meissner, R., Eckhardt, K.U., and Seeger, J. 1999. Management effects on
forms of phosphorus in soil and leaching losses. Eur. J. Soil Sci. 50: 413-424.
Liu, Y. and Chen, J. 2008. Phosphorus cycle. Global Ecol. 2715- 2724.
Ludwig, B., Khanna, P.K., Bauhus, J., and Hopmans, P. 2002. Near infrared spectroscopy
of forest soils to determine chemical and biological properties related to soil
sustainability. Forest. Ecol. Manag. 171: 121- 132.
Magid, J. 1993. Vegetation effects on phosphorus fractions in set-aside soils. Plant Soil.
149: 111-119.
Maleki, M.R., Van Holm, L., Ramon, H., Merckx, R., De Baerdemaeker, J., and Mouazen,
A.M. 2006. Phosphorus Sensing for Fresh Soils using Visible and Near Infrared
Spectroscopy. Biosys. Engin. 95: 425- 436.
Mattingly, G.E.G. 1975. Labile phosphate in soils. Soil Sci. 119: 369- 375.
Martens, H., and T. Naes. 2001. Multivariate calibration by data compression. In: P.C.
Williams and K.H. Norris, editors, Near infrared technology in the agricultural and
food industries. 2nd ed. Am. Assoc. of Cereal Chemists, St. Paul, MN. pp. 59–100.
McCarty, G.W., and Reeves, J.B. 2006. Comparaison of Near infrared and mid infrared
diffuse reflectance spectroscopy for field scale measurement of soil fertility
parameters. Soil Sci. 171: 94-112.
McGechan, M.B., and Lewis, D.R. 2002. Sorption of phosphorus by soil, Part 1: Principles,
equations and models. Biosyst. Eng. 82: 1-24.
27
McKelvie, I.D. 2005. Separation, preconcentration and speciation of organic phosphorus in
environmental samples. Turner, B.L., Frossard, E., and Baldwin, D.S. Eds. 399 p.
McKenzie, R.H., Stewart, J.W.B., Dormaar, J.F., and Schaalje, G.B. 1992a. Long-term
crop rotation and fertilizer effects on phosphorus transformations: I. In a chernozemic
soil. Can. J. Soil Sci. 72: 569-579.
McKenzie, R.H., Stewart, J.W.B., Dormaar, J.F., and Schaalje, G.B. 1992b. Long-term
crop rotation and fertilizer effects on phosphorus transformations: II. In a luvisolic soil.
Can. J. Soil Sci. 72: 581-589.
McLaughlin, M.J., Alston, A.M., and Martin, J.K. 1988. Phosphorus cycling in water-
pasture rotations. II.*The role of the microbial biomass in phosphorus cycling. Aust. J.
Soil Res. 26: 333-342.
Mehlich, A. 1984. A modification of Mehlich 2 extractant. Commun. Soil Sci. Plant.Anal.
15: 1409–1416.
Mehta, N.C., Legg, J.O., Goring, C.A.I., and Black, C.A. 1954. Determination of organic
phosphorus in soils. I. Extraction method. Soil Sci. Soc. Am. Proc. 18: 443–449.
Messiga, A.J., Ziadi, Z., Morel, C., and Parent, L.É. 2010. Soil phosphorus availability in
no-till versus conventional tillage following freezing and thawing cycles. Can. J. Soil
Sci. 90: 419-428.
Morel, R. 1989. Les sols cultivés. Technique et Documentation-Lavoisier, Paris, France.
373 p.
Morel, C. 2002. Caractérisation de la phytodisponibilité du phosphore du sol par la
modélisation du transfert des ions phosphates entre le sol et la solution. Mémoire
préparé en vue de l’obtention du diplôme d’habilitation à diriger des recherches. Institut
National Polytechnique de Lorraine. INRA, Bordeaux, 80 p.
Motavalli, P.P., and Miles, R.J. 2002. Soil phosphorus fractions after 111 years of animal
manure and fertilizers applications. Bio. Fert. Soils. 36: 35-42.
Nduwamungu, C., Ziadi, N., Tremblay, G.F., and Parent, L.-É.. 2009a. Near-infrared
reflectance spectroscopy prediction of soil properties: Effects of sample cups and
preparation. Soil Sci. Soc. Am. J. 73: 1896–1903.
Nduwamungu, C., Ziadi, N., Parent, L.-É., and Tremblay, G.F.. 2009b. Mehlich 3
extractable nutrients as determined by near-infrared reflectance spectroscopy. Can. J.
Soil Sci. 89: 579–587.
Negassa, W., and Leinweber, P. 2009. How does the Hedley sequential phosphorus
fractionation reflect impacts of land use and management on soil phosphorus: A
review. J. Plant Nutr. Soil Sci. 172: 305–325.
Newman, R.H., and Tate, K.R. 1980. Soil phosphorus characterisation by 31P nuclear
magnetic resonance. Commun. Soil Sci. Plant Anal. 11: 835-842.
O’Halloran, I.P. and Cade-Menun, B.J. 2007. Total and organic phosphorus. In Soil
Sampling and Methods of analysis. Carter, M.R., and Gregorich, E.G., (2nd eds). pp.
265–291.
Oberson, A., and Joner, E.J. 2005. Microbial turnover of phosphorus in soil. Organic
phosphorus in the environment. Turner, B.L., Frossard, E., and Baldwin, D.S. Eds.
399 p.
Oehl, F., Oberson, A., Sinaj, S., and Frossard, E. 2001. Organic phosphorus mineralization
studies using isotopic dilution techniques. Soil Sci. Soc. Am. J. 65: 780–787.
28
Olsen, S.R., Cole, C.V., Watanabe, F.S. and Dean, L.A. 1954. Estimation of available
phosphorus by extraction with sodium bicarbonate. USDA Circ. 939, U.S. Govt.
Print. Office, Washington, D.C.
Olsen, S.R. and Sommers, L.E. 1982. Phosphorus. In A.L. Page, R.H., Miller and D.R.,
Keeney, Eds. Methods of Soil analysis. Part 2. (2nd ed.) Agronomy No. 9. American
Society of Agronomy, Madison, WI, USA. pp. 403–430.
Parent, L.É., Parent, S.É., Hébert-Gentile, V., Naess, K., and Lapointe, L. 2013a. Mineral
Balance Plasticity of Cloudberry (Rubus chamaemorus) in Quebec-Labrador Bogs.
Am. J. Plant Sci. 4: 1508-1520.
Parent, L.É., Parent, S.É., and Ziadi, N. 2014. Biogeochemistry of soil inorganic and
organic phosphorus: A compositional analysis with balances. J. Geochem. Expl. 141:
52–60.
Parent, L.É, Almeida, C.X., Hernandes, A., Egozcue, J.J., Gülser, C, Bolinder, M.A.,
Kätterer, T., Andrén, O., Parent, S.É., Anctil, F., Centurion, J.F., and Natale, W. 2012b.
Compositional analysis for an unbiased measure of soil aggregation. Geoderma 179-
180: 123–131.
Parent S.É., Karam A., and Parent L.E. 2011. Compositional modeling of C mineralization
of organic materials in soils. 2nd Int. Symp. Agricultural Waste Management, 15-17
March 2011 (Sigera Proceedings paper), Foz do Iguaçu, Brazil.
Parent, S.É., Parent, L.E., Rozanne, D.E., Hernandes, A. and Natale, W. 2012a. Nutrient
Balance as Paradigm of Soil and Plant Chemometrics. INTECH.
http://dx.doi.org/10.5772/53343.
Parent, S.É, Parent, L.É., Rozane, D.E. and Natale, W. 2013b. Plant ionome diagnosis
using sound balances: case study with mango (Mangifera Indica). Front. Plant Sci. 4: 1-
12.
Parkinson, J.A. and Allen, S.E. 1975. A wet oxidation procedure suitable for the
determination of nitrogen mineral nutrients in biological material. Comm. Soil Sci.
Plan. 6: 1–11.
Paytan, A., Cade-Menun, B.J., McLaughlin, K. and Faul, K.L. 2003. Selective phosphorus
regeneration of sinking marine particles: evidence from 31P NMR. Marine Chem. 82:
55–70.
Pawlowsky-Glahn, V., and Egozcue, J.J. 2006. Compositional data and their analysis: an
introduction. Compositional data analysis in the geosciences: from theory to pratice.
Buccianti, Mateu-Figueras, G., and Pawlowsky-Glahn, V. Eds. pp: 1-10.
Petersen, G.W. and Corey, R.B. 1966. A modified Chang and Jackson Procedure for
Routine Fractionation of Inorganic Soil Phosphates. Soil Sci. Soc. Am. J. 30: 563-565.
Pierzynski, G.M., McDowell, and Sims, J.T. 2005. Chemistry, cycling and potential
movement of inorganic phosphorus in soils. Phosphorus: agriculture and the
environment. Sims, J.T. and Sharpley, A.N. Eds. 53-86.
Quiquampoix, H. and Mousain, D. 2005. Enzymatic Hydrolysis of Organic Phosphorus. In
Phosphorus: agriculture and the environment. Sims, J.T. and Sharpley, A.N. Eds. pp:
89-112.
Redel, Y.D., Escudey, M., Alvear, M., Conrad, J. and Borie, F. 2011. Effects of tillage and
crop rotation on chemical phosphorus forms and some related biological activities in a
Chilean Ultisol. Soil Use Manag. 27: 221–228.
29
Redel, Y.D., Rubio, R., Rouanet, J.L., and Borie, F. 2007. Phosphorus bioavailability
affected by tillage and crop rotation on a chilean volcanic derived ultisol. Geoderma,
139: 388-396.
Richardson, A.E. 2001. Prospects for using soil microorganisms to improve the acquisition
of phosphorus by plants. Austr. J. Plant Physiol.. 28: 897- 906.
Ross, D.J., Tate, K.R., Scott, N.A., and Feltham, C.W. 1999. Land-use change: Effects on
soil carbon, nitrogen and phosphorus pools and fluxes in three adjacent ecosystems.
Soil Biol. Biochem. 31: 803-813.
Saunders, W.M.H. and Williams, E.G. 1955. Observations on the determination of total
organic phosphorus in soils. J. Soil Sci. 6: 254–267.
Sharpley, A.N., and Smith, S.J., 1994. Wheat tillage and water quality in the Southern
plains. Soil Tillage Res. 30: 33–38.
Shi, Y., Ziadi, N., Messiga, J.A., Lalande, R., and Hu, Z. 2013. Changes in soil phosphorus
fractions for a long-term corn-soybean rotation with tillage and phosphorus
fertilization. Soil Sci. Soc. Am. J. 77: 1402–1412.
Simard, R.R., Beauchemin, S., and Haygarth, P.M. 2000. Potentiel for preferential
pathways of phosphorus transport. J. Environ. Qual. 29: 97-105.
Sims, J.T., and Pierzynski, G.M. 2005. Chemistry of phosphorus in soil. In: Tabatabai A.M,
Sparks D.L. (eds) Chemical processes in soil, SSSA book series 8. SSSA, Madison, pp.
151-192.
Sissingh, H.A. 1971. Analytical technique of the Pw method used for the assessment of the
phosphate status of arable soils in the Netherlands. Plant Soil 34: 483-486.
Smernik, R.J., and Dougherty, W.J. 2007. Identification of phytate in phosphorus-31
nuclear magnetic resonance spectra: the need for spiking. Soil Sci. Soc. Am. J. 71:
1045- 1050.
Smith, A.N. 1965. Distinction between iron and aluminium phosphate in Chang and
Jackson’s procedure for fractionating inorganic soil phosphorus. Agrochimica, 9: 162-
168.
Stenberg, B., Rossel, R.A.V., Mouazen, A.M., and Wetterlind, J. 2010. Visible and near
infrared spectroscopy in soil science. Adv. Agron. 107: 163- 215.
Stevenson, F.J. 1986. Cycles of soil: carbon, nitrogen, phosphorus, sulfur, micronutrients.
380 p.
Steward, J.H. and Oades, J.M. 1972. The determination of organic phosphorus in soils. J.
Soil Sci. 23: 38–49.
Stewart, J.W.B., and Tiessen, H. 1987. Dynamics of soil organic phosphorus. Biogeochem.
4: 41-60.
Tiessen, H., and Moir, J.O. 2007. Characterization of available P by sequential extraction.
Soil sampling and methods of analysis. Carter, M.R., and Grecorich, E.G. Eds. pp. 75-
86.
Tiessen, H., Stewart, J.W.B., and Bettany, J.R. 1982. Cultivation effect on the amounts and
concentrations of carbon, nitrogen and phosphorus in grassland soils. Agron. J. 74:
831–835.
Tiessen, H., Stewart, J.W.B., and Cole, C.V. 1984. Pathways of phosphorus
transformations in soils of differing pedogenesis. Journal paper no. R361 of the
Saskatchewan Institute Of pedology, University of Saskatchewan. 8p.
30
Tolosana-Delgado, R. and K.-G. van den Boogart. 2011. Linear models with compositions
in R. pp. 356-371. In V. Pawlowsky-Glahn, and A. Buccianti, editors. Compositional
data analysis: Theory and Applications. John Wiley and Sons, New York.
Turner, B.L., Cade-Menun, B.J., Condron, L.M. and Newman, S. 2005. Extraction of soil
organic phosphorus. Talanta, 66: 294-306.
Turner, B.L., Mahieu, N. and Condron, L.M. 2003. Phosphorus composition of temperate
pasture soils determined by NaOH–EDTA extraction and solution 31P NMR
spectroscopy. Org. Geochem. 34: 1199–1210.
Tran, T.S., and N’Dayegamiye, A. 1995. Long term effects of fertilizers and manure
applications on the forms and availability of soil phosphorus. Can. J.Soil Sci. 75: 281-
285.
U.S. Geological Survey. 2014. Mineral commodities summaries. (Online). Available at
http://minerals.usgs.gov/minerals/pubs/commodity/phosphate_rock/ (verified 15
September 2014).
Van Groenigen, J.W., Mutters, C.S., Horwath, W.R. and van Kessel, C. 2003. NIR and
DRIFT-MIR spectrometry of soils for predicting soil and crop parameters in a flooded
field. Plant Soil. 250: 155- 165.
Veverka, L., Jelinkova, M., Hron, K., Balik, J., Stávek, J., and Barták, P., 2012. Chemical
markers in the aroma profiles of south Moravian red wine distillates. Czech J. Food Sci.
30: 369-376.
Vincent, A . G ., Turnerb, B. L., and Tannera, E. V. J. 2010. Soil organic phosphorus
dynamics following perturbation of litter cycling in a tropical moist forest. Eur. J. Soil
Sci. 61: 48–57.
Williams, J. D. H., Syers, J. K., and Walker, T. W. 1967. Fractionation of Soil Inorganic
Phosphate by a Modification of Chang and Jackson's Procedure. Soil Sci.Soc. Am. J.
31: 736-739.
Williams, J.D.H., Syers, J.K., Walker, T.W., and Rex, R.W. 1970. A comparison of
methods for the determination of soil organic phosphorus. Soil Sci. 110: 13–18.
Zhang, T.Q., and Mackenzie, A.F. 1997. Changes of soil phosphorous fractions under long-
term corn monoculture. Soil Sci. Soc. Am. J. 61: 485-493.
Zheng, Z., Simard, R.R., Lafond, J., and Parent, L.E. 2001. Changes in phosphorus
fractions of a humic gleysol as influenced by cropping systems and nutrient sources.
Can. J. Soil Sci. 81: 175-183.
Zheng, Z., Simard, R.R., Lafond, J., and Parent, L.E. 2002. Pathways of soil phosphorus
transformations after 8 years of cultivation under contrasting cropping practices. Soil
Sci. Soc. Am. J. 66: 999-1007.
Zheng, Z., Simard, R.R., and Parent, L-É. 2003. Anion exchange and Mehlich-III
phosphorus in Humaquepts varying in clay contents. Soil Sci. Soc. Am.J. 67: 1287-
1295.
Zuo, R., Xia, Q., and Wang, H., 2013. Compositional data analysis in the study of
integrated geochemical anomalies associated with mineralization. App. Geochem. 28:
202-211.
Ziadi, N., Whalen, J.K., Messiga, A.J., and Morel, C. 2013. Assessment and modeling of
soil svailable phosphorus in sustainable cropping systems. Adv. Agron. 122: 85-126.
31
CHAPITRE III: PREDICTING SOIL PHOSPHORUS-RELATED
PROPERTIES USING NEAR-INRARED REFLECTANCE
SPECTROSCOPY
Dalel Abdi1,2, Gaëtan F. Tremblay1, Noura Ziadi1, Gilles Bélanger1, and Léon-Étienne
Parent2
1Agriculture and Agri-Food Canada, Soils and Crops Research and Development Centre,
2560 Hochelaga Boulevard, Québec, QC, Canada G1V 2J3.
2Department of Soils and Agri-Food Engineering, Université Laval, Québec, QC, Canada
G1K 7P4.
Soil Science Society of America Journal, 2012. 76 (6): 2318–2326.
32
3.1 RÉSUMÉ
La spectroscopie dans le proche infra-rouge (SPIR) est une technique d’analyse
rapide, précise et peu coûteuse. Les objectives de cette étude étaient d’évaluer le potentiel
de la SPIR dans la prédiction (i) du P du sol extrait avec deux méthodes [Mehlich-3 (PM3)
et à l’eau (Cp)], P total (TP), P prelevé par les plantes, et bilan annuel de P, et (ii) d’autres
propriétés chimiques du sol [C total (TC), N total (TN), pH, K, Al, Fe, Ca, Mg, Cu et Zn
extraits au Mehlich-3]. Des échantillons du sol (n = 448) ont été prelevés d’un site
expérimental situé à Lévis, Québec. Les modèles de prédiction de la SPIR ont été
developpés en utilisant 80% des échantillons pour la calibration et 20% pour la validation.
Les résulats ont démontré que le PM3, Cp, P prelevé, bilan annuel de P, K et Cu n’ont pas
été prédictibles par la SPIR. Cependant, des prédictions fiables ont été trouvées pour TP,
TC, TN, Al, Fe, Zn, Mg, Ca, Mn, et pH.
33
3.2 ABSTRACT
Near infrared reflectance spectroscopy (NIRS) is a rapid, inexpensive, and accurate
analysis technique for a wide variety of materials and it is increasingly used in soil science.
The objectives of our study were to examine the potential of NIRS to predict: (i) soil P
extracted by two methods [Mehlich 3 (M3P) and water (Cp)], soil total P (TP), annual crop
P-uptake, and annual P-budget, and (ii) other soil chemical properties [total C (TC), total N
(TN), pH, and K, Al, Fe, Ca, Mg, Mn, Cu, and Zn extracted by Mehlich 3]. Soil samples (n
= 448) were taken over a 7-yr period from an experimental site in Lévis (Québec, Canada)
where timothy (Phleum pratense L.) was grown under four combinations of P and N
fertilizer. The NIRS equations were developed using 80% of the samples for calibration
and 20% for validation. The predictive ability of NIRS was evaluated using the coefficient
of determination of validation (Rv2) and the ratio of standard error of prediction to standard
deviation (RPD). Results show that M3P, Cp, crop annual P-uptake, and annual P-budget
were not accurately predicted by NIRS (Rv2 < 0.70 and RPD < 1.75). Similar results were
found for K and Cu. However, NIRS predictions were moderately useful for TP, TN, Fe,
and Zn (0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25), moderately successful for TC and Al
(0.80 ≤ Rv2 < 0.90 and 2.25 ≤ RPD < 3.00), successful for pH and Mg (0.90 ≤ Rv
2 ≤ 0.95
and 3.00 ≤ RPD ≤ 4.00), and excellent for Ca and Mn (Rv2 > 0.95 and RPD > 4.00). The
NIRS predictive ability of several soil properties appears to be related to their relationship
with soil organic C. Although NIRS can predict several soil properties, prediction of total P
was the only soil P-related property, correlated to soil C, that was moderately useful.
Abbreviations: b, slope of linear regression; Cp, P extracted in water; CV, coefficient of
variation; DM, dry matter; ICP, inductively coupled plasma; M3, Mehlich 3; M3P_Col, soil
P content extracted using the Mehlich 3 method and analysed by colorimetry; M3P_ICP,
soil P content extracted using the Mehlich 3 method and analysed by ICP; N, total number
of samples; NIRS, near infrared reflectance spectroscopy; PLSR, partial least squares
regression method; Rc2, coefficient of determination of calibration; Rv
2, coefficient of
determination of validation; Rep File, repeatability file; RPD, ratio of standard error of
prediction to standard deviation; SD, standard deviation; SEC, standard error of calibration;
SECV, standard error of cross-validation; SEP, standard error of prediction; SNVD,
standard normal variate and detrending; TC, total carbon; TN, total nitrogen; TP, total
phosphorus; 1-VR, coefficient of determination of cross-validation.
34
3.3 INTRODUCTION
Phosphorus is an essential nutrient and one of the most limiting for crop production.
Mineral and organic P fertilizers are often applied to agricultural soils to achieve optimal
crop yield but amounts exceeding crop requirements can have a negative environmental
impact.
Several methods and/or techniques of soil analysis, including chemical extraction
methods, have been developed to estimate the quantity of plant-available P in soils. Current
soil P extraction methods, such as Mehlich 3 (Mehlich, 1984), Olsen (Olsen et al., 1954),
Bray 1, Bray 2 (Bray and Kurtz, 1945), and water (Morel et al., 2000), are expensive,
destructive, and both time and space consuming. The recommended Mehlich 3 method for
a large range of soil types (Ziadi and Tran, 2007) requires five chemical reagents (acetic
acid, ammonium fluoride, ammonium nitrate, nitric acid, and ethylenediaminetetraacetic
acid).
Near infrared reflectance spectroscopy (NIRS) is a cost-effective, time-saving, non-
destructive, and environmentally-sound technique that can predict many constituents from
the single spectrum of a soil sample (Coûteaux et al., 2003; Viscarra Rossel et al., 2006),
including P (Chang et al., 2001; Ludwig et al., 2002; McCarty and Reeves, 2006).
Combined with minimal conventional reference methods, NIRS provides a good alternative
to routine soil analysis (Nduwamungu et al., 2009a) with at least an 80% reduction in
chemical use and laboratory costs (Foley et al., 1998). The NIRS measures the radiation
absorbed by various bonds of C-H, C-C, C-N, N-H, and O-H found in organic constituents
resulting in bending, twisting, stretching, or scissoring (Miller, 2001). Diffusely reflected
near infrared radiation is then correlated to measured material properties using various
multivariate calibration techniques (Martens and Naes, 2001; Mouazen et al., 2010).
Successful NIRS predictions have been reported for soil organic matter and texture
(Ben-Dor and Banin, 1995; Fidêncio et al., 2002; Coûteaux et al., 2003; Viscarra Rossel et
al., 2006; Stenberg, 2010) and for other soil properties including pH, CEC, N, P, K, Al, Fe,
Ca, and Zn (Reeves et al., 1999; Chang et al., 2001; Nduwamungu et al., 2009a). Although
some results appear promising, most studies use a limited number of samples (Malley et al.,
2004; Nduwamungu et al., 2009a, b, c). Nduwamungu et al. (2009b) report moderately
useful NIRS predictions for Mehlich 3 extractable Ca, Cu, and Mg, and less reliable
35
predictions for Al, Fe, K, Mn, P, and Zn. They conclude that further studies should
incorporate larger sample sizes and more diverse soils. To our knowledge, NIRS was used
to estimate water soluble P and total P (Bogrekci and Lee, 2005) but it has never been used
to estimate P extracted in water (Cp) according to the method of Morel et al. (2000).
Extracted soil P is often correlated with plant growth and P-uptake under controlled
conditions (Simard et al., 1991; Tran et al., 1992) and from field studies (Ziadi et al., 2001;
Messiga et al., 2010). Predicting crop P-uptake and P-budget from soil spectra would
eliminate the need to establish relationships between soil test P and crop response to P
fertilization. Börjesson et al. (1999), Terhoeven-Urselmans et al. (2008), and more recently
St. Luce et al. (2012) link NIRS soil spectra to winter cereal N-uptake and report good
predictions (Rv2 ≥ 0.70), but to our knowledge, the prediction of crop P-uptake and annual
P-budget by NIRS has not been documented.
The objective of this study was to evaluate the potential of NIRS to predict soil P-related
properties (total soil P, soil P extracted using a Mehlich 3 solution or water, annual crop P-
uptake, and annual P-budget) and other soil properties (pH, TC, TN, K, Al, Fe, Ca, Mg,
Mn, Cu, and Zn).
3.4 MATERIALS AND METHODS
3.4.1 Experimental site description
Detailed information on the experimental site is provided by Bélanger and Ziadi (2008)
and Bélanger et al. (2008). Briefly, the experiment was conducted between 1998 and 2007
on a gravely-sandy loam soil of the Saint-André series located at the Agriculture and Agri-
Food Canada research farm at Lévis, QC, Canada (46°47’ N, 71°07’ W, elevation 65 m).
The experimental design was a split-plot with four P treatments (0, 15, 30, and 45 kg P ha-
1) as main plots and four N treatments (0, 60, 120, and 180 kg N ha-1) as sub-plots. The
experiment had four replicates with a total of 64 sub-plots of equal size (1.5 m × 2.1 m).
Nitrogen (calcic ammonium nitrate) and P (triple superphosphate) fertilizers were applied
each year, during the first week of May from 1999 to 2006, prior to the start of timothy
growth. Potassium (KCl; 84 kg K ha-1) was applied with N and P to satisfy crop
requirements. The soil Mehlich-3 P content was 35.2 mg P kg1 when the experiment was
initiated in 1998.
36
3.4.2 Soil and plant analyses
Soil from plots (n=64) was sampled to a depth of 15 cm in the spring before N and P
was applied, each year from 2001 to 2007. Each sample consisted of 3 to 4 soil cores (2.5-
cm diameter) taken randomly within the experimental plot. The composite samples were
carefully mixed on site, air-dried, and gently crumbled by hand to pass through a 2-mm
sieve.
Soil P available to plants was characterized in the whole sample set (n=448, set 1) by
using two methods: water extraction to determine the concentration (Cp, mg P L-1) of P
ions in solution (Morel et al., 2000; Messiga et al., 2010) and the Mehlich-3 extraction
(Mehlich, 1984). To determine Cp, 2 g of air-dried soil were mixed with 20 mL of distilled
water and 200 μL of toluene to inhibit microbial activity. The solution was gently shaken
for 16 h on a horizontal roller shaker (40 cycles min-1) before passing through a disposable
cellulose acetate filter with a 0.2-μm cut-off (Minisart, Sartorius Gottingen, Germany). For
the Mehlich-3 extraction, 2.5 g of air-dried soil was mixed with 25 mL of a Mehlich-3
solution (0.25 M NH4NO3 + 0.015 M NH4F + 0.001 M EDTA + 0.2 M CH3COOH + 0.013 M
HNO3 buffered at pH 2.3), shaken for 5 min, and then filtered through Whatman No. 42
paper. Total soil P concentration was determined in 192 samples (collected in 2001, 2003,
and 2006, set 2, Table 1) using a method adapted from Nelson (1987) and used by Messiga
et al. (2012). Briefly, 0.1 g of finely ground soil (0.2 mm) was mixed in a 50-mL boiling
flask with 0.5 g K2S2O8 and 10 mL 0.9 M H2SO4, and digested at 121ºC in an autoclave for
90 min. Following Mehlich 3, water, and total P extractions, P was quantified by the
colorimetric blue method (Murphy and Riley, 1962). Also, Mehlich-3 P (M3P) was
measured by the inductively coupled plasma (ICP) emission spectroscopy (M3P_ICP) in
192 samples collected in 2005, 2006, and 2007 (set 3, Table 1).
Soil pH was measured in distilled water with a 1:2 soil to solution ratio (Hendershot et
al., 1993). Total C and TN were quantified by dry combustion with a LECO CNS-1000
analyzer (LECO Corp., St. Joseph, MI). The concentrations of K, Al, Fe, Ca, Mg, Mn, Cu,
and Zn were measured by ICP emission spectroscopy after the Mehlich 3 extraction
(Mehlich, 1984). Potassium, Al, and Fe were analyzed from samples collected in 2005,
2006, and 2007 (n=192, set 3), whereas Ca, Mg, Mn, Cu, and Zn were determined from
37
samples collected in 2005 and 2007 (n=128, set 4, Table 1). One chemical determination of
each soil property was done on soil samples.
From 2001 to 2006, timothy was harvested twice a year (n=378, set 5); the first harvest
was in mid-June, at the late heading stage of development, and the second harvest in early
August. Dry matter (DM) yield of each plot was determined from strips (0.91 m wide × 2.1
m long) harvested at a 5-cm height using a self-propelled flail forage harvester (Carter
MGF Co. Inc., Brookston, IN). A forage sample of approximately 500 g was collected from
each plot, dried at 55°C in a forced-draft oven for 3 d, and ground with a Wiley mill
(Standard model 3, Arthur H. Thomas Co., Philadelphia, PA) fitted with a 1-mm screen.
Plant samples of 0.1 g were digested using a mixture of sulphuric and selenious acids, as
described by Isaac and Johnson (1976). Phosphorus concentration was measured with a
QuikChem 8000 Lachat autoanalyzer (Lachat Instruments) using the Lachat method 13-
115-01-2-A (Lachat Instruments, 2011). The P-uptake at each harvest was calculated as the
product of forage P concentration and DM yield. Annual DM yield and crop P-uptake were
the sum of their first and second harvest values. The annual P-budget was computed as the
difference between P applied as fertilizer and annual P-uptake, as reported in Messiga et al.
(2012).
3.4.3 Near-infrared reflectance spectroscopy spectrum acquisition
Each soil sample was mixed and scanned by measuring his absorbance [log (1/R), where
R is reflectance] in the visible and near-infrared regions between 400 and 2500 nm at 2-nm
intervals using a NIRSystems 6500 monochromator Instrument (Foss NIRSystems Inc.,
Silver Spring, MD) with a cup (quarter cup, rectangular ¼) containing approximately 25
mL of the soil sample. This NIRS instrument is equipped with a tungsten-halogen light
source, a silicon detector for wavelength between 400-1100, a Pbs (Lead (II) Sulfide)
detector for wavelength in the range of 1100-2500 nm, and two intern standards
(polystyrene and didymium) that are used during sample spectrum acquisition. Each
spectrum was the mean of 16 co-added scans. A check test was performed prior to scanning
the soil sample and a performance test was done daily. One randomly selected soil sample
was scanned 12 times to create a repeatability file that was used to account for possible
operator errors and to improve calibration equations by minimizing errors associated with
soil heterogeneity and compaction (Nie et al., 2009).
38
3.4.4 Pretreatment, calibration, and cross-validation
To improve the calibration models of each property, the following 40 spectral pre-
treatments (2 × 2 × 2 × 5 factorial arrangement) were tested using WinISI III (ver.1.61)
software (Infrasoft International, LLC, Silver Spring, MD): (critical T-outlier values of 2.0
and 2.5) × (with and without a repeatability file) × (400-2500 nm and 1100-2500 nm
wavelength section) × (1-4-4-1, 2-8-6-1, 2-10-10-1, 0-4-4-1, and 0-8-6-1). Low and high
limits of the critical T-statistic, for T-outlier detection, were set to 2.0 and 2.5, respectively.
The math treatments that were compared are identified with four numbers (i.e., 1-4-4-1);
the first number is the derivative order, the second is the size of the gap in nm, the third is
the number of smoothing points, and the last is the second smooth (Ludwig et al., 2002;
Coûteaux et al., 2003). For each property, two criteria were used to select the best of the 40
spectral pre-treatments: simultaneous low standard error (SE) and high coefficient of
determination in cross-validation (1-VR) [Nduwamungu et al., 2009a]. The spectral pre-
treatments selected for each soil property are listed in tables 2 and 3, and only the results
using the best pre-spectral treatments are presented. Scatter correction with standard normal
variate and detrending (SNVD) was used to remove, or reduce, particle size and noise
effects (Brunet et al., 2007). The modified partial least squares regression method (PLSR)
of the WinISI III software was used to develop calibration equations for the soil and crop
properties. To maximize the probability of developing a robust calibration equation for
each property, a maximal number of soil samples, corresponding to, 80% of each soil
sample set, were randomly selected by the software to be used for the calibration set, and
the remainder samples were used for the validation set (Ludwig et al., 2002; Brunet et al.,
2007; St. Luce et al., 2012). General calibration equations were selected based on Martens
and Naes (2001) as follows: Reference data = f (spectral data) + SEC, where f () means
“function of” and SEC is the standard error of calibration. The best NIRS calibration
equations were the ones that minimize the SEC. Cross-validation was performed by using
four sub-groups from the calibration set in order to choose the optimal number of terms and
to avoid over-fitting the calibration model (Shenk and Westerhaus, 1991).
39
3.4.5 Validation
Calibration equations were validated using WinISI III software by comparing predicted
against reference values. Predicted values were generated using the modified PLSR method
of the WinISI III software according to Martens and Naes (2001): Predicted values = f
(spectrum data) = Reference data + error.
The accuracy of NIRS predictions was assessed with the following statistics: the
coefficient of determination of validation (Rv2) and the ratio of standard error of prediction
to standard deviation (RPD), which is the standard deviation of samples in the validation
set (SD) divided by the standard error of prediction corrected for the bias (SEP(C)) [RPD =
SD/SEP(C)]. Calibration equations were considered to be excellent when Rv2 > 0.95 and
RPD > 4.00; successful when 0.90 ≤ Rv2 ≤ 0.95 and 3.00 ≤ RPD ≤ 4.00; moderately
successful when 0.80 ≤ Rv2 < 0.90 and 2.25 ≤ RPD < 3.00; moderately useful when
0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25; and less reliable when Rv
2 < 0.70 and
RPD < 1.75 (Malley et al., 2004). The coefficient of variation (CV) in the reference set was
defined as the SD divided by the mean of chemical values, whereas the coefficient of
variation in the calibration set was computed as the ratio of standard error of calibration
(SEC) to the mean of calibration data (Williams, 2001).
Predictive graphs, illustrating the relationships between predicted and reference values
for P-related properties, were created with SigmaPlot for Windows (SYSTAT, 2012,
version 12.1). Pearson correlations between soil total C and soil and plant properties
analyzed on samples collected in 2001, 2003, and 2006 [soil P content extracted using the
Mehlich 3 method and analysed by colorimetry (M3P_Col), M3P_ICP, Cp, TP, P-uptake,
P-budget, TN, and pH] were computed with the SPSS 19 software (SPSS, 2010).
3.5 RESULTS AND DISCUSSION
3.5.1 Reference data
The CV values for the measured properties were relatively high (> 50%, Table 1) for Cp,
annual P-budget, and Mn and intermediate (20–50%) for M3P_Col, M3P_ICP, annual P-
uptake, K, Mg, and Zn. The CV values for the measured TP, TC, TN, pH, Al, Fe, Ca, and
Cu were low (< 20%). This variability of the soil properties is mostly related to direct and
indirect effects of the different N and P fertilization treatments during the seven years of the
40
study. According to Dardenne et al. (2000), a wide range of values for a given property is
required to obtain high NIRS calibration accuracy and thus good predictive performance.
3.5.2 Spectral pretreatments
Spectral pre-treatments that gave the best calibration equation for each property are
provided in Tables 2 and 3. Based on the Rv2 and RPD statistics, selected calibration
equations performed slightly better for M3P_Col, Cp, TC, TN, pH, K, Fe, Mg, Cu, and Zn
when a repeatability file was used. Also, M3P_ICP, P-uptake, TP, TC, TN, Fe, and Ca were
better predicted with the wavelength region of 1100–2500 nm than with the 400–2500 nm
region, whereas the other properties were more accurately predicted with the whole visible-
near-infrared spectrum. McCarty and Reeves (2006) report that the use of mid-infrared
spectra may yield better calibrations, than the NIR spectral region, for K, Ca, and Mg
extracted with Mehlich 1 and analysed using atomic absorption spectroscopy.
3.5.3 Near-infrared reflectance spectroscopy prediction of soil and crop P properties
Statistics for calibration and cross-validation are listed in Table 2. The number of T-
outliers was 20 for Cp and less than 8 for the other P properties, indicating that the
development of calibration equations was based on more than 94% of the soil samples in
the calibration set. Based on the high standard error and CV (> 23%) and the low
coefficients of determination of calibration equations (Rc2 ≤ 0.70) for M3P_Col, M3P_ICP,
Cp, P-uptake, and P-budget, NIRS calibration performances were considered poor for these
properties. As a result, cross-validation of calibration equations showed high standard error
of cross-validation (SECV) and low coefficient of determination of cross-validation (1-VR)
values (≤ 0.55). However, the calibration for TP resulted in an acceptable coefficient of
determination (Rc2 = 0.78) while the cross-validation was acceptable with a 1-VR of 0.76.
The number of soil samples used in the validation set varied between 38 and 90
depending on the P-related property (Table 2). Slopes of the linear regression (b) between
reference and predicted values of M3P_Col, M3P_ICP, Cp, P-uptake, and P-budget from
the validation set were less than 0.60 while the Rv2 (≤ 0.49) and RPD values (≤ 1.37) were
low (Fig. 3.1). The relationship between reference and predicted values for these properties
was therefore poor (b < 0.80, Rv2 < 0.70, RPD < 1.75; Nduwamungu et al., 2009c),
indicating that these properties cannot be predicted by NIRS.
41
This result agrees with Nduwamungu et al. (2009b) which reports a poor calibration
performance for M3P when analysed by ICP emission spectroscopy on a limited number of
soil samples (n = 150). Chang et al. (2001) also reports a low accuracy for M3P_ICP
prediction using NIRS with a principal component regression technique (Rv2 = 0.40, RPD =
1.18). Similarly, McCarty and Reeves (2006) found that soil P content extracted using the
Mehlich 1 method and analysed by colorimetry cannot be predicted by NIRS (Rv2 = 0.21).
However, NIRS has been shown to be useful (Rv2 = 0.71; RPD = 1.81) to predict P when
measured by the Olsen method (van Groenigen et al., 2003). Thus, it appears that the
performance of NIRS calibration could be affected by the reference method used that might
produce different reference values. Morón and Cozzolino (2007) report that NIRS
predictions of soil P were slightly more reliable when based on a resin extracted P method
(Rv2 = 0.61, RPD = 2.2) rather than the Bray method (Rv
2 = 0.58, RPD = 1.72). Sørensen
and Dalsgaard (2005) suggest that NIRS could be useful to predict soil P if there is an
indirect relationship between soil P and organic components, which means that P relates to
NIRS by covariation (Stenberg, 2010). Indeed, Ludwig et al. (2002) reports useful
calibration for soil P measured by the Olsen method, which was highly correlated with soil
C content (r = 0.67). In our study, soil C content was not significantly correlated to
M3P_Col (r = -0.04, P = 0.56), M3P_ICP (r = 0.10, P = 0.45), P-uptake (r = 0.08,
P = 0.30), and P-budget (r = 0.02, P = 0.80), and significantly but weakly correlated to Cp
(r = 0.31, P < 0.001, data not shown). Moderately useful NIRS prediction (Rv2 = 0.78) was
previously reported for water soluble P in 150 fine sandy soil samples collected from three
sites in Florida (USA, Bogrekci and Lee, 2005), but to our knowledge, the potential of
NIRS to predict Cp (Morel et al., 2000), P-uptake, and P-budget has not previously been
studied.
Successful calibration was found for TP as indicated by high Rc2 and 1-VR values,
which resulted in a moderately useful prediction of TP from the validation set (Fig. 1) with
Rv2 = 0.75 and RPD = 1.98. This result can be explained, in part, by the fact that TP
contains a certain proportion of organic P that is related to organic matter (Turner et al.,
2005). Bogrekci and Lee (2005) report successful NIRS prediction of TP (Rv2 = 0.92) in
150 fine sandy soil samples. Future research is needed to verify whether the ability of NIRS
42
to predict soil TP is related to soil texture and to evaluate the potential for NIRS to predict
soil organic P, since it is highly correlated with the concentration of organic matter.
3.5.4 Near-infrared reflectance spectroscopy prediction of other soil properties
Statistics of calibration and validation for all other soil properties are provided in Table
3. The number of T-outliers excluded from the calibration set was < 14; at least 89% of the
calibration samples were used to generate the prediction equations (Table 3). The NIRS
predictions were moderately successful for TC (Rv2 = 0.87; RPD = 2.82) and moderately
useful for TN (Rv2 = 0.79; RPD = 2.2). Chang and Laird (2002) also report successful NIRS
predictions (0.86 < Rv2 < 0.91; 2.8 < RPD < 4.4) for these two properties from 108 samples
obtained from a wide range of soil groups and cropping histories. Soil pH was successfully
predicted by NIRS (Table 3); our predictions were better than those reported by Dunn et al.
(2002) [Rv2 = 0.80, RPD = 2.3] and He et al. (2007) [Rv
2 = 0.82]. Chang et al. (2001) link
the accurate prediction of pH to its significant correlation with clay content and soil organic
matter. In our study, pH was significantly correlated to soil TC (r = 0.52, P < 0.001) and
TN (r = 0.54, P < 0.001, data not shown). In contrast, Nduwamungu et al. (2009a) report a
less reliable prediction for pH due to weak correlations between soil pH and soil TC and
TN.
Excellent NIRS predictions were found for Ca and Mn (Rv2 > 0.95, RPD > 4.0).
Furthermore, predictions were successful for Mg and moderately successful for Al. The Fe
and Zn calibration equations were moderately useful, while the K and Cu calibration
equations had the lowest Rv2 (< 0.70) and RPD (< 1.75) which means that their NIRS
predictions were unacceptable.
Compared with our results, Nduwamungu et al. (2009b) report lower NIRS prediction
performance for Al, Fe, Ca, Mn, and Zn, and greater calibration accuracy for Mg and Cu
when extracted by the Mehlich 3 method and analysed by atomic absorption spectroscopy.
The ability of NIRS to predict certain soil properties may be related to their correlation with
primary soil properties such as texture (Chang et al., 2001; Nduwamungu et al., 2009a), or
organic matter content (Dalal and Henry, 1986). The CV values in the reference data used
by Nduwamungu et al. (2009b) were actually greater than those from our data set (Table 1),
except for K and Mn; however, we found better NIRS predictions for the majority of the
soil properties. Similarly, Chang et al. (2001) report similar NIRS prediction performance
43
for soil K and Cu to ours, extracted by Mehlich 3 and determined by ICP, but their
predictions for Fe, Ca, Mg, Mn, and Zn were less accurate despite having a reference data
set with a large variability. Hence, high CVs that reflect a large range in soil properties do
not always guarantee good NIRS predictions.
3.6 CONCLUSIONS
The current study showed that soil P-related properties of a gravely loam soil,
including M3P_Col, M3P_ICP, Cp, annual P-uptake, and annual P-budget, were not
accurately predicted by NIRS. These unsatisfactory NIRS predictions may be related to the
low correlation observed between the P-related properties and soil C content. However,
NIRS predictions were considered moderately useful for soil TP, TN, Fe, and Zn,
moderately successful for TC and Al, successful for pH and Mg, and excellent for Ca and
Mn. Although NIRS can predict several soil properties, prediction of total P was the only
soil P-related property, correlated to soil C, that was moderately useful. These findings on
homogenous textured soil samples should be validated on soils of diverse textures.
3.7 ACKNOWLEDGEMENTS
The authors acknowledge the technical assistance of Mario Laterrière, Claude
Levesque, and Danielle Mongrain (Agriculture and Agri-Food Canada, Soils and Crops
Research and Development Centre, Québec, Canada). We also thank Mervin St. Luce for
his comments on an early version of this manuscript and acknowledge the assistance of
Christina McRae, from Editworks (Nova Scotia, Canada), for the structural editing of this
manuscript. This study was funded by The Sustainable Agriculture Environmental Systems
(SAGES) initiative of Agriculture and Agri-Food Canada.
44
3.8 REFERENCES
Bélanger, G., G.F. Tremblay, and D. Mongrain. 2008. Yield and nutritive value of the
spring growth of an aging timothy sward. Can. J. Plant Sci. 88:457–464.
Bélanger, G., and N. Ziadi. 2008. Phosphorus and nitrogen relationships during spring
growth of an aging timothy sward. Agron. J. 100:1757–1762.
Ben-Dor, E., and A. Banin. 1995. Near infrared analysis as a rapid method to
simultaneously evaluate several soil properties. Soil Sci. Soc. Am. J. 59:364–372.
Bogrekci, I., and W.S. Lee. 2005. Spectral phosphorus mapping using diffuse reflectance of
soils and grass. Biosystems. Eng. 91:305–312.
Börjesson, T., B. Stenberg, B. Lindén, and A. Jonsson. 1999. NIR spectroscopy, mineral
nitrogen analysis and soil incubations for the prediction of crop uptake of nitrogen
during the growing season. Plant Soil 214:75–83.
Bray, R.H., and L.T. Kurtz. 1945. Determination of total, organic, and available forms of
phosphorus in soils. Soil Sci. 59:39–45.
Brunet, D., B.G. Barthès, J.-L. Chotte, and C. Feller. 2007. Determination of carbon and
nitrogen contents in Alfisols, Oxisols and Ultisols from Africa and Brazil using NIRS
analysis: Effects of sample grinding and set heterogeneity. Geoderma 139:106–117.
Chang, C.W., and D.A. Laird. 2002. Near-infrared reflectance spectroscopic analysis of soil
C and N. Soil Sci. 167:110–116.
Chang, C.W., D.A. Laird, M.J. Mausbach, and C.R.Jr. Hurburgh. 2001. Near-infrared
reflectance spectroscopy – Principal components regression analyses of soil
properties. Soil Sci. Soc. Am. J. 65:480–490.
Coûteaux, M.M., B. Berg, and P. Rovira. 2003. Near infrared reflectance spectroscopy for
determination of organic matter fractions including microbial biomass in coniferous
forest soils. Soil Biol. Biochem. 35:1587–1600.
Dalal, R.C., and R.J. Henry. 1986. Simultaneous determination of moisture, organic carbon,
and total nitrogen by near infrared reflectance, spectroscopy. Soil Sci. Soc. Am. J.
50:120–123.
Dardenne, P., G. Sinnaeve, and V. Baeten. 2000. Multivariate calibration and chemometrics
for near infrared spectroscopy: which method? J. Near Infrared Spectrosc. 8:229–237.
Dunn, B.W., H.G. Beecher, G.D. Batten, and S. Ciavarella. 2002. The potential of near-
infrared reflectance spectroscopy for soil analysis – a case study from the Riverine
Plain of south-eastern Australia. Aus. J. Exp. Agri. 42:607–614.
Fidêncio, P.H., R.J. Poppi, and J.C. de Andrade. 2002. Determination of organic matter in
soils using radial basis function networks and near infrared spectroscopy. Anal. Chim.
Acta 453:125–134.
Foley, W.J., A. McIlwee, I.R. Lawler, L.V. Aragones, A.P. Woolnough, and N. Berding.
1998. Ecological applications of near infra-red spectroscopy - a tool for rapid, cost-
effective prediction of the composition of plant and animal tissues and aspects of
animal performance. Oecologia 116:293–305.
He, Y., M. Huang, A. García, A. Hernández, and H. Song. 2007. Prediction of soil
macronutrients content using near-infrared spectroscopy. Comput. Electron. Agric.
58:144–153.
45
Hendershot, W.H., H. Lalande, and M. Duquette. 1993. Soil reaction and exchangeable
acidity. p. 141–145. In Carter, M.R. (ed.) Soil Sampling and Methods of Analysis for
Canadian Society of Soil Science. Lewis Publishers, Boca Raton, FL.
Isaac, R.A., and W.C. Johnson. 1976. Determination of total nitrogen in plant tissue, using
a block digestor. J. Assoc. Anal. Chem. 59:98–100.
Lachat Instruments. 2011. Methods list for automated ion analyzers (flow injection
analyses, ion chromatography). Available at
http://www.lachatinstruments.com/applications/Methods.asp (verified 27 Feb. 2012).
Lachat Instruments, Loveland, CO.
Ludwig, B., P.K. Khanna, J. Bauhus, and P. Hopmans. 2002. Near infrared spectroscopy of
forest soils to determine chemical and biological properties related to soil
sustainability. F. Ecol. Mang. 171:121–132.
Malley, D.F., E. Ben-Dor, and P.D. Martin. 2004. Application in analysis of soils. p. 729–
84. In Roberts, C.A., J. Jr. Workman, and J.B. Reeves III (ed.) Near-infrared
Spectroscopy in Agriculture. Am. Soc. Agr., Crop Sc. Soc. Am., and Soil Sc. Soc.
Am., Madison, USA.
Martens, H., and T. Naes. 2001. Multivariate calibration by data compression. p. 59–100.
In Williams, P.C., and K.H. Norris (ed.) Near Infrared Technology in the Agricultural
and Food Industries. 2nd ed. American Association of Cereal Chemists, St. Paul, MN.
McCarty, G.W., and J.B. Reeves, III. 2006. Comparison of near infrared and mid infrared
diffuse reflectance spectroscopy for field-scale measurement of soil fertility
parameters. Soil Sci. 171:94–102.
Mehlich, A. 1984. Mehlich 3 soil test extractant: A modification of Mehlich 2 extractant.
Commun. Soil Sci. Plant Anal. 15:1409–1416.
Messiga, A.J., N. Ziadi, C. Morel, and L.É. Parent. 2010. Soil phosphorus availability in
no-till versus conventional tillage following freezing and thawing cycles. Can. J. Soil
Sci. 90:419–428.
Messiga, A. J., N. Ziadi, G. Bélanger, and C. Morel. 2012. Process-based mass-balance
modeling of soil phosphorus availability in a grassland fertilized with N and P. Nutr.
Cycling Agroecosyst. 92:272-287.
Miller, C.E. 2001. Chemical principles of near-infrared technology. p. 19–8. In Williams,
P.C., and K.H. Norris (ed.) Near Infrared Technology in the Agricultural and Food
Industries 2nd ed. American Association of Cereal Chemists, St. Paul, MN.
Morel, C., H. Tunney, D. Plenet, and S. Pellerin. 2000. Transfer of phosphate ions between
soil and solution: Perspectives in soil testing. J. Environ. Qual. 29:50–59.
Morón, A., and D. Cozzolino. 2007. Measurement of phosphorus in soils by near infrared
reflectance spectroscopy: Effect of reference method on calibration. Commun. Soil
Sci. Plant Anal. 38:1965–1974.
Mouazen, A.M., B. Kuang, J. De Baerdemaeker, and H. Ramon. 2010. Comparison among
principal component, partial least squares and back propagation neural network
analyses for accuracy of measurement of selected soil properties with visible and near
infrared spectroscopy. Geoderma 158:23–31.
Murphy, J., and J.P. Riley. 1962. A modified single solution method for the determination
of phosphate in natural waters. Anal. Chim. Acta 27:31–36.
Nduwamungu, C., N. Ziadi, G.F. Tremblay, and L.-É. Parent. 2009a. Near-infrared
reflectance spectroscopy prediction of soil properties: Effects of sample cups and
preparation. Soil Sci. Soc. Am. J. 73:1896–1903.
46
Nduwamungu, C., N. Ziadi, L.-É. Parent, and G.F. Tremblay. 2009b. Mehlich 3 extractable
nutrients as determined by near-infrared reflectance spectroscopy. Can. J. Soil Sci.
89:579–587.
Nduwamungu, C., N. Ziadi, L.-É. Parent, G.F. Tremblay, and L. Thuriès. 2009c.
Opportunities for, and limitations of, near infrared reflectance spectroscopy
applications in soil analysis: A review. Can. J. Soil Sci. 89:531–541.
Nelson, N.S. 1987: An acid‐persulfate digestion procedure for determination of phosphorus
in sediments. Commun. Soil Sci. Plant Anal. 18:359–369.
Nie, Z., G.F. Tremblay, G. Bélanger, R. Berthiaume, Y. Castonguay, A. Bertrand, R.
Michaud, G. Allard, and J. Han. 2009. Near-infrared reflectance spectroscopy
prediction of neutral detergent-soluble carbohydrates in timothy and alfalfa. J. Dairy
Sci. 92:1702–1711.
Olsen, S.R., C.V. Cole, F.S. Watanabe, and L.A. Dean. 1954. Estimation of available
phosphorus by extraction with sodium bicarbonate. USDA Circ. 939, U.S. Govt.
Print. Office, Washington, D.C.
Reeves, III J.B., G.W. McCarty, and J.J. Meisinger. 1999. Near infrared reflectance
spectroscopy for the analysis of agricultural soils. J. Near Infrared Spectrosc. 7:179–
193.
Simard, R.R., T.S. Tran, and J. Zizka. 1991. Strontium chloride-citric acid extraction
evaluated as a soil testing procedure for phosphorus. Soil Sci. Soc. Am. J. 55:14–421.
Sørensen, L.K., and S. Dalsgaard. 2005. Determination of clay and other soil properties by
near infrared spectroscopy. Soil Sci. Soc. Am. J. 69:159–167.
SPSS. 2010. IBM SPSS Statistics 19 Core System User’s Guide. Version 19. IBM
Company, US.
Shenk, J.S., and M.O. Westerhaus. 1991. Population definition, sample selection, and
calibration procedures for near infrared reflectance spectroscopy. Crop Sci. 31:469–
474.
St. Luce, M., N. Ziadi, J. Nyiraneza, G.F. Tremblay, B.J. Zebarth, J.K. Whalen, and M.
Laterrière. 2012. Near infrared reflectance spectroscopy prediction of soil nitrogen
supply in humid temperate regions of Canada. Soil Sci. Soc. Am. J. 76:1454–1461.
Steffens, D. 1994. Phosphorus release kinetics and extractable phosphorus after long-term
fertilization. Soil Sci. Soc. Am. J. 58:1702–1708.
Stenberg, B. 2010. Effects of soil sample pretreatments and standardised rewetting as
interacted with sand classes on Vis-NIR predictions of clay and soil organic carbon.
Geoderma 158:15–22.
SYSTAT. 2012. SigmaPlot 12.0. SYSTAT Software, Inc., Point Richmond, CA.
Terhoeven-Urselmans, T., H. Schmidt, R. Georg Joergensen, and B. Ludwig. 2008.
Usefulness of near-infrared spectroscopy to determine biological and chemical soil
properties: Importance of sample pre-treatment. Soil Biol. Biochem. 40:1178–1188.
Tran, T.S., R.R. Simard, and M. Tabi. 1992. Evaluation of the electro-ultrafiltration
technique (EUF) to determine available P in neutral and calcareous soils. Commun.
Soil Sci. Plant Anal. 23:2261–2281.
Turner, B.L., B.J. Cade-Menun, L.M. Condron, and S. Newman. 2005. Extraction of soil
organic phosphorus. Talanta 66:294–306.
van Groenigen, J.W., C.S. Mutters, W.R. Horwath, and C. van Kessel. 2003. NIR and
DRIFT-MIR spectrometry of soils for predicting soil and crop parameters in a flooded
field. Plant Soil 250:155–165.
47
Viscarra Rossel, R.A.V., D.J.J. Walvoort, A.B. McBratney, L.J. Janik, and J.O. Skjemstad.
2006. Visible, near infrared, mid infrared or combined diffuse refelectance
spectroscopy for simultaneous assessment of various soil properties. Geoderma
131:59–75.
Williams, P.C. 2001. Implementation of near-infrared spectroscopy. p. 145–169. In
Williams, P.C., and K. Norris (ed.) Near-infrared Technology in the Agricultural and
Food Industries. 2nd ed. American Association of Cereal Chemists, St. Paul, MN.
Ziadi, N., R.R. Simard, T.S. Tran, and G. Allard. 2001. Evaluation of soil-available
phosphorus for grasses with Electro-Ultrafiltration technique and some chemical
extractions. Can. J. Soil Sci. 81:167–174.
Ziadi, N., and T. Tran. 2007. Mehlich 3-extractable elements. In: M.R. Carter et al., editors,
Soil sampling and methods of analysis. Lewis, Boca Raton, p. 81-88.
48
Tableau 3-1 Descriptive statistics† for the soil P-related and other properties analyzed
using reference methods.
Property Sampled years N (set #) Reference
method Min Max Mean SD CV (%)
mg kg-1
M3P_Col 2001–2007 448 (1) Colorimetry 4 102 40 19 48
M3P_ICP 2005–2007 192 (3) ICP 17 124 54 24 44
TP 2001+ 2003 + 2006 192 (2) Colorimetry 450 1242 729 136 19
mg L-1
Cp 2001–2007 448 (1) Colorimetry 0.06 1.28 0.33 0.20 61
kg ha-1
P-uptake 2001–2006 378 (5) Colorimetry 1.6 27.7 13.5 5.9 44
P-budget 2001–2006 378 (5) -26.5 42.4 9 17 189
g kg-1
TC 2001+ 2003 + 2006 192 (2) Dry combustion 20.2 31.3 25.2 2.2 9
TN 2001+ 2003 + 2006 192 (2) Dry combustion 1.8 2.8 2.2 0.2 9
pH 2001+ 2003 + 2006 192 (2) Water (1:2) 4.6 6.4 5.5 0.4 7
mg kg-1
K 2005–2007 192 (3) ICP‡ 46 332 133 61 46
Al 2005–2007 192 (3) ICP 769 1225 995 95 10
Fe 2005–2007 192 (3) ICP 162 315 233 30 13
Ca 2005 + 2007 128 (4) ICP 986 2314 1559 271 17
Mg 2005 + 2007 128 (4) ICP 134 449 280 81 29
Mn 2005 + 2007 128 (4) ICP 13 110 36 23 64
Cu 2005 + 2007 128 (4) ICP 1.1 2.6 1.6 0.3 19
Zn 2005 + 2007 128 (4) ICP 0.8 4.2 2 0.6 30 †N, number of samples; Min, minimum; Max, maximum, SD, standard deviation; CV, coefficient of variation
[(SD/mean) × 100]. ‡ICP, inductively coupled plasma.
49
Tableau 3-2 NIRS spectral pre-treatments and statistics† of calibration, cross-validation,
and validation for the P-related soil properties.
M3P_Col M3P_ICP Cp TP Annual
P-uptake Annual
P-budget
Statistic (mg kg-1)‡ (mg L-1)‡ (mg kg-1)‡ (kg P ha-1)‡
Pre-treatment Math treat 2,10,10,1 0,8,6,1 1,4,4,1 0,4,4,1 1,4,4,1 0,8,6,1 T 2.5 2.5 2.5 2.5 2.5 2.5 Region (nm) 400–2500 1100–2500 400–2500 1100–2500 1100–2500 400–2500 Rep File Yes No Yes No No No
Calibration Nc 350 151 338 153 295 298 T-outliers 8 3 20 1 7 4 Mean 40 54 0.30 721 13.7 8.9 SEC 16 21 0.12 61 3.1 13.2 CV (%) 40 39 40 8 23 148 Rc
2 0.43 0.23 0.46 0.78 0.70 0.40 Cross-validation
SECV 16 22 0.12 63 3.9 14.4 1-VR 0.30 0.17 0.43 0.76 0.55 0.29
Validation Nv 90 38 90 38 76 76 Mean 38.0 54.3 0.30 762 13.1 9.3 SD 18.1 24.6 0.17 156.1 5.8 15.7 SEP(C) 15.7 21.4 0.13 78.8 4.2 14.8 †Math treat, mathematical treatment; T, critical outlier value; Region, wavelength region of the spectrum that
was used; Rep File, repeatability file; Nc, number of samples used for calibration; T-outliers, outliers
eliminated during calibration; SEC, standard error of calibration; CV, coefficient of variation defined as the
ratio of SEC to the mean; Rc2, coefficient of determination of calibration; SECV, standard error of cross-
validation; 1-VR, coefficient of determination of cross-validation; Nv, number of samples used for validation;
SD, standard deviation; SEP(C), standard error of prediction corrected for the bias. ‡The units apply only to Means, SEC, SECV, SD, and SEP(C).
50
Tableau 3-3 NIRS spectral pre-treatment and statistics† of calibration, cross-validation, and validation for the other soil
properties.
TC TN pH K Al Fe Ca Mg Mn Cu Zn
Statistic (g kg-1)‡ (mg kg-1)‡
Pre-treatment
Math treat 2,10,10,1 2,8,6,1 2,10,10,1 1,4,4,1 1,4,4,1 2,8,6,1 1,4,4,1 2,8,6,1 2,8,6,1 0,8,6,1 0,4,4,1
T 2.5 2.5 2.5 2.5 2 2 2 2 2.5 2 2 Region (nm) 1100–2500 1100–2500 400–2500 400–2500 400–2500 1100–2500 1100–2500 400–2500 400–2500 400–2500 400–2500 Rep File Yes Yes Yes Yes No Yes No Yes No Yes Yes
Calibration
Nc 148 153 151 150 140 141 94 93 97 91 93 T-outliers 6 1 3 4 14 13 8 9 5 11 9 Mean 25.3 2.2 5.5 129 997 232 1564 287 35 1.7 1.9 SEC 0.5 0.1 0.1 25 31 13 33 13 3 0.1 0.2 CV (%) 1.9 4.5 1.8 19.4 3.1 5.6 2.1 4.5 8.5 5.8 10.5 Rc
2 0.93 0.79 0.94 0.79 0.89 0.81 0.98 0.97 0.98 0.84 0.83
Cross-validation
SECV 0.7 0.1 0.1 36 43 16 62 18 5 0.1 0.2 1-VR 0.88 0.73 0.89 0.58 0.80 0.70 0.94 0.94 0.95 0.79 0.78
Validation
Nv 38 38 38 38 38 38 25 25 25 25 25 Mean 24.9 2.2 5.4 147 989 231 1487 260 38 1.5 2.0 SD 2.23 0.22 0.42 71 81 29 285 83 23 0.22 0.52 SEP(C) 0.79 0.10 0.13 44.6 31.6 14.1 56.3 20.5 4.1 0.19 0.25 Rv
2 0.87 0.79 0.91 0.62 0.85 0.77 0.96 0.94 0.97 0.37 0.78 RPD 2.82 2.20 3.23 1.59 2.56 2.05 5.06 4.04 5.60 1.16 2.08 Prediction§ MS MU S LR MS MU E S E LR MU †Math treat, mathematical treatment; T, critical outlier value; Rep File, repeatability file; Nc, number of samples used for calibration; T-outliers, outliers
eliminated during calibration; SEC, standard error of calibration; CV, coefficient of variation defined as the ratio of SEC to the mean multiplied by 100; Rc2,
coefficient of determination of calibration; SECV, standard error of cross-validation; 1-VR, coefficient of determination of cross-validation, Nv, number of
51
samples used for validation, SD, standard deviation; SEP(C), standard error of prediction corrected for the bias; Rv2, coefficient of determination of validation;
RPD, ratio of standard error of prediction to standard deviation which is the SD of samples in the validation set divided by the SEP(C). ‡The units apply only to Means, SD, SEC, SECV, SD, and SEP(C). §Based on validation statistics, the NIRS predictions were considered excellent (E) when Rv
2 > 0.95 and RPD > 4; successful (S) when 0.90 ≤ Rv2 ≤ 0.95 and
3 ≤ RPD ≤ 4; moderately successful (MS) when 0.80 ≤ Rv2 < 0.90 and 2.25 ≤ RPD < 3; moderately useful (MU) when 0.70 ≤ Rv
2 < 0.80 and 1.75 ≤ RPD < 2.25;
and less reliable (LR) when Rv2 < 0.70 and RPD < 1.75 (Malley et al., 2004).
52
0 20 40 60 80 100
0
20
40
60
80
100
(a) M3P_Col (mg kg-1
)
y = 0.31x + 28.86
Rv
2 = 0.25, RPD = 1.15, LR
0 20 40 60 80 100 120
0
20
40
60
80
100
120
y = 0.27x + 39.55
Rv
2 = 0.24, RPD = 1.15, LR
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
(c) Cp (mg L-1
)
y = 0.41x + 0.18
Rv
2= 0.45, RPD = 1.33, LR
400 600 800 1000 1200 1400
400
600
800
1000
1200
1400
(d) TP (mg kg-1
)
y = 0.73x + 170.98
Rv
2 = 0.75, RPD = 1.98, MU
0 5 10 15 20 25 30
0
5
10
15
20
25
30
(e) Annual P-uptake (kg ha-1
)
y = 0.60x + 6.67
Rv
2 = 0.49, RPD = 1.37, LR
(b) M3P_ICP (mg kg-1
)
-30 -20 -10 0 10 20 30 40 50
-30
-20
-10
0
10
20
30
40
50
(f) Annual P-budget (kg ha-1
)
NIR
S P
red
icte
d V
alu
es
Measured Values
y = b x + ay = x
y = 0.24x + 7.03
Rv
2 = 0.16, RPD = 1.06, LR
53
Figure 3-1 NIRS predicted values against measured values of (a) soil P content extracted
using the Mehlich 3 method and analysed by colorimetry (M3P_Col); (b) soil P content
extracted using the Mehlich 3 method and analysed by ICP (M3P_ICP); (c) soil P content
extracted with water and analysed by colorimetry (Cp); (d) total soil P; (e) annual timothy
crop P-uptake, and; (f) annual P-budget. Based on validation statistics reported here and in
Table 3.2, NIRS predictions were considered moderately useful (MU) when
0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25, and less reliable (LR) when Rv
2 < 0.70 and
RPD < 1.75 (Malley et al., 2004).
54
Prédiction du phosphore organique du sol par la spectroscopie dans le
proche infra-rouge
Dans le chapitre 3, nous avons démontré que le phosphore extrait à la solution du
Mehlich-3; méthode de référence au Québec, et à l’eau ne sont pas prédictibles par la
spectroscopie dans le proche infra-rouge (SPIR). Le P total du sol, par contre, est
prédictible par cette technique. Nous avons présumé que la corrélation du P à la matière
organique du sol affecterait le potentiel de le prédire par la SPIR, et nous avons conclu que
les modèles de prédiction obtenus dans cette étude pour un sol sableux-limoneux
podzolique devraient être validés dans d’autres sites de textures différentes. De ce fait, nous
avons évalué, dans le chapitre 4, le potentiel de la SPIR à prédire le P organique, PT, PM3
et Al, Fe, Mg et Mn extraits selon la méthode Mehlich-3 pour des Mollisols loameux et
argileux-loameux riches en P organique étant donné qu’ils étaient sous semis direct.
55
CHAPITRE IV: PREDICTING SOIL ORGANIC PHOSPHORUS
USING NEAR-INFRARED REFLECTANCE SPECTROSCOPY
Abdi D.a,c, Cade-Menun B.J.c, Ziadi N.b, Tremblay G.F.b, and Parent L.É.c
aAgriculture and Agri-Food Canada, Soils and Crops Research and Development Centre,
2560 Hochelaga Boulevard, Québec, QC, Canada G1V 2J3.
bAgriculture and Agri-Food Canada, Semiarid Prairie Agricultural Research Centre, P.O.
Box 1030 Swift Current, SK, Canada, S9H 3X2.
cDepartment of Soils and Agri-Food Engineering, Université Laval, Québec, QC, Canada
G1K 7P4.
sera soumis à Geoderma
56
4.1 RÉSUMÉ
La mesure du phosphore organique (Po) du sol se fait jusqu’à date par différence
entre le phosphore total et le phosphore inorganique. L’objectif de cette étude est d’évaluer
le potentiel de la spectroscopie dans le proche infrarouge (SPIR) à prédire (i) le Po total, et
(ii) d’autres propriétés chimiques du sol [P total (PT)], matière organique (MO), et P, Al,
Fe, Ca, Mg et Mn extraits au Mehlich-3]. Des échantillons du sol (n = 360) ont été prelevés
d’un site expérimental sous semis direct à court et à long terme en Saskatechwan. Les
équations de calibration de la SPIR ont été développées avec 80% des échantillons alors
que 20% a été utilisé pour la validation. Les résultats ont démontré que les prédictions du
Po ont été acceptables pour l’ensemble des échantillons et pour le site sous semis direct à
long terme, et fiables pour le site sous semis direct à court terme. Les prédictions ont été
acceptables pour PM3, Fe, et Mg, fiables pour la MO, et non acceptables pour PT, Al et Mn
pour les deux sites. Cette étude démontre que la SPIR est une technique prometteuse pour
quantifier le Po dans les Mollisols.
57
4.2 ABSTRACT
To date, there is no direct method to quantify the total concentration of organic
phosphorus (OP) in soils. Near-infrared reflectance spectroscopy (NIRS) is a direct, rapid,
inexpensive, and accurate analysis technique for a wide variety of materials and it is
increasingly used in soil science. The aim of this study was to examine the potential of
NIRS to predict (i) total soil OP, and (ii) other soil chemical properties [total P (TP),
organic matter (OM), and Mehlich-3 extractable P, Al, Fe, Ca, Mg and Mn]. Soil samples
(n = 360) were taken from an experimental site near Indian Head, SK, Canada, from short-
term (8 yr, n = 180) and long-term (31 yr, n = 180) conservation tillage plots of a field pea ̶
spring wheat rotation receiving five P fertilizer rates annually. Samples were collected at
three soil depths (0-7.5, 7.5-15, and 15-30 cm). Calibration NIRS equations were developed
using 80% of the soil samples and the partial least squares regression while the remaining
20% of samples were used for validation. The predictive ability of NIRS was evaluated
using the coefficient of determination of validation (RV2) and the ratio of standard error of
prediction to standard deviation (RPD). Results show that NIRS predictions for total OP
were classified as moderately useful (0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25) for the total
soil sample set and for the long-term no-till set, and were moderately successful
(0.80 ≤ Rv2 < 0.90 and 2.25 ≤ RPD < 3.00) for the short-term no-till set. Predictions were
moderately useful for Mehlich-3 P, Fe, and Mg, and successful for OM, but were not
acceptable for TP, Al, and Mn. This study demonstrated that NIRS is a promising analysis
technique for OP.
58
4.3 INTRODUCTION
Phosphorus (P) occurs in inorganic and organic forms in soils. Inorganic P as
orthophosphate (HPO42- or H2PO4
-) is the directly available form for plant or microbial
uptake. Nevertheless, organic P (OP) is widely found in the natural environment and acts as
a resource of long-term bio-available P (Oehl et al., 2001). Despite this, its role, mobility
and bio-availability are poorly understood due to analytical difficulties. Indeed, total OP
cannot be estimated directly, and instead is determined indirectly as the difference between
total P (TP) and inorganic P (IP) by ignition (Saunders and Williams, 1955) or extraction
(Hedley et al., 1982; Tiessen and Moir, 2007). These procedures involve treating soil with
several chemical reagents and are time consuming, and there is potential for error in the
steps for TP and IP, increasing the potential for errors in OP measurement.
Near infrared reflectance spectroscopy (NIRS) is a rapid, inexpensive, accurate and
environmentally friendly technique. It is used to predict the concentration of the soil
attributes of interest from an empirical model developed based on complex spectra of a
subset of soil samples. Hence, it could be a good alternative to routine soil analysis
methods. The NIRS spectra are a result of the radiation absorbed by various chemical
bonds (e.g. C-H, C-C, and O-H) found in soil constituents (Miller, 2001). Using a single
spectrum, many soil attributes can be predictable.
Some soil constituents, such as organic matter (OM), carbon, nitrogen, and pH can
be successfully predicted by NIRS (Cozzolino and Morón, 2006; Brunet et al., 2007)
because they have a theoretical basis for NIRS prediction and are considered to be primary
properties (Chang et al., 2001). Others soil constituents related to these properties, such as
aluminium (Al), iron (Fe), P, potassium (K), magnesium (Mg), manganese (Mn), calcium
(Ca), and copper (Cu) may be predictable by NIRS (Chang et al., 2001). Abdi et al. (2012)
found that available P determined using Mehlich-3 and water extraction methods for
gravelly sandy soil samples were not accurately predicted by NIRS (R2 < 0.70 and RPD <
1.75). However, NIRS prediction was moderately useful for total P (0.70 ≤ R2 < 0.80 and
1.75 ≤ RPD < 2.25). They explained these results by the fact that total P contains a certain
proportion of OP that is related to OM (Turner et al., 2005). The NIRS prediction accuracy
can be also related to soil texture type (Chang et al., 2001). To our knowledge, the NIRS
prediction potential of soil OP has not previously been assessed.
59
The objectives of this study were to assess the potential for NIRS to predict (i) total
OP, and (ii) other chemical attributes: TP, OM, and Mehlich-3 extractable P, Al, Fe, Ca,
Mg, and Mn for loam and clayey loam soil samples.
4.4 MATERIALS AND METHODS
4.4.1 Experimental site description
The experimental site description is provided in detail in Lafond et al. (2011), and a
brief summary follows. The experiment was located approximately 19 km south-east of
Indian Head, Saskatchewan, Canada (50.42° N, 103.58° W), on short-term no-till (ST-NT;
8 yr) and long-term no-till (LT-NT; 31 yr) plots. The soil of the LT-NT field was loamy
with a pH of 6.8, while the soil of the ST-NT field was clayey loam with a pH of 7.3. The
soil type for both fields was Orthic Black Chernozem. A continuous cropping no-till system
has been established since 1978 on the LT-NT field, and since 2001 on the adjacent ST-NT
field. Prior to this, the ST-NT field has been managed using a fallow-crop system involving
extensive tillage to 10 cm soil depth. The experimental design was a split-plot with crop
rotation (pea and spring wheat) as main plots and five P treatments (0, 11, 22, 33, and 45 kg
P2O5 ha-1) as sub-plots. Both crops were present each year, and there were two replicates
for each combination of crop, tillage and fertilizer treatment (40 plots total). Soils from this
no-till experimental site were selected for this study due to the higher soil OP content.
4.4.2 Soil sampling and analysis
Soil samples were collected in the fall of 2008 (n = 120) and in the spring and the
fall of 2009 (n = 240) at three depths (0 - 7.5, 7.5 - 15.0, and 15.0 – 30.0 cm), air-dried,
sieved, and ground (< 2 mm). Total P and OP concentrations were determined for the
whole soil sample set (n = 360). Total P was extracted using the wet acid digestion method
of Parkinson and Allen (1975). Briefly, 0.5 g of soil was mixed with 3.75 mL of H2SO4 and
3 mL of digestion solution (175 mL H2O2 + 0.21 g Se + 7 g LiSO4.H2O), and digested at
360°C for 2.5 h. Total OP was determined by the difference between 0.5 M H2SO4
extractable P in a 0.5 g soil sample ignited at 550° C and an unignited sample according to
Saunders and Williams (1955). Concentrations of TP and OP were measured in the soil
samples extracts by the colorimetric molybdate blue method (Murphy and Riley, 1962).
Soil OM was determined by loss on ignition for soil samples (n = 90) collected in 2008
60
from LT-NT field where pea was grown, and from ST-NT fields where pea and spring
wheat were grown. Soil P available to plants and exchangeable Al, Fe, Ca, Mg, and Mn
were determined in samples collected in 2008 (n = 120) by ICP emission spectroscopy after
a Mehlich 3 extraction (Mehlich, 1984). Briefly, 2.5 g of air-dried soil was mixed with 25
mL of a Mehlich 3 solution (0.25 M NH4NO3 + 0.015 M NH4F + 0.001 M EDTA + 0.2 M
CH3COOH + 0.013 M HNO3 buffered at pH 2.3), shaken for 5 min, and then filtered
through Whatman no. 42 paper.
4.4.3 Near-infrared reflectance spectroscopy spectrum acquisition
The absorbance of soil samples [log (1/R), where R is the reflectance] was
measured in the visible and near-infrared regions between 400 and 2498 nm at 0.5 nm
intervals using a NIRSTM DS 2500 monochromator Instrument (Foss NIRSystems Inc.,
Silver Spring, MD), with the transport cup containing approximately 25 mL of soil sample.
Each spectrum was the average of 16 co-added scans. To take into account the operator
error, a repeatability file was created by collecting 20 spectra for one randomly selected soil
sample.
4.4.4 Spectral pre-treatment
Prior to calibration, different pre-treatments were selected to improve calibration
models, including the critical T-statistics values of 2.0 and 2.5 for T-outlier detection, and
the following math treatments: 1-16-16-1, 2-32-24-1, 1-20-20-1, 2-20-20-1, 1-40-40-1, 2-
40-40-1. These four numbers (i.e., 1-16-16-1) are derivative treatments, the first indicating
the order of the derivative, the second, the gap over which the derivative is was calculated,
the third, the number of the smoothing points, and the last, the second smooth. Scatter
correction with standard normal variate and detrending (SNVD) was applied to all spectra
using the WinISI IV software to reduce scatter and particle size effect, and to remove linear
or curvilinear trend of each spectrum (Barnes et al., 1989). Trying to find the best NIRS
prediction accuracy for TP and OP, calibration equations were developed separately for
LT-NT plus ST-NT, LT-NT, and ST-NT soil sample sets.
4.4.5 Calibration, cross-validation and validation
Calibration models of soil properties have been developed using the modified partial
least squares regression (PLSR) method of WinISI IV with 80% of each soil sample set
61
randomly selected. The remaining 20% of soil samples were used as external validation set.
Cross-validation was performed using four groups from the calibration set to avoid over-
fitting the calibration model (Nduwamungu et al., 2009a). To select the best calibration
model, two criteria were simultaneous used: the low standard error (SE) and high
coefficient of determination (R2) in all cross-validations and in validation sets (Van Vuuren
et al., 2006).
Validation of generated calibration models was performed using WinISI IV by
comparing predicted and reference data. Calibration accuracy, i.e., closeness to reference
data, was assessed as described in Abdi et al. (2012), based on the coefficient of
determination of validation (Rv2) and the ratio of standard error of prediction to standard
deviation (RPD), which is the standard deviation of samples in the validation set (SD)
divided by the standard error of prediction corrected for the bias (SEP(C)) [RPD =
SD/SEP(C)]. Calibration equations were considered to be excellent when Rv2 > 0.95 and
RPD > 4.00; successful when 0.90 ≤ Rv2 ≤ 0.95 and 3.00 ≤ RPD ≤ 4.00; moderately
successful when 0.80 ≤ Rv2 < 0.90 and 2.25 ≤ RPD < 3.00; moderately useful when
0.70 ≤ Rv2 < 0.80 and 1.75 ≤ RPD < 2.25; and less reliable when Rv
2 < 0.70 and
RPD < 1.75 (Malley et al., 2004). The coefficient of variation (CV) in the reference set was
defined as the SD divided by the mean of chemical values, whereas the coefficient of
variation in the calibration set was computed as the ratio of standard error of calibration
(SEC) to the mean of calibration data (Williams, 2001).
4.5 RESULTS AND DISCUSSION
4.5.1 Soil reference data
Descriptive statistics for soil reference properties are provided in Tables 4.1 and 4.2.
The OP concentration was relatively variable in the soil samples collected from LT-NT and
ST-NT fields (CV = 30.8%). This variability was lower in LT-NT experiment (CV =
26.5%) and higher under ST-NT management (CV = 34.2%). Soil total P content and
variability were lower in both fields and were comparable between the two no-till systems.
For the Mehlich-3 extracted soil nutrients, their coefficients of variation decreased in the
order of P > Ca > Al > Mg > Mn > Fe. Soil organic matter content was higher (> 5 %) and
relatively homogenous distributed in both experimental sites (22.8%).
62
4.5.3 Spectral pre-treatment, calibration, and prediction of soil organic P
The best calibration model for OP measured in the whole soil sample set (LT-NT and
ST-NT; n = 360) was developed with the math treatment of 2-20-20-1 and the critical T
value of 2.0 (Table 4.3). Thirty four soil samples were excluded from this model, indicating
that 88% of soil samples remained in the calibration set. Based on the relatively low
standard error of calibration (26 mg kg-1) and cross-validation (34 mg kg-1) sets, the
medium coefficient of variation (29%), and the high coefficients of determination of
calibration (Rc2 = 0.89) and cross-validation (1-VR = 0.82) equations, the performance of
NIRS calibration for OP was considered acceptable. The validation of the calibration model
also showed acceptable coefficient of determination (Rv2 = 0.75) and RPD (2.01), with a
moderately useful prediction of OP for these soil samples. This NIRS prediction
performance might be related to organic bonds such as C-C and C-H (Miller, 2001;
Viscarra Rossel et al., 2006) in organic forms of P.
The OP calibration models were generated with the math treatment of 1-20-20-1 and T
value of 2.5 for the LT-NT sample set and with the math treatment of 1-40-40-1 and T
value of 2.0 for the ST-NT sample set. The first model was based on 95% of the whole
sample set of calibration (n = 180), and the second one on 88% of the sample set free of
outliers. Near-infrared reflectance spectroscopy prediction of OP in LT-NT soil samples of
the validation set (n = 36) was moderately useful (Rv2 = 0.70, RPD = 1.81; Fig. 4.1).
However, the NIRS prediction of OP in the ST-NT soil sample of the validation set (n =
36) was moderately successful (Rv2 = 0.88, RPD = 2.49; Fig. 4.1). This improvement of
NIRS prediction’s accuracy could be attributed to the higher OP mean concentration and
coefficient of variation in the ST-NT compared to the LT-NT soil sample calibration set.
According to Dardenne et al. (2000), a soil constituent with a wide dispersion is likely more
easily predictable by NIRS.
4.5.4 Spectral pre-treatment, calibration, and prediction of soil total and Mehlich-3 P
The calibration equation generated to predict TP concentration for the whole set of soil
samples (n = 360, Table 4.3) showed low coefficients of determination (RC2 = 0.77) and
variation (CV = 14%), which resulted in low coefficients of determination in cross-
validation (1-VR = 0.68) and validation (Rv2 = 0.60) models, and low RPD (1.54). Hence,
63
the NIRS prediction performance was not acceptable for soil TP. The same result was
obtained for the NIRS prediction of TP concentration in LT-NT (Rv2 = 0.51, RPD = 1.34)
and in ST-NT (Rv2 = 0.44, RPD = 1.34) soil samples. Abdi et al. (2012) found a moderately
useful NIRS prediction accuracy (Rv2 = 0.75, RPD = 1.98) for TP in 192 gravely sandy
loam soils, while Bogrekci and Lee (2005) successfully predicted TP (Rv2 = 0.92) in 150
fine sandy soil samples using NIRS. Thus, NIRS prediction performance for TP may be
related to soil texture.
The R2 in calibration (0.83) and validation (0.72), and the RPD value of 1.86 for M3-P
(Table 4.4) indicated that the NIRS prediction was moderately useful. Conversely, M3-P
was previously found to be not predictable by NIRS in 192 gravely sandy loam soil
samples (Abdi et al., 2012) and in 150 clayey soil samples (Nduwamungu et al., 2009b). A
possible explanation for these contradictory results is the higher value of coefficient of
variation for M3-P found in our study (CV = 119%) compared to the low values of 48 and
61% respectively found in the studies of Abdi et al. (2012) and Nduwamungu et al.
(2009b).
4.5.5 Spectral pre-treatment, calibration, and prediction of soil organic matter and
Mehlich-3 nutrients
Statistics of calibration, cross-validation, and validation for total set of soil extracted
Mehlich-3 constituents and organic matter are listed in Table 4.4. Soil organic matter was
successfully predicted by NIRS (Rv2 = 0.91, RPD = 3.02). According to Chang et al.
(2001), organic matter is considered primary property and has a theoretical basis for NIRS
prediction, which was well documented in the literature (St-Luce et al., 2014, Stevens et al.,
2013; Brian and Daniel, 2012). Near-infrared reflectance spectroscopy prediction for Ca
extracted from the whole set of loam and clay loam soil samples was moderately successful
(Rv2 = 0.86, RPD = 2.25). Moderately useful NIRS predictions were found for Fe and Mg
(Rv2 = 0.78, RPD = 2.1). However, Al and Mn were poorly predicted (Rv
2 < 0.70;
RPD < 1.75). These NIRS predictions were lower than in the study by Abdi et al. (2012)
except for Fe which was more dispersed in this study. Conversely, the range of Mn was
smaller in our study. Thus, it appears that the range of reference data may affect the
performance of NIRS calibration model as suggested by Dardenne et al. (2000).
64
4.6 CONCLUSION
This work showed that NIRS prediction potential for total organic P determined for
360 loam and clayey loam soils samples was moderately useful. This prediction precision
was improved to moderately successful for OP determined in clay loam soil probably due
to higher content and coefficient of variation of reference data. The NIRS predictions were
considered reliable for soil Mehlich-3 extracted P, Fe, Ca, and Mg, and not acceptable for
TP, Al, and Mn. We conclude that NIRS can be a cost-effective, time-saving and accurate
alternative technique for soil OP. Further studies are needed to evaluate the potential of
NIRS for predicting the chemical forms of soil OP.
65
4.7 REFERENCES
Abdi, D., Tremblay, G.F., Ziadi, N., Bélanger, G., and Parent, L.-E., 2012. Predicting soil
phosphorus-related properties using near-infrared reflectance spectroscopy. Soil Sci.
Soc. Am. J. 76: 2318–2326.
Barnes, R.J., Dhanoa, M.S., and Lister, S.J., 1989. Standard normal variate transformation
and detrending of near-infrared diffuse reflectance spectra. Appl. Spectrosc. 43:
772-777.
Bogrekci, I., and W.S. Lee. 2005. Spectral phosphorus mapping using diffuse reflectance of
soils and grass. Biosystems. Eng. 91:305–312.
Brian, K. N. and Daniel, J.A. 2012. Near Infrared Reflectance-Based Tools for Predicting
Soil Chemical Properties of Oklahoma Grazinglands. Agr. J. 104: 1122-1129.
Brunet, D., Barthès, B.G., Chotte, J.-L. and Feller, C. 2007. Determination of carbon and
nitrogen contents in Alfisols, Oxisols and Ultisols from Africa and Brazil using NIRS
analysis: Effects of sample grinding and set heterogeneity. Geoderma, 139: 106–117.
Chang, C.W., D.A. Laird, M.J. Mausbach, and Hurburgh, C.R.Jr. 2001. Near-infrared
reflectance spectroscopy – Principal components regression analyses of soil
properties. Soil Sci. Soc. Am. J. 65: 480–490.
Cozzolino, D., and Morón, A. 2006. Potential of near-infrared reflectance spectroscopy and
chemometrics to predict soil organic carbon fractions. Soil Tillage Res. 85: 78–85.
Dardenne, P., Sinnaeve, G. and Baeten, V. 2000. Multivariate calibration and chemometrics
for near infrared spectroscopy: which method? J. Near Infrared Spectrosc. 8: 229-
237.
Lafond, G.P.,Walley, F., Maya, W.E., and Holzapfel, C.B. 2011. Long term impact of no-
till on soil properties and crop productivity on the Canadian prairies. Soil Till. Res.
117: 110–123.
Malley, D.F., Ben-Dor, E. and Martin, P.D. 2004. Application in analysis of soils. p. 729–
84. In Roberts, C.A., J. Jr. Workman, and J.B. Reeves III (ed.) Near-infrared
Spectroscopy in Agriculture. Am. Soc. Agr., Crop Sc. Soc. Am., and Soil Sc. Soc.
Am., Madison, USA.
Miller, C.E. 2001. Chemical principles of near-infrared technology. p. 19–8. In Williams,
P.C., and K.H. Norris (ed.) Near Infrared Technology in the Agricultural and Food
Industries 2nd ed. American Association of Cereal Chemists, St. Paul, MN.
Murphy, J., and Riley, J.P. 1962. A modified single solution method for the determination
of phosphate in natural waters. Anal. Chim. Acta 27: 31–36.
Nduwamungu, C., Ziadi, N., Parent, L.É., Tremblay, G.F., Thuriès, L., 2009a.
Opportunities for and limitations of near infrared reflectance spectroscopy
applications in soil analysis: A review. Can. J. Soil Sci. 89: 531-541.
Nduwamungu, C., Ziadi, N. Parent, L.-É. and Tremblay, G.F. 2009b. Mehlich 3 extractable
nutrients as determined by near-infrared reflectance spectroscopy. Can. J. Soil Sci.
89: 579–587.
Oehl, F., Oberson, A., Sinaj, S. and Frossard, E. 2001. Organic Phosphorus Mineralization
Studies Using Isotopic Dilution Techniques. Soil Sci. Soc. Am. J. 65: 780–787.
Parkinson, J.A, and Allen, S.E. 1975. A wet oxidation procedure suitable for the
determination of nitrogen and mineral nutrients in biological material. Commun.
Soil Sci. Plant Anal. 6: 1-11.
66
St. Luce M., Ziadi N., Zebarth B.J., Grant C.A., Tremblay G.F, and Gregorich E.G. 2014.
Rapid determination of soil organic matter quality indicators using visible near
infrared reflectance spectroscopy. Geoderma 232–234: 449–458.
Saunders, W.M.H. and Williams, E.G. 1955. Observations on the determination of total
organic phosphorus in soils. J. Soil Sci. 6: 254–267.
Stevens A, Nocita M, To´th G, Montanarella L, and van Wesemael B. 2013. Prediction of
soil organic carbon at the european scale by visible and near infrared reflectance
spectroscopy. PLoS ONE 8(6): e66409. doi:10.1371/journal.pone.0066409
Tiessen, H. and Moir, J.O. 2007. Characterization of available P by sequential extraction.
In Soil Sampling and Methods of analysis. Carter, M.R., and Gregorich, E.G., (2nd
eds). pp. 293–306.
Turner, B.L., B.J. Cade-Menun, L.M. Condron, and S. Newman. 2005. Extraction of soil
organic phosphorus. Talanta, 66: 294–306.
Van Vuuren, J. A. J., Meyer, J. H. and Claassens, A. S. 2006. Potential use of near infrared
reflectance monitoring in precision agriculture. Commun. Soil Sci. Plan. 37: 2171-
2184.
Viscarra Rossel, R.A., Walvoort, D.J.J., McBratney, A.B., Janik, L.J., and Skjemstad, J.O.,
2006. Visible, near infrared, mid infrared or combined diffuse reflectance
spectroscopy for simultaneous assessment of various soil properties. Geoderma,
131: 59-75.
Williams, P.C. 2001. Implementation of near-infrared spectroscopy. p. 145–169. In
Williams, P.C., and K. Norris (ed.) Near-infrared Technology in the Agricultural
and Food Industries. 2nd ed. American Association of Cereal Chemists, St. Paul,
MN.
67
Tableau 4-1 Descriptive statistics for the soil organic (OP) and total (TP) P analysed for long- and short-term no-till (NT)
treatments.
OP TP
Treatment Long- and short-
term NT Long-term NT Short-term NT
Long- and short-term
NT Long-term NT Short-term NT
N 360 180 180
359 180 179
mg kg-1
Min 16.2 72.6 16.2
353.7 353.7 408.2
Max 492.7 406.1 492.7
758.1 703.0 758.1
Mean 274.9 269.1 280.6
556.7 543.3 570.1
SD 84.7 71.4 96.0
76.6 79.9 70.8
%
CV (%) 30.8 26.5 34.2
13.8 14.7 12.4
Tableau 4-2 Descriptive statistics for the soil Mehlich-3 extracted nutrients and organic matter for long- and short-term no-till
(NT) treatments.
Treatment M3-P M3-Al M3-Fe M3-Ca M3-Mg M3-Mn
OM
N 120 120 120 120 120 120 90
mg kg-1 %
Min 0.0 0.0 19.0 1046.1 290.0 7.5 3.5
Max 65.7 736.0 157.4 28643.7 4639.9 97.0 9.0
Mean 11.4 254.8 76.5 6295.2 849.7 46.5 6.0
SD 13.5 236.7 27.7 7180.9 619.9 25.5 1.4
%
CV 119.0 92.9 36.2 114.1 73.0 54.7 22.8
68
Tableau 4-3 Statistics of near-infrared reflectance spectroscopy calibration, cross-validation, and validation for soil (OP) and
total (TP) P analysed for long- and short-term no-till (NT) treatments.
Calibration
Cross-
validation Validation
T Math treat. Nc T-outliers Mean SD SEC R2c CV (%)
SECV 1-VR
Nv Mean SD SEP(C) R2
v RPD Predi.
Long- and short-term NT
OP (mg kg-1) 2.0 2,20,20,1 254 34 272 79 26 0.89 29
34 0.82
72 287 79 39 0.75 2.01 MU
TP (mg kg-1) 2.5 1,16,16,1 279 9 556 77 37 0.77 14
43 0.68
72 556 73 47 0.60 1.54 NA
Long-term NT
OP (mg kg-1) 2.5 1,20,20,1 137 7 265 70 31 0.80 26
32 0.79
36 292 61 34 0.70 1.81 MU
TP (mg kg-1) 2.5 1,16,16,1 143 1 540 83 36 0.80 15
45 0.71
36 552 63 47 0.51 1.34 NA
Short-term NT
OP (mg kg-1) 2.0 1,40,40,1 127 17 283 91 37 0.83 32
39 0.82
36 292 96 38 0.88 2.49 MS
TP (mg kg-1) 2.5 1,16,16,1 137 6 565 67 35 0.72 12
46 0.52
36 587 74 55 0.44 1.34 NA
69
Tableau 4-4 Statistics of near-infrared reflectance spectroscopy calibration, cross-validation, and validation for soil Mehlich-3
extracted nutrients and organic matter for long- and short-term no-till treatments.
Calibration
Cross-
validation Validation
T Math treat. Nc T-outliers Mean SD SEC R2c CV (%)
SECV 1-VR
Nv Mean SD SEP(C) R2
v RPD Predi.
M3-P (mg kg-1) 2.5 1,16,16,1 74 22 13 12 5 0.83 92 6.25 0.73 22 11 14 7.5 0.72 1.86 MU
M3-Al (mg kg-1) 2.0 2,32,24,1 67 29 324 215 77 0.87 66 109 0.74 19 328 253 180 0.71 1.41 NA
M3-Fe (mg kg-1) 2.5 2,32,24,1 81 15 78 23 13 0.67 30 16 0.50 24 72 27 13 0.78 2.08 MU
M3-Ca (mg kg-1) 2.0 2,32,24,1 76 20 3771 3257 951 0.92 86 1522 0.78 24 8136 8585 3815 0.86 2.25 MS
M3-Mg (mg kg-1) 2.5 1,16,16,1 85 11 710 345 106 0.96 49 144 0.82 24 914 548 258 0.78 2.12 MU
M3-Mn (mg kg-1) 2.5 1,16,16,1 80 16 52 24 12 0.77 46 13 0.70 24 41 25 15 0.66 1.67 NA
OM (%) 2.5 1,16,16,1 55 41 6 1.45 0.31 0.95 24 0.38 0.92 18 6.25 1.33 0.44 0.91 3.02 S
70
NIR
S P
redic
ted v
alu
es f
or
soil
org
anic
P (
mg k
g-1
)
Measured values of soil organic P (mg kg-1)
Figure 4-1 Near-infrared reflectance spectroscopy (NIRS) predicted vs. measured values
of soil organic P analysed for (a) long- and short term no-till, (b) long-term no-till,
and (c) short-term no-till treatments.
71
ANALYSE STATISTIQUE NON-BIAISÉE DES FORMES DU PHOSPHORE DU
SOL DÉTERMINÉES PAR RMN-31P
Dans les chapitres 3 et 4, nous avons démontré que les analyses chimiques des
formes de P total, de P inorganique et organique du sol, peuvent être remplacées en partie
par la spectroscopie dans le proche infrarouge, une nouvelle technique analytique de chimie
verte. L’objectif du chapitre 5 est de démontrer l’utilité du nouveau concept mathémathique
de l’analyse compositionnelle dans la caractérisation des formes chimiques du P par rapport
aux approches classiques. Nous avons utilisé deux bases de données publiées de deux sites
expérimentaux canadiens différents (Île-du-Prince-Édouard et Québec). Une de ces bases
de données provient du chapitre 6 de cette thèse.
72
73
CHAPITRE V: UNBIASED STATISTICAL ANALYSIS OF SOIL 31P-
NMR
Dalel Abdia,c, J. Barbara Cade-Menunb, Noura Ziadia and Léon-Étienne Parentc
aAgriculture and Agri-Food Canada, Soils and Crops Research and Development Centre,
2560 Hochelaga Boulevard, Québec, QC, Canada G1V 2J3.
bAgriculture and Agri-Food Canada, Semiarid Prairie Agricultural Research Centre, P.O.
Box 1030 Swift Current, SK, Canada, S9H 3X2.
cDepartment of Soils and Agri-Food Engineering, Université Laval, Québec, QC, Canada
G1K 7P4.
Highlights
- Soil 31P-NMR forms are compositional data
- Ordinary log transformation generated statistically erroneous results depending on
measurement scale
- Compositional analysis using clr and ilr transformations avoids statistical analysis
biases
Soumis à Geoderma, 2014
74
5.1 RÉSUMÉ
Les formes du phosphore (P) déterminées par la spectroscopie de la résonance magnétique
nucléaire du 31P sont des données compositionnelles. L’objectif de cette étude était de
comparer les analyses statistiques classiques et la nouvelle approche de l’analyse
compositionnelle des espèces de P. Deux bases de données publiées ont été utilisées.
L’analyse de la variance et la corrélation avec le pH du sol ont été conduites pour les
espèces du P exprimées en pourcentage du P total ou en concentrations brutes, ainsi qu’à
leurs transformations logarithmiques simples ou compositionnelles. Les valeurs statistiques
de F de l’analyse de variance et les coefficients de corrélation obtenus pour les données
brutes et celles ordinairement transformées exprimées en pourcentage ou en concentration
sont contradictoires. Cependant, les résultats statistiques obtenus avec les transformations
compositionnelles sont conformes quel que soit l’échelle de mesure. L’analyse
compositionnelle permet d’obtenir une analyse non biaisée des formes de P dans le sol.
Mots clés : espèces de P de 31P-RMN, analyse compositionnelle, log-ratio centré, log-ratio
isométrique, analyse non biaisée
75
5.2 ABSTRACT
Phosphorus (P) forms determined by 31P nuclear magnetic resonance spectroscopy (31P–
NMR) are compositional data. Because compositional data are intrinsically related to each
other within a closed pre-defined compositional space, a simple log transformation,
variable by variable, or any other transformation of the compositional variables may
produce statistically erroneous results. However, most studies analyze the P forms as single
components rather than parts of some whole such as total P (TP) or soil dry mass, leading
systematically to methodological biases and possibly conflicting interpretations.
Compositional data analysis using centred log-ratio (clr) or isometric log-ratio (ilr)
coordinates avoids such difficulties and preserves sub-compositional coherence in the
analysis. The objective of this study was to compare classical and compositional methods
for the statistical analysis of 31P–NMR P data expressed as proportions of TP or
concentrations relative to soil dry mass. Two published datasets were used. Analyses of
variance and regression analysis with soil pH were conducted on P species percentages
scaled on TP or as raw concentrations scaled on a soil dry-weight basis as well as their
ordinary log, centred log-ratios (clr) and isometric log-ratios (ilr). Contradictory F-statistics
values and coefficients of correlation with soil pH were obtained for the raw and ordinary
log transformed 31P-NMR P data expressed as proportions or concentrations, indicating
spurious correlations. In contrast, statistical results were the same regardless of the
measurement unit when P compound percentages were clr-transformed. Using orthogonal
ilr coordinates, 31P-NMR P data were correlated to soil properties and to each other and
synthesized into a multivariate distance without methodological bias. We conclude that the
variance and regression analyses of molecular P species are scale-dependent and that the
clr- and the ilr-transformations should be used to unbiasedly analyze the P fractions and
avoid conflicting interpretations.
Key words: 31P NMR-P species, compositional analysis, centred log-ratio, isometric log-
ratio, unbiased analysis.
76
5.3 INTRODUCTION
Phosphorus turnover in soils is controlled by a combination of interrelated chemical and
biological factors. At the top of system hierarchy, the C, N, S and P cycles are inter-
connected (McGill and Cole, 1981; Stevenson, 1986). Understanding relationships among
P species within the P cycle can advance our knowledge of soil P bioavailability. Inorganic
phosphorus (Pi), specifically orthophosphate, is the primary source of P for most
organisms; however, organic phosphorus (Po) is generally more abundant. The Po
encompasses a large spectrum of ionic and molecular entities that can be identified using
solution 31P-NMR spectroscopy. Based on the nature of C-P bonds, Po species are
classified into phosphonate, phosphate monoesters and phosphate diesters (Condron et al.,
2005).
When P species are determined by advanced spectroscopic techniques such as 31P-NMR
and P K-edge X-ray absorption near-edge structure spectroscopy (P-XANES), the
proportions of each species are determined as relative percentages rather than absolute
concentrations (e.g. Liu et al., 2013). Although few studies using these techniques analyze
replicated results for full statistical analyses, most studies attempt some simple analyses,
such as correlation to other soil parameters. However, the relative percentages determined
by 31P-NMR and P-XANES, being compositional data, are not normally distributed
(Aitchison, 1986), and thus require transformation prior to statistical analyses. This is often
done using simple log transformations.
When the data from these methods are analyzed statistically, the P species are
commonly expressed as proportions of total P (TP) and are thus constrained between 0 and
100%. Confidence intervals that reach below 0% or above 100% are physically impossible.
Such data are intrinsically related to each other while the proportion of one P species can be
computed by difference between 100% and the sum of the proportions of other P species
(Aitchison, 1986). Thus, a change in the percentage of any one P form must affect the
percentage of at least one other P form. In addition, P species do not follow a normal
distribution and statistical analyses may return different results depending on measurement
scale due to spurious correlations between proportions across scales (Reimann and
Filzmoser, 2000).
77
Simple log transformation or any other transformation of compositional variables may
generate statistically erroneous results (Filzmoser et al., 2009). Alternatively, additive log-
ratio (alr) and centred log-ratio (clr) transformations allow the projection of compositional
data into an unconstrained real space of scale-invariant variables (Aitchison, 1986).
However, the alr has been criticized as being subjective, because the results and
interpretation of univariate analysis depend on the choice of denominator (Bacon-Shone,
2011). This arbitrariness can be avoided using clr where the geometric mean of all
proportions is selected as the denominator. The clr simplifies the interpretation of the
transformed variables because one could think in terms of the original variable (Filzmoser
and Hron, 2009). Nevertheless, the clr cannot be used in multivariate analysis because the
matrix is singular (the D clr variates add up to 0; Bacon-Shone, 2011). The isometric log-
ratio (ilr) transformation (Egozcue et al., 2003) overcomes this problem using a sequential
binary partition (SBP) of balances with orthonormal basis.
The objectives of this paper were to (i) demonstrate the dependence of variance and
correlation analyses of 31P-NMR P data on measurement scale, and (ii) statistically analyse
P species using clr and ilr transformations. We hypothesized that the statistical analysis of
P forms expressed as proportions or concentrations using the classical approach of log
transformation could lead to conflicitng results, which could be avoided using clr and ilr
transformations.
5.4 MATERIALS AND METHODS
5.4.1 Datasets
We conducted univariate and multivariate analyses on two Canadian datasets reporting
soil P species. Abdi et al. (2014) analyzed the long-term effect of no-till [NT] or
mouldboard plowing [MP]) and P fertilization (0 and 35 kg P ha-1) on P species distribution
in the soil profile. The trial was conducted with a corn-soybean rotation in Québec, Canada
since 1992. The soil is a clay loam (claey, mixed, mesic Typic Humaquept). Thirty six soil
samples were collected at three depths (0-5 cm, 5-10 cm and 10-20 cm) and analyzed by
solution 31P-NMR spectroscopy. For the second study, Cade-Menun et al. (2010) examined
the P forms in a long-term experiment that compared MP to NT systems in Prince Edward
Island, Canada. The experiment was established in 1985 on sandy loam soil (Orthic
78
Podzol). Soil was sampled at six depths (0-5 cm, 5-10 cm, 10-20 cm, 20-30 cm, 30-40 cm
and 40-60 cm) and was also analyzed for P species.
5.4.1.1 Compositional data transformations
5.4.1.1.1 Centred log-ratio transformation
The ith clr is computed as follows (Aitchison, 1986):
(1)
where is the ith component and is the geometric mean across proportions.
The Euclidean distance between any two compositions, called the Aitchison distance, is
computed as follows:
(2)
where is the ith clr of the reference composition.
5.4.1.1.2 Isometric log-ratio transformation
The ilr is the log ratio between geometric means of two non-overlapping subsets of parts
(tagged with + and - signs) called orthonormal balances and computed as follows (Egozcue
et al., 2003):
(3)
where ilri is the ith balance between two sub-compositions, i Є [1, D-1], and are
number of components at numerator and denominator, respectively, and are
geometric means of components in subsets. Similarly to clr, ilr can scan the real space (±∞)
and is scale-invariant (the ratio between components of sets of components is a way to
eliminate the unit or scale of measurement). An advantage of ilr over clr is that the ilr
transformations have matrix rank of D-1. As a result, the ilr values associated with pre-
defined sub-compositions do not change if composition includes more sub-compositions.
79
The choice of non-overlapping subsets to compute geometric means is formalized by a
sequential binary partition (SBP). Orthogonal coefficients are computed as
allowing ilrs to be geometric coordinates in the Euclidean space.
The Aitchison distance A between two equal-length compositions x (diagnosed) and y
(reference or control) is computed as follows (Egozcue and Pawlowsky-Glahn, 2006):
(4)
where I is identity matrix and T indicates a transposed matrix.
The Mahalanobis distance M between two equal-length compositions x (diagnosed)
and y (reference or control) is computed as follows:
(5)
where COV is the covariance matrix.
5.4.1.1.3 Choice of SBP
Inorganic P (Pi) is a mineral product or the product of Po hydrolysis that can be
converted back to Po forms by soil microorganisms (Baldwin et al., 2005). The Pi and Po
species can be connected to each other using a mobile-fulcrums-buckets design balancing
subsets of components a priori defined in a SBP (Parent et al., 2012). The SBP can be
generated by default (Comas-Cufí and Thió-Henestrosa, 2011) or arranged hierarchically
according to some theory or practice (Egozcue and Pawlowsky-Glahn, 2006, Parent et al.,
2012). However, the choice of SBP does not influence the multivariate distance due to
balance orthogonality. The selected SBPs are presented in Tables 5.1 and 5.2 for processing
the Abdi et al. (2014) and the Cade-Menun et al. (2010) datasets, respectively.
The P compositional vector can be closed to total P or soil dry matter. In any event, the
first balances contrast the P fractions with filling values to the unit or scale of
measurement. If P fractions are scaled on soil dry matter, a filling value is computed
between the unit of measurement (e.g. kg of dry soil) and the sum of P fractions. If P
fractions are scaled on total P, a filling value is computed as residual P by difference
80
between total P and P fractions; residual P may also be quantified and P fractions added up
to total P. The filling value is needed to convert ilr means back to P fractions expressed on
a familiar scale or unit such as proportions. Thereafter, we contrasted Pi and Po forms.
Inorganic P is further partitioned into orthophoosphate and other Pi forms (pyro- and
polyphosphates) and the Po arranged into in subsets according to their bioavailability,
sorption and hydrolysis.
5.4.1.1.4 Ordinary logarithmic transformation
A log ratio is a log contrast or difference between two ordinary logarithmic
transformations. The ordinary logarithmic transformation returns an Euclidean distance that
differs from clr if geometric means differ between compositions (hence rejecting the ceteris
paribus assumption) as follows, where the star refers to some benchmark composition:
(6)
The ordinary logarithmic transformation returns a Mahalanobis distance that differs
from the ilr if the denominator differs between two compositions. As per example for a
dual /Po ratio, the computation is as follows after discarding the orthogonal coefficient
for simplification:
(7)
Where the difference between Pi in two samples is the difference between Pi
concentrations only if Po concentrations are the same (ceteris paribus assumption).
Similarly, difference between geometric means at numerator of two compositions is the
difference between geometric means at numerator only if geometric means at denominator
remains the same (ceteris paribus assumption). Equations 6 and 7 show that the ceteris
paribus equation is not applicable to compositional data. The Aitchison or Mahalanobis
distances across ordinary log-transformed proportions are thus inflated by compositional
discrepancies at denominator.
81
5.4.1.2 Statistical analysis
The 31P-NMR P forms were expressed as percentages of TP or as raw concentrations on
a soil dry-weight basis to show the influence of measurement scale on the results of
statistical analyses. For comparison, analyses of variance (ANOVA) were conducted using
clr-transformed data. Soil pH (0.01 M CaCl2) was correlated with P species percentages of
TP or as raw concentrations on a soil dry-weight basis as well as ordinary log [log (n+1)]
and isometric log-ratios to show spurious correlations. Compositional and statistical
analyses were conducted using R “compositions” (van den Boogaart et al., 2011). Rounded
zeroes for data below the detection limit for 31P-NMR were replaced according to Martín-
Fernández et al. (2003). The effects of tillage system, P fertilization and soil depth, and
their interactions were tested using Proc Mixed of SAS (SAS Institute, 2001).
5.5 RESULTS AND DISCUSSION
5.5.1 Biased analysis of variance
As a result of scale dependency and data redundancy, ANOVA may return conflicting
values for the significance of treatment effects and treatment means. Tables 5.3 and 5.4
show that treatment effects as well as their interactions differed considerably in terms of
significance whether the ANOVA was conducted using ordinary logarithmic
transformation of raw concentrations of P species or their proportions relative to total soil P
for both the Abdi et al. (2014) and Cade-Menun et al. (2010) studies.
In the Abdi et al. (2014) study, tillage, TP and FD effects were not significant
whatever the variable tested (Table 3). Among P, D, TD and TPD sources of variation,
there were 13 significant effects (P < 0.05) using log-transformed concentrations of P
species (mg P kg-1 dry soil), seven significant effects (P < 0.05) using log-transformed
proportions of P species (vs. Total P), and seven significant effects (P < 0.05) using centred
log-ratios. The interpretation of treatment effects thus differed considerably depending on
the transformation being used. For example, there was a significant effect of soil depth on
pyrophosphate concentration, but no such effect using log- and clr-transformed
pyrophosphate values. There were significant TD and TPD interactions using log-
82
transformed concentrations and proportions but no such significant using clr-transformed
orthophosphate values.
In the Cade-Menun et al. (2010) study (Table 5.4), there were five significant effects of
tillage treatments (P < 0.05) at five soil depths using log-transformed concentrations
compared with two significant effects (P < 0.05) using log-transformed proportions, and
five significant effects (P < 0.05) using clr-transformed values. Moreover, significant
effects were not always on the same P species at the same soil depth. For example, tillage
treatment significantly influenced (P < 0.05) the level of AMP in the 10-20 cm layer, of
myo-inositol hexakisphosphate in the 20-30 cm layer, and of -glycerophosphate in the 40-
60 cm layer, but these effects were not significant using ordinary log transformations.
Similar to the results in Tables 5.3 and 5.4, treatment means were also dependent on
the data transformation technique (Table 5.5). Moreover, the ordinary log back-transformed
treatment means were similar to the raw means (Table 5.6). In contrast, clr back-
transformed P species differed from the raw data indicating bias in raw and ordinary log-
transformed data.
Many studies investigating P cycling in soils in various environments collect replicate
field samples and conduct replicate analysis of various soil chemical parameters such as
pH, but do not conduct 31P-NMR or P-XANES analyses on these replicate samples, either
selecting a single sample or compositing the replicate samples into one sample for these
advanced analyses (e.g. Turner, 2008; Sato et al., 2009; Redel et al., 2011; Cheesman et al.,
2012; Liu et al., 2013; Wei et al., 2014). Other research groups use composite samples for
all aspects of their studies of P cycling (e.g. Turner and Engelbrecht, 2011; Vincent et al.,
2013; Hashimoto and Watanabe, 2014). The two data sets used here (Cade-Menun et al.,
2010; Abdi et al., 2014) are among the very few 31P-NMR studies to use individual samples
from replicate field plots for 31P-NMR to assess changes in soil forms with management
practices. In our opinion, all studies should try where ever possible to conduct 31P-NMR or
P-XANES analyses on replicate samples rather than composite samples, to allow the results
to be statistically analyzed. This will make the studies more scientifically rigorous, rather
than merely descriptive. However, the results from this current study clearly show that care
83
must be taken when transforming these results. We recommend using clr-transformed data
for statistical analysis of P forms determined by 31P-NMR or P-XANES.
5.5.2 Spurious correlations
It has long been recognized (Pearson, 1897) that the same measurement made on
different scales generates spurious correlations. More recently, Aitchison (1986) showed
that redundancy generates at least one negative correlation because any increase in one
proportion must be associated with a decrease in one or more proportions in a closed
system. When measurement scales vary from raw concentrations on dry soil basis and raw
proportions of total P to their log-transformed values, scale-dependent coefficients of
correlation can be measured by the discrepancies in correlation coefficient, sign and
significance (Table 5.7). Discrepancies are shown when correlating soil pH with (1) log-
transformed concentrations and proportions of residual P in the Abdi et al. (2014) study and
(2) orthophosphate concentrations and proportions and their log-transformed expressions in
the Cade-Menun et al. (2010) study. The pH is significantly correlated either negatively or
positively depending on measurement scale. This is why a scale-invariant expression that
avoids redundancy is required to conduct correlation analysis of compositional data.
These results raise concerns about the many studies that have correlated 31P-NMR and
P-XANES results to soil parameters such as pH with little or no data transformation (e.g.
Cheesman et al., 2012; Turner and Blackwell, 2013; Wei et al., 2014). Any relationships of
soil P forms to soil chemical parameters developed in these studies should be treated with
caution. We recommend that the authors of these and other similar studies reanalyze their
correlations after correctly transforming their data.
Because multivariate analyses are based on correlations between variables, using raw
concentrations or proportions or their log-transformation must distort the results of
multivariate analysis (Aitchison, 1986). The degree of distortion in multivariate distances
such as the Mahalanobis distance can be shown by an inflation of the Mahalanobis distance
using raw concentrations or proportions or their log-transformation (Fig. 5.1). As shown by
Eqs. 6 and 7, the distortion is attributable to the difference in the geometric means between
compositions. An unbiased multivariate distance can be obtained using ilr variables
because the ilr variables are orthogonal to each other in a Euclidean space, hence
84
measuring differences between compositions as straight lines. The ilr variables are the most
appropriate transformation to conduct multivariate analyses on compositional data
(Filzmoser et al., 2009). The ilr variables can be back-transformed into familiar units to
facilitate interpreting the results, solving for the D-1 ilr values and the sum of components
to the unit of measurement.
5.6 CONCLUSIONS
In this paper, we demonstrated that variance and correlation analyses of 31P-NMR P
data depended on measurement scale, which is to be expected for compositional data. The P
species could be statistically analyzed using clr transformations to facilitate interpreting
ANOVA results and ilr transformations for correlation and multivariate analyses. The clr
and ilr transformations avoid methodological biases in classical procedures that often lead
to conflicting interpretations using the same compositional vector but different scaling
procedures. The ilr transformation has considerable advantage over other data
transformation methods because P species can be arranged into interpretable orthonormal
balances according to some theory that connects the P species measured in the P cycle to
components of the C, N, and S cycles. There is a need for paradigm shift in studies on
elemental cycles like P considering that components of any cycle are connected to each
other in a coherent system analysis.
85
5.7 REFERENCES
Abdi, D., Cade-Menun, B.J., Ziadi, N., and Parent, L-É., 2014. Long-term impact of tillage
and P fertilization on soil P forms as deterimed by 31P-NMR spectroscopy. J.
Environ. Qual. doi:10.2134/jeq2013.10.0424.
Aitchison, J., 1986. The statistical analysis of compositional data, first ed. London, UK.
Bacon-Shone, J., 2011. A short history of compositional data analysis, in: Pawlowsky-
Glahn, V., Buccianti, A. (Eds.), Compositional data analysis: Theory and
Applications. John Wiley & Sons, New York. pp. 3-11.
Baldwin, D.S, Howitt. J.A., and Beattie, J.K., 2005. Abiotic degradation of organic
phosphorus compounds in the environment, in: Turner, B.J., Frossard, L., Baldwin,
D. (Eds.), Organic Phosphorus in the Environment. CABI Publishing, Oxfordshire,
pp. 75-88.
Cade-Menun, B.J., Carter, M.R., James, D.C., and Liu, C.W., 2010. Phosphorus forms and
chemistry in the soil profile underlong-term conservation tillage: A phosphorus-31
nuclear magnetic resonance study. J. Environ. Qual. 39: 1647-1656.
Cade-Menun, B.J., and Preston, C.M., 1996. A comparison of soil extraction procedures for 31P NMR spectroscopy. Soil Sci. 161: 770-785.
Cheesman, A.W., Turner, B.L., and Reddy, K.R., 2012. Soil phosphorus forms along a
strong nutrient gradient in a tropical ombrotrophic wetland. Soil Sci. Soc. Am. J.
76: 1496-1506.
Comas-Cufí, M., Thió-Henestrosa, S., 2011. and CoDaPack 2.0: a stand-alone, multi-
platform compositional software. Proceedings 4th International Workshop on
Compositional Data Analysis, Sant Feliu de Guíxols, Spain, 9-13 May 2011
(congress.cimne.com/codawork11/Admin/Files/FilePaper/p28.pdf).
Condron, L.M., Turner, B.L., and Cade-Menun, B.J., 2005. Chemistry and dynamics of soil
organic phosphorus. in: Sims, J.T. Sharpley, A.N. (Eds.), Phosphorus, Agriculture
and the Environment. Monograph no 46, Soil Science Society of America,
Madison, WI. pp. 87-121.
Egozcue, J.J., and Pawlowsky-Glahn, V., 2006. Simplicial geometry for compositional
data, in: Buccianti, A., Mateu-Figueras, G., Pawlowsky-Glahn, V. (Eds.),
Compositional Data Analysis in the Geosciences: From Theory to Practice. Special
Publications, 264, Geological Society, London. pp. 67–77.
Egozcue, J.J., and Pawlowsky-Glahn, V., Mateu-Figueras, G., Barceló-Vidal, C., 2003.
Isometric log-ratio transformations for compositional data analysis. Math. Geol. 35:
279-300.
Filzmoser, P., and Hron, K., 2009. Correlation analysis for compositional data. Math.
Geosc. 41: 905-919.
Filzmoser, P., Hron, K., and Reimann, C., 2009. Univariate statistical analysis of
environmental (compositional) data: Problems and possibilities. Sci. T. Environ.
407: 6100-6108.
86
Hashimoto, Y., and Watanabe, Y., 2014. Combined applications of chemical fractionation,
solution 31P-NMR and P K-edge XANES to determine phosphorus speciation in
soils formed on serpentine landscapes. Geoderma, 230-231, 143-150.
Liu, J., Yang, J., Cade-Menun, B.J., Liang, X., Hu, Y., Liu, C.W., Zhao, Y., Li, L., and Shi,
Y. 2013. Complementary phosphorus speciation in agricultural soils by sequential
fractionation, solution 31P nuclear magnetic resonance, and phosphorus K-edge X-
ray absorption near-edge structure spectroscopy. J. Environ. Qual. 42: 1763-1770.
Martin-Fernandez, J.A., Barcelo-Vidal, C., and Pawlowsky-Glahn, V., 2003. Dealing with
zeros and missing values in compositional data sets using nonparametric imputation.
Math. Geol. 35: 253–278.
McGill, W.B., and Cole, C.V., 1981. Comparative aspects of cycling of organic C, N, S and
P through soil organic matter. Geoderma 26, 267-286.
Parent, S.- É., Parent, L. E. Rozane, D. E. Hernandes, A., and Natale, W., 2012. Nutrient
balance as paradigm of plant and soil chemometrics, in: Issaka, R.N. (Eds.), Soil
Fertility. InTech Publishing, New York, pp. 83-114.
http://www.intechopen.com/books/soil-fertility.
Pearson, K., 1897. Mathematical contributions to the theory of evolution. On a form of
spurious correlation which may arise when indices are used in the measurement of
organs. Proceedings of the Royal Society of London, LX, pp. 489-502.
Redel, Y.D., Escudey, M., Alvear, M., Conrad, J., and Borie, F., 2011. Effects of tillage
and crop rotation on chemical phosphorus forms and some related biological
activities ina Chilean Ultisol. Soil Use Manage. 27: 221-228.
Reimann, C., and Filzmoser, P., 2000. Normal and lognormal data distribution in
geochemistry: death of a myth. Consequences for the statistical treatment of
geochemical and environmental data. Environ. Geol. 39: 1001-1014.
SAS Institute. 2001. The SAS system for Windows. Release 8.2. SAS Inst., Cary, NC.
Sato, S., Neves, E.G., Solmon, D., Liang, B., and Lehmann, J., 2009. Biogenic calcium
phosphate transformation in soils over millennial time scales. J. Soils Sediments, 9:
194-205.
Stevenson, F.J., 1986. Cycles of soils. Carbon, nitrogen, phosphorus, micronutrients, first
ed.Wiley-Interscience, New York.
Turner, B.L., 2008. Soil organic phosphorus in tropical forests: an assessment of the
NaOH-EDTA extraction procedure for quantitative analysis by solution 31P NMR
spectroscopy. Eur. J. Soil Sci. 59: 453-466.
Turner, B.L., and Blackwell, M.S.A., 2013. Isolating the influence of pH on the amounts
and forms of soil organic phosphorus. Eur. J. Soil Sci. 64: 249-259.
Turner, B.L., and Engelbrecht, B.M.J., 2011. Soil organic phosphorus in lowland tropical
rain forests. Biogeochem. 103: 297-315.
Turner, B.L., Cade-Menun, B.J., Condron, L.M., and Newman, S., 2005. Extraction of soil
organic phosphorus. Talanta, 66: 294-306.
87
van den Boogaart, K.G., Tolosana-Delgado, R., and Bren, M., 2011. Compositions:
Compositional data analysis, R Package Version 1.10 2.
http://cran.rproject.org/web/packages/compositions/compositions.pdf
Vincent, A.G., Vestergren, J., Gröbner, G., Persson, P., Schleucher, J., and Giesler, R.,
2013. Soil organic phosphorus transformations in a boreal forest chronosequence.
Plant Soil. 367: 149-162.
Wei, K., Chen, Z.H., Zhang, X.P., Liang, W.J., and Chen, L.J., 2014. Tillage effects on
phosphorus composition and phosphatase activities in soil aggregates. Geoderma,
217-218, 37-44.
88
Table 5-1 Sequential binary partition of soil 31P-NMR P species analyzed by Abdi et al. (2014).
ilr Orthoa Pyro Poly Phos Myo Neo Scyllo Gluc α-glyc β-glyc Nucl Chol DNA Res
P
r s
1 1 1 1 1 1 1 1 1 1 1 1 1 1 -1 13 1
2 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 0 3 10
3 1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 1 2
4 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 1 1
5 0 0 0 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 0 1 9
6 0 0 0 0 1 1 1 1 1 1 1 1 -1 0 8 1
7 0 0 0 0 1 -1 -1 -1 -1 -1 -1 -1 0 0 1 7
8 0 0 0 0 0 1 -1 -1 -1 -1 -1 -1 0 0 1 6
9 0 0 0 0 0 0 1 -1 -1 -1 -1 -1 0 0 1 5
10 0 0 0 0 0 0 0 1 -1 -1 -1 -1 0 0 1 4
11 0 0 0 0 0 0 0 0 1 -1 -1 -1 0 0 1 3
12 0 0 0 0 0 0 0 0 0 1 -1 -1 0 0 1 2
13 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 1 1 aOrthophosphate (Ortho), pyrophosphate (Pyro), polyphosphate (Poly), phosphonate (Phos), myo-inositol hexakisphosphate (Myo),
neo-inositol hexakisphosphate (Neo), scyllo-inositol hexakisphosphate (Scyllo), glucose-6 phosphate (Gluc), α-glycerophosphate (α-
Glyc), β-glycerophosphate (β-Glyc), nucleotides (Nucl), choline-phosphate (Chol), deoxyribonucleic acid (DNA), residual
phosphate (Res P).
r: number of components at numerator
s: number of components at denominator
89
Table 5-2 Sequential binary partition of soil 31P-NMR P species analyzed by Cade-Menun et al. (2010).
ilr Orthoa Pyro Poly Phos Myo Scyllo Gluc α-glyc AMP Mono1 Mono2 Mono3 Oth.Di DNA r s
1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 3 11
2 1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 1 2
3 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 1 1
4 0 0 0 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 10
5 0 0 0 0 1 1 1 1 1 1 1 1 -1 -1 8 2
6 0 0 0 0 1 1 -1 -1 -1 -1 -1 -1 0 0 2 6
7 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 1 1
8 0 0 0 0 0 0 1 1 1 -1 -1 -1 0 0 3 3
9 0 0 0 0 0 0 1 -1 -1 0 0 0 0 0 1 2
10 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 1 1
11 0 0 0 0 0 0 0 0 0 1 -1 -1 0 0 1 2
12 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 1 1
13 0 0 0 0 0 0 0 0 0 0 0 0 1 -1 1 1 aOrthophosphate (Ortho), pyrophosphate (Pyro), polyphosphate (Poly), phosphonate (Phos), myo-inositol hexakisphosphate (Myo),
scyllo-inositol hexakisphosphate (Scyllo), glucose-1 phosphate (Gluc), α-glycerophosphate (α-Glyc), adenosine-5-monophosphate
(AMP), orthophosphate monoesters regions 1, 2, or 3 (mono1, mono2, mono3), orthophosphate diesters other than DNA (Oth.Di),
deoxyribonucleic acid (DNA)
r: number of components at numerator
s: number of components at denominator
90
Table 5-3 ANOVA of the effect of tillage (T), P fertilization (P) and soil depth (D) on log-
and clr-transformed P compositions (Abdi et al., 2014). P species defined in Table 1.
P species Tillage (T) Fertilization
(P)
Depth
(D)
T x
P
T x
D
F x
D
T x P
x D
log (mg P kg-1 sol)
Orthophosphate ns * ** ns * † **
Pyrophosphate ns ns * † * ns ns
Polyphosphate ns ns ns ns ns ns ns
Phosphonate ns ** ns ns ns ns ns
Myo-IP6 ns ns ns ns ns ns *
Neo-IP6 ns ns † ns ns ns ns
Scyllo-IP6 ns ns * ns † ns ns
Glucose-6 phosphate ns ns ns ns ns ns ns
α-Glycerophosphate ns ns ns ns ns ns ns
β-Glycerophosphate ns ns ns ns ns ns ns
Nucleotides ns ns * ns * ns ns
Choline-phosphate ns † ns ns ns ns ns
DNA ns ns ns ns ** ns *
Residual phosphate † ns ns ns ns ns ns
log (% total P)
Orthophosphate ns † ** ns * ns *
Pyrophosphate ns ns ns ns ns ns ns
Polyphosphate ns ns ns ns ns ns ns
Phosphonate ns ns ns ns ns ns ns
Myo-IP6 ns ns ns ns ns ns ns
Neo-IP6 ns ns ns ns ns ns ns
Scyllo-IP6 ns ns ** ns † ns †
Glucose-6 phosphate ns ns ns ns ns ns ns
α-Glycerophosphate ns † † ns ns ns ns
β-Glycerophosphate ns † † ns ns ns ns
Nucleotides ns ns † ns * ns ns
Choline-phosphate ns * ns ns ns † ns
DNA ns ns † ns * ns †
Residual phosphate ns † ns ns ns ns ns
centred log-ratio transformation
Orthophosphate ns * * ns ns ns †
Pyrophosphate ns ns † † † ns ns
Polyphosphate ns ns ns ns ns ns ns
Phosphonate ns * ns ns ns ns ns
91
Myo-IP6 ns ns ns ns ns ns ns
Neo-IP6 ns ns ns ns ns ns ns
Scyllo-IP6 ns ns * ns † ns ns
Glucose-6 phosphate ns ns ns ns ns ns ns
α-Glycerophosphate ns † † ns ns ns ns
β-Glycerophosphate ns † † ns ns ns ns
Choline-phosphate ns † ns ns ns ns ns
Nucleotides ns ns * ns ** ns ns
DNA ns ns ns ns ** ns ns
Residual phosphate ns ns ns ns ns ns ns
† Significant at P < 0.1
* Significant at P < 0.05
** Significant at P < 0.01
ns, nonsignificant at the 0.10 level
92
Table 5-4 ANOVA of the effect of tillage and soil depth on log- and clr-transformed P
compositions (Cade-Menun et al., 2010). P species defined in Table 2.
P species 0-5 cm 5-10 cm 10-20 cm 20-30 cm 30-40 cm 40-60 cm
log (mg P kg-1 soil)
Orthophosphate ns * † ns ns ns ns
Pyrophosphate ns ns ns ns ns ns ns
Polyphsophate ns ns † ns ns ns ns
Phosphonate ns * ns ns ns ns ns
Myo ns * † † ns ns ns
Scyllo ns ns ns ns ns ns ns
Glucose-1 P ns ns ns ns ns ns ns
α-GlyceroP ns ns ns † ns ns ns
AMP ns * ns ns ns ns ns
Mono1 ns ns ns ns ns ns ns
Mono2 ns ns ns ns ns ns ns
Mono3 ns ns ns ns ns ns ns
OthDdi. ns * ns ns ns ns ns
DNA ns ns ns ns ns ns ns
log (% total P)
Orthophosphate ns ns ns ns ns ns ns
Pyrophosphate ns ns ns ns ns ns ns
Polyphsophate ns ns ns ns ns ns ns
Phosphonate ns ns ns ns ns ns ns
Myo ns * * † ns ns ns
Scyllo ns ns ns ns ns ns ns
Glucose-1 P ns ns ns ns ns ns ns
α-GlyceroP ns ns ns ns ns ns ns
AMP ns ns † ns ns ns ns
Mono1 ns ns ns ns ns ns ns
Mono2 ns ns ns † ns ns ns
Mono3 ns ns ns ns ns ns ns
Oth.Di. ns ns ns ns ns ns ns
DNA ns ns ns ns ns ns ns
centred log-ratio transformation
Orthophosphate ns ns ns ns ns ns ns
Pyrophosphate ns ns ns ns ns ns ns
Polyphsophate ns ns † ns ns ns ns
Phosphonate ns ns ns ns ns ns ns
Myo ns † † * ns ns ns
Scyllo ns ns ns ns ns ns ns
Glucose-1 P ns ns ns ns ns ns ns
93
α-GlyceroP ns ns ns ns ns *
AMP ns ns * ns ns ns
Mono1 ns ns † ns * ns
Mono2 ns ns ns † † ns
Mono3 ns ns ns ns ns ns
Oth.Di. ns * ns ns ns ns
DNA ns ns ns ns † †
† Significant at P < 0.1
* Significant at P < 0.05
** Significant at P < 0.01
ns, nonsignificant at the 0.10 level
94
Table 5-5 Treatment means of P species concentrations and proportions for main
effect of tillage in the Abdi et al. (2014) and Cade-Menun et al. (2010) studies.
P species Transformation
mg P kg-1
soil
log(mg P kg-1
soil) clr
% total
P
log(% total
P) clr
mg P kg-1 soil % total P
Abdi et al. (2014) study
Orthophosphate 260.2 2.4 2.5 41.8 1.6 2.5
Pyrophosphate 6.2 0.9 -1.3 1.0 0.3 -1.3
Polyphosphate 5.7 0.8 -1.6 0.7 0.2 -1.6
Phosphonate 17.6 1.3 -0.3 2.7 0.6 -0.3
Myo-IP6 61.6 1.8 1.0 9.9 1.0 1.0
Neo-IP6 24.5 1.4 0.1 4.0 0.7 0.1
Scyllo-IP6 24.4 1.4 0.1 3.9 0.7 0.1 Glucose-6
phosphate 14.9 1.2 -0.4 2.3 0.5 -0.4
α-Glycerophosphate 10.7 1.1 -0.8 1.6 0.4 -0.8
β-Glycerophosphate 21.4 1.3 -0.1 3.2 0.6 -0.1
Nucleotides 36.5 1.6 0.4 5.5 0.8 0.4
Choline-phosphate 13.5 1.2 -0.5 2.1 0.5 -0.5
DNA 10.7 1.1 -0.7 1.7 0.4 -0.7
Residual phosphate 118.8 2.1 1.7 19.5 1.3 1.7 Cade-Menun et al. (2010) study
Orthophosphate 503.4 2.7 3.6 68.9 1.8 3.6
Pyrophosphate 8.0 1.0 -0.7 1.0 0.3 -0.7
Polyphsophate 4.9 0.8 -0.9 0.7 0.2 -0.9
Phosphonate 5.7 0.8 -0.9 0.8 0.3 -0.9
Myo 98.1 2.0 1.9 13.0 1.1 1.9
Scyllo 19.5 1.3 0.3 2.6 0.6 0.3 Glucose-1
phosphate 6.1 0.9 -0.9 0.8 0.3 -0.9
α-Glycerophosphate 14.1 1.2 -0.1 1.8 0.4 -0.1
AMP 11.1 1.1 -0.3 1.4 0.4 -0.3
Mono1 5.7 0.8 -0.9 0.8 0.3 -0.9
Mono2 42.5 1.6 1.0 5.5 0.8 1.0
Mono3 5.1 0.8 -1.0 0.7 0.2 -1.0
Oth.Di. 5.6 0.8 -0.9 0.7 0.2 -0.9
DNA 9.9 1.0 -0.4 1.3 0.4 -0.4
95
Table 5-6 Back-transformed treatment means of P species concentrations and
proportions for main effect of tillage in the Abdi et al. (2014) and Cade-Menun et al.
(2010) studies.
P species Back-transformation
mg P kg-1
soil
log(mg P kg-1
soil) clr
% total
P
log(% total
P) Clr
mg P kg-1 soil % total P
Abdi et al. (2014) study
Orthophosphate 259.3 259.3 262.1 41.4 41.4 41.8
Pyrophosphate 6.6 6.6 6.0 1.0 1.0 1.0
Polyphosphate 5.7 5.7 4.5 0.9 0.9 0.7
Phosphonate 17.8 17.8 17.2 2.8 2.8 2.7
Myo-IP6 61.3 61.3 62.2 9.8 9.8 9.9
Neo-IP6 24.5 24.5 25.0 3.9 3.9 4.0
Scyllo-IP6 24.4 24.4 24.1 3.9 3.9 3.8 Glucose-6
phosphate 14.8 14.8 14.7 2.4 2.4 2.3
α-Glycerophosphate 10.4 10.4 10.1 1.7 1.7 1.6
β-Glycerophosphate 20.8 20.8 20.3 3.3 3.3 3.2
Nucleotides 35.9 35.9 34.3 5.7 5.7 5.5
Choline-phosphate 13.8 13.8 13.1 2.2 2.2 2.1
DNA 11.1 11.1 10.9 1.8 1.8 1.7
Residual phosphate 121.5 121.5 122.4 19.4 19.4 19.5
Cade-Menun et al. (2010) study
Orthophosphate 503.4 503.4 515.0 68.9 68.9 69.6
Pyrophosphate 8.0 8.0 7.1 1.0 1.0 1.0
Polyphsophate 4.9 4.9 5.6 0.7 0.7 0.8
Phosphonate 5.7 5.7 5.7 0.8 0.8 0.8
Myo 98.1 98.1 95.3 13.0 13.0 12.9
Scyllo 19.5 19.5 18.4 2.6 2.6 2.5 Glucose-1
phosphate 6.1 6.1 5.8 0.8 0.8 0.8
α-Glycerophosphate 14.1 14.1 13.1 1.8 1.8 1.8
AMP 11.1 11.1 10.2 1.4 1.4 1.4
Mono1 5.7 5.7 5.6 0.8 0.8 0.8
Mono2 42.5 42.5 37.5 5.5 5.5 5.1
Mono3 5.1 5.1 5.3 0.7 0.7 0.7
Oth.Di. 5.6 5.6 5.5 0.7 0.7 0.7
DNA 9.9 9.9 9.5 1.3 1.3 1.3
96
Table 5-7 Correlation between soil pH and P species expressed as raw or log-
transformed concentrations or proportions or as isometric log ratios (ilr). P species and
ilrs defined in Tables 1 (Abdi et al., 2014) and 2 (Cade-Menun et al., 2010).
P species mg P kg-1
soil
% total P Log (mg P kg-1
soil)
Log (% total
P)
Ilr
r r
Abdi et al. (2014) study
Orthophosphate -0.26 0.37† -0.32† 0.33* ilr1 -0.49**
Pyrophosphate 0.28 0.49† 0.32† 0.54* ilr2 0.53**
Polyphosphate -0.19 -0.01 -0.06 0.07 ilr3 -0.25
Phosphonate -0.44† -0.05 -0.43* -0.07 ilr4 0.17
Myo-IP6 -0.68† -0.37† -0.69* -0.36* ilr5 0.14
Neo-IP6 -0.64† -0.04 -0.64* -0.04 ilr6 -0.77**
Scyllo-IP6 -0.85† -0.45† -0.82* -0.49* ilr7 0.41*
Glucose-6 phosphate -0.67† -0.31* -0.65* -0.34* ilr8 0.62**
α-Glycerophosphate -0.79† -0.42† -0.83* -0.45* ilr9 0.05
β-Glycerophosphate -0.76† -0.64† -0.78* -0.66* ilr10 0.28
Nucleotides -0.43† -0.10 -0.48* -0.13 ilr11 0.38*
Choline-phosphate -0.91† -0.74† -0.89* -0.74* ilr12 -0.11
DNA 0.06 0.49† 0.05 0.51* ilr13 -0.54**
Residual phosphate -0.35† 0.39† -0.38* 0.38*
Cade-Menun et al. (2010) study
Orthophosphate 0.54* -0.66** 0.53** -0.65** ilr1 -0.28
Pyrophosphate 0.58* 0.42* 0.57** 0.42* ilr2 -0.59**
Polyphsophate 0.37 0.05 0.15 -0.02 ilr3 0.26
Phosphonate 0.68** 0.22 0.64** 0.22 ilr4 -0.27
Myo 0.72** 0.46* 0.67** 0.50* ilr5 0.36†
Scyllo 0.78** 0.64** 0.71** 0.66** ilr6 0.24
Glucose-1 phosphate 0.41† 0.09 0.47* 0.09 ilr7 -0.38†
α-Glycerophosphate 0.57* 0.39† 0.56** 0.42* ilr8 0.32
AMP 0.79** 0.70** 0.75** 0.73** ilr9 -0.51*
Mono1 0.43† -0.30 0.37† -0.30 ilr10 -0.49*
Mono2 0.85** 0.58** 0.71** 0.62** ilr11 -0.60**
Mono3 0.70** 0.64** 0.69** 0.65** ilr12 0.63**
Oth.Di. 0.65** 0.28 0.62** 0.29 ilr13 -0.06
DNA 0.56* 0.24 0.52* 0.24
† Significant at P < 0.1
* Significant at P < 0.05
** Significant at P < 0.01
97
Figure 5-1 Relationship between Mahalanobis distance from ilr with (a) ordinary log
transformed 31P NMR-P species concentration, and (b) raw of 31P NMR-P species
concentrations (data from Abdi et al., 2014).
98
Effet à long-terme du travail du sol et de la fertilisation phosphatée sur
les formes de P déterminées par RMN-31P
Le chapitre 5 a démontré l’importance du traitement des formes du P par la nouvelle
approche mathématique de l’analyse compositionnelle qui permet de générer des résultats
non biaisés, et par conséquent, des interprétations cohérentes. Nous présentons, dans
l’annexe, un modèle de balance que nous avons développé en utilisant les coordonnées du
log ratio isométrique pour décrire les relations entre des pools de P. Nous avons utilisé des
données publiées pour des Mollisols canadiens cultivés. Ce travail a été présenté lors d’un
congrès international sur l’analyse des données compositionnelles. Dans le chapitre 6, nous
avons utilisé la spectroscopie de la résonance magnétique nucléaire du 31P pour
charactériser les espèces ioniques et moléculaires du P, et étudié leur distribution dans le
profil du sol perturbé par des pratiques culturales en utilisant l’analyse compositionnelle.
99
CHAPITRE VI: LONG-TERM IMPACT OF TILLAGE
PRACTICES AND P FERTILIZATION ON SOIL P FORMS AS
DETERMINED BY 31P NUCLEAR MAGNETIC RESONANCE
SPECTROSCOPY
Dalel Abdi, Barbara J. Cade-Menun*, Noura Ziadi and Léon-Étienne Parent
D. Abdi and N. Ziadi, Agriculture and Agri-Food Canada, Soils and Crops Research and
Development Centre, 2560 Hochelaga Boulevard, Québec, QC, Canada G1V 2J3; B. J.
Cade-Menun, Agriculture and Agri-Food Canada, Semiarid Prairie Agricultural Research
Centre, P.O. Box 1030 Swift Current, SK, Canada, S9H 3X2; D. Abdi and L.É. Parent,
Department of Soils and Agri-Food Engineering, Université Laval, Québec, QC, Canada
G1K 7P4.
Journal of Environmental Quality, 2014. 43 (4): 1431-1441.
100
6.1 RÉSUMÉ
Les pratiques de conservation du sol sont de plus en plus utilisées pour réduire
l’érosion, améliorer la capacité de rétention de l’eau et la teneur en matière organique.
L’objectif de cette étude était d’évaluer l’effet du travail du sol (labour concentionel; MP,
et semis direct, NT) à long-terme et la fertilisation phosphatée (0 et 35 kg P ha-1) sur la
distribution des espèces de P dans le profil du sol. Les échantillons de sol ont été prelevés
d’un site expérimental établi au Québec, Canada, sur une rotation maïs-soya dans trois
profondeurs (0-5, 5-10, et 10-20 cm). Les résultats des analyses chimiques des échantillons
du sol ont montré que le PM3 et les orthophosphates s’accumulaient à la surface du sol
fertilisé sous semis direct, alors que les formes organiques du P (monoesters et nucléotides)
s’accumulaient en profondeur du sol non labouré. Nous avons conclu que le semis direct et
la fertilisation phosphatée changeaient la distribution des formes de P tout au long du profil
du sol et pourrait augmenter le risque de perte de P.
101
6.2 ABSTRACT
Conservation tillage practices have become increasingly common in recent years
to reduce soil erosion, improve water conservation, and increase soil organic matter.
However, research suggests that conservation tillage can stratify soil test phosphorus (P),
but little is known about the effects on soil organic P. This study was conducted to assess
the long term effects of tillage practices (no-till [NT] and mouldboard plowing, [MP]),
and P fertilization (0 and 35 kg P ha-1) on the distribution of P species in the soil profile.
Soil samples from a long term corn-soybean rotation experiment in Québec, Canada, were
collected from three depths (0-5, 5-10, and 10-20 cm). These were analysed for total P
(TP), total C (TC), total N (TN), pH, and Mehlich-3 P (PM3); P forms were characterized
with solution phosphorus-31 nuclear magnetic resonance spectroscopy (31P-NMR).
Results showed a stratification of TP, TC, TN, pH, PM3 and Mehlich-3 extractable
aluminum (Al) and magnesium (Mg) under NT management. The PM3 and
orthophosphate concentrations were greater at the soil surface (0 – 5 cm) of the NT-P35
treatment. Organic P forms (orthophosphate monoesters scyllo-IP6 and nucleotides) had
accumulated in the deep layer of NT treatment possibly due to preferential movement. We
found evidence that the NT system and P fertilization changed the distribution of P forms
along the soil profile, potentially increasing soluble inorganic P loss in surface runoff and
organic P in drainage, and decreasing bioavailability of both inorganic and organic P in
deeper soil layers.
Abbreviations: α-Glyc, α-glycerophosphate; β-Glyc, β-glycerophosphate; Chol-P, choline
phosphate; clr, centered log ratio; Gluc-6P, glucose-6-phosphate; IP6, inositol
hexakisphosphate; MP, mouldboard plow; Nucl, nucleotides; NT, no-till; Ortho,
orthophosphate; PM3, soil P content extracted using the Mehlich-3 method; P0, soil
treatment with 0 kg P ha−1; P35, soil treatment with 35 kg P ha−1; Phos, phosphonate; Poly,
polyphosphate; Pyro, pyrophosphate; Res P, residual P; TC, total carbon; TN, total
nitrogen; TP, total P.
102
6.3 INTRODUCTION
Phosphorus is one of the essential and commonly limiting macronutrients for plant
growth, but is a major cause of freshwater eutrophication (Elser et al., 2007). Conservation
tillage practices (minimum or no-till; NT) are characterized by minimal soil disturbance and
mixing, and have been used more frequently in recent years to reduce off-site losses of
nutrients associated with eroded particles, including P (Chichester and Richardson, 1992;
Shepard, 2005). Some benefits of NT over conventional tillage include reduced wind and
water erosion (Dick, 1992; Olson and Ebelhar, 2009) and greater soil biological activity
(Duiker and Beegle, 2006). Further advantages of NT for crop growth and yields (Lafond et
al., 2011) and economic performance (Holm et al., 2006) are well recognized.
However, by maintaining crop residues and fertilizers on the soil surface and
reducing their mixing into the plow layer, the relatively immobile nutrients that do not
readily move down the soil profile will remain at or near the soil surface (Sharpley, 2003;
Sharpley and Smith, 1994). Therefore, NT management systems often result in high
concentrations of nutrients at the soil surface (0 -5 cm) but sharply decreasing
concentrations below this depth (Selles et al., 1999). Studies have shown that NT
management has induced the stratification of soil organic carbon ( Poirier et al., 2009;
Zibilske et al., 2002), nitrate (Grant and Lafond, 1994; Lupwayi et al., 2006), potassium
(Fernández et al., 2008) and extractable P ( Sharpley, 2003; Sharpley and Smith, 1994;
Zibilske et al., 2002).
Stratification of P is of particular environmental and agronomic concern. Indeed,
high concentrations at the soil surface potentially increase the loss of dissolved P in runoff
(Cade-Menun et al., 2013; Kleinman et al., 2009), which is readily available to aquatic
organisms, while low concentrations at depth may limit plant P uptake by decreasing
available P in the rooting zone (Lupwayi et al., 2006). Dissolved P could be lost from soil to
water through drains (Haygarth et al., 1998), where dissolved organic P can represent a large
fraction (McDowell and Koopmans, 2006) and could be used by algae, especially when low
dissolved inorganic P concentrations limit growth (Whitton et al., 1991). Therefore, it is
important to characterize P stratification and to determine its possible causes. This can allow
103
producers to apply fertilizer appropriately for plant requirements while minimizing the
potential for P loss.
Phosphorus stratification may be due to the minimal mixing of surface-applied
fertilizers under NT management (Sharpley, 2003), and/or due to leaching of P from crop
residues that are retained on the soil surface (Hussain et al., 1999). To indicate the P source,
Cade-Menun et al. (2010) identified specific inorganic and organic P forms and
characterized their distribution patterns with soil depth in a long-term tillage study, using
solution 31P nuclear magnetic resonance (31P -NMR) spectroscopy. To date, few studies have
investigated the effects of tillage systems (Redel et al., 2011) and P fertilization (Ahlgren et
al., 2013) on the stratification of 31P-NMR P forms in soil profiles, and their interaction
effect has never been studied.
Solution 31P-NMR spectroscopy is by far the most widely used spectroscopic
technique for the speciation of soil organic P, because it provides the most detailed and
accurate information in most circumstances (Doolette and Smernik, 2011). However, few
31P-NMR studies to date have used replicate field samples, allowing for statistical analysis
of P forms. Caution must be used during statistical analysis of 31P-NMR data, because this
technique produces results that are not normally distributed and must be transformed prior to
statistical analysis. The 31P–NMR P forms are generally computed as proportions of total P
and hence are constrained between 0 and 100%. Therefore, following Aitchison (1986) and
Egozcue et al. (2003), P compounds can be defined as compositional data, i.e. strictly
positive data characterized by a constant sum and considered to be parts of a whole that only
provides relative information. Because compositional data are intrinsically related to each
other and have logistic-normal distribution, they should be expressed in terms of log ratios
(Aitchison, 1986). A simple log transformation, variable by variable, or any other
transformation of the single compositional variables may provide statistically erroneous
results, and thus may lead to contradictory interpretations (Filzmoser et al., 2009).
Compositional data analysis using log ratio transformation can avoid these difficulties.
The objectives of this study were: (i) to use 31P-NMR spectroscopy to identify the P
forms along the soil profile under a long-term tillage and P fertilization study in Québec,
Canada; (ii) to use compositional analysis to examine the effect of tillage systems and P
104
fertilization on their distribution patterns with depth; and (iii) to determine the possible causes
and risks for the stratification of soil P forms.
6.4 MATERIELS AND METHODS
6.4.1 Experimental site
The long-term crop rotation experiment was established in 1992 at the l’Acadie
Research Station (45°18’N; 73°21’W), Agriculture and Agri-Food Canada. The soil is a clay
loam (clayey, mixed, mesic Typic Humaquept) with 364 g kg-1 clay and 204 g kg-1 sand in the
Ap horizon. This deep soil originates from a fluvial deposit and evolved from a fine-textured,
greyish-to-brown parent material. The site is tile-drained with slope less than 1% and was
cropped with alfalfa (Medicago sativa) before 1992. Corn (Zea mays L.) was grown in 1992 to
1994, followed by soybean (Glycine max L.) in 1995; after that, corn and soybean were grown
in an annual rotation until 2010.
The chemical characteristics of the topsoil when the experiment was established were,
on average: soil organic matter (SOM) 38 g kg-1, PM3 135 kg ha-1, Mehlich-3 saturation ratio
(P/Al) 4.3%, and pH 6.3 (1:2 soil/water) (Légère et al., 2008; Tremblay et al., 2003). The mean
annual temperature in the area of the study is 6.3°C, and the mean total annual precipitation is
1100 mm (Poirier et al., 2009).
The experimental set-up was a split plot with two tillage practices (NT and MP)
assigned to main plots and nine fertilization combinations consisting of three N (0, 80 and 160
kg N ha-1) and three P (0, 17.5 and 35 kg P ha-1) applications to subplots. Experimental
treatments were replicated in four blocks, with individual plots measuring 25 m long and 4.5 m
wide. The MP treatment consisted of one moldboard plowing operation in the fall after harvest
to a depth of 20 cm, followed by disking and harrowing to 10 cm each spring before seeding.
For the NT treatment, plots had previously been ridge-tilled from 1992 to 1997 and were flat
direct-seeded from 1998 onward. For direct seeding, crop residues were left on the ground after
harvest. There were six rows per subplot unit, and corn and soybean were sown at rates of 74 x
103 and 45 x 104 plants ha-1, respectively. Mineral fertilizers (N and P) were band-applied (5
cm from the seeding row) only during the corn phase of the rotation (11 years) using a disk
opener (3 – 4 cm deep), according to local recommendations. The P treatments were applied in
a single application at planting as triple super-phosphate (0-46-0). Nitrogen treatments were
105
first band-applied at seeding at rates of 0, 48 and 48 as urea, and completed with 0, 32, and 112
kg N ha-1 side-dressed as ammonium nitrate at approximately the eight-leaf stage.
6.4.2 Soil sampling and chemical analysis
Although the experiment was conducted with four replicate blocks per treatment, only
three blocks with subplots receiving 160 kg N ha-1 and 0 or 35 kg P ha-1 were selected for this
study. Soil profiles were sampled to depths of 0-5 cm, 5-10 cm and 10-20 cm during fall 2010
for a total of 36 samples (2 tillage x 2 P-fertilization x 3 replicates x 3 depths). Samples were
air-dried and sieved to < 2 mm. Soil pH was measured in distilled water with 1:2 soil to
solution ratio (Hendershot et al., 2008). Soils were extracted by shaking 2.5 g of soil with 25 ml
of Mehlich-3 solution (pH 2.3) for 5 min (Mehlich, 1984) and the concentrations of P, Al, iron
(Fe), calcium (Ca), and magnesium (Mg) were determined with an Inductively Coupled
Plasma Optical Emission Spectrometer (ICP-OES; Model 4300DV, Perkin Elmer, Shelton,
CT). Total soil P was determined as described in Nelson (1987). Briefly, 0.1 g of finely ground
soil (0.2 mm) was mixed in a 50-mL boiling flask with 0.5 g K2S2O8 and 10 mL 0.9 M H2SO4,
and then digested at 121◦C in an autoclave for 90 min. The solution was analyzed by the
ammonium molybdate-ascorbic acid method (Murphy and Riley, 1962). Total C and N were
determined by dry combustion on 0.20 mm ground soils with a LECO CNS-1000 analyzer
(LECO Corp., St. Joseph, MI).
6.4.3 Solution 31P-NMR spectroscopy
Samples were analyzed by solution 31P-NMR spectroscopy using a modified version
of the Cade-Menun and Preston (1996) procedure. This involved shaking 2.5 g of soil with
25 ml of combined 0.25 mol L-1 NaOH and 0.05 mol L-1 Na2EDTA solutions for 6 h,
followed by centrifugation for 20 min at approximately 1500 x g. A 1-mL aliquot was
removed and diluted to 10 ml with deionized water for determination of TP, Al, Fe, Ca, and
manganese (Mn) by ICP-OES. The remaining supernatants were frozen and freeze-dried.
Freeze-dried samples were re-dissolved in 1.0 mL D2O, 0.6 mL of 10 mol L-1 NaOH, and
0.6 mL of the NaOH-EDTA extracting solution. The samples were centrifuged (1500 x g)
for 20 min to remove particles > 0.1 µm in diameter and transferred to 10 mm NMR tubes.
Solution 31P-NMR spectra were acquired on a 600 MHz spectrometer (202.5 MHz for P;
INOVA; Varian, Palo Alto, CA) equipped with a 10 mm broadband probe. The NMR
106
experimental parameters were: pulse width 18 µs (90o), acquisition 0.675 s, delay time 4.32
s; 2200-4300 scans (4-6 h); no proton decoupling. Although spin-lattice relaxation (T1)
times were not measured for these samples, this delay time was estimated to be sufficient
based on the ratio of P/(Fe+Mn) in the extracts (McDowell et al., 2006; Cade-Menun and
Liu, 2013).
Chemical shifts of signals were determined in parts per million (ppm) relative to an
external orthophosphoric acid standard (85%), and the orthophosphate peak was
standardized to 6 ppm for each sample. Signals were assigned to P compounds based on the
literature (Cade-Menun, 2005; Cade-Menun et al., 2010; Turner et al., 2012). Peak areas
were calculated by integration on spectra processed with 1 Hz and 7 Hz line broadening,
using NMR Utility Transform Software (NUTS, Acorn NMR, Livermore CA, 2000 edition).
6.4.4 Compositional data analysis
Soil chemical properties, excluding pH, were closed to total mass of soil and
represented as pertaining to a sample space called the simplex SD which is defined as
follows:
SD = {TP + Al + Fe + Ca + Mg + TC + TN + Res = 100%} [1]
where Res represent the residual mass in the soil. Total P comprised PM3 and its
complementary value to total P. To conduct statistical analyses in this study, we used the
centred log-ratio (clr) from Eq. [2] to transform, and back-transform, the soil chemical
properties (excluding pH), and the P forms computed as relative percentages of TP or as
raw concentrations on a dry-weight basis using the R ‘compositions’ package (van den
Boogaart et al., 2011). The clr transformation computes the geometric mean (g(x)) across
components ( ix ) as follows:
[2]
The clr transformation simplifies the interpretation of the transformed variables
because one could think in terms of the original variables of the simplex (Filzmoser and
Hron, 2009).
)(ln)(
xg
xxclr i
i
107
The R ‘robCompositions’ package was used to replace the rounded zeros for
concentrations of P failing to be detected (values below detection limit of the 31P-NMR), and
the whole composition was adjusted accordingly (Martín-Fernández et al., 2003).
Confidence intervals were computed about clrs means then back-transformed to familiar
measurement units using the R ‘compositions’ package (van den Boogaart et al., 2011).
6.4.5 Statistical analysis
Data were tested for normality using the SAS univariate procedure (SAS Institute,
2001). Analyses of variance (ANOVA) for centred log-ratio transformed 31P-NMR P
forms, PM3, TP, Al, Fe, Ca, Mg, TC and TN, and pH were conducted using Proc Mixed of
SAS (SAS Institute, 2001) to test the effects of tillage, P fertilization, depth, and their
interactions. Least Significant Difference (LSD) was computed to separate treatments
means when ANOVA test was significant (p < 0.05) or considered as a trend (0.05 < p <
0.1).
6.5 RESULTS
6.5.1 Chemical soil properties
The statistical ANOVA results for chemical soil properties are shown in Table 1.
There was a significant interaction of tillage, fertilization, and depth for TP (Table 1, Fig. 1).
The LSD results showed significant differences for the amount of TP between the top soil
layer (0-5 cm) and the subjacent layers (5-10 cm and 10-20 cm) under NT-P0, and between
the 0 to10 cm and 10 to 20 cm soil layers in the NT-P35 treatment. Similarly, a significant
tillage x P fertilization x depth interaction (p < 0.05) was obtained for PM3 (Table 1). The
highest concentrations for PM3 were recorded under NT in P fertilized treatment at 0 to 5 cm
(Fig. 1). Concentrations of Al and Mg were affected by tillage x soil depth interaction (Fig. 2).
However, Fe was affected only by the depth variation, and was significantly lower at 0-5 cm
(212 ± 14mg kg-1) than at the deeper layers (5-10 cm, 226± 14 mg kg-1; 10-20 cm, 229± 17
mg kg-1). There were no significant treatment effects for Ca (Table 1), which was 2432 ± 131
mg kg-1 under MP and 2768 ± 244 mg kg-1 under NT. Significant differences in TC and TN
were detected in NT treatment among the soil depths, where they were higher in the top 5 cm
soil layer (Fig. 2). Soil pH was significantly higher in the NT treatment at 0 to 5 cm (6.33 ±
108
0.14; p < 0.01) compared to 5 to 10 cm (5.95 ± 0.25) and 10 to 20 cm (5.96 ± 0.32). None of
the soils parameters tested, except PM3, showed any effect of P fertilization.
6.5.2 Identification of phosphorus forms by 31P nuclear magnetic resonance
spectroscopy
The 31P-NMR peaks for P compounds detected in this study fall between 25 and -25
ppm (Table 2). An example spectrum from MP-P35 treatment at 0 to 5 cm depth is shown in
Fig. 3. Three groups of inorganic P forms were detected: orthophosphate at 6.00 ± 0.01 ppm
chemical shift, pyrophosphate at -4.02 ± 0.02 ppm, and polyphosphates between -4.23 and -
24.76 ppm, with the polyphosphate end group detected at -3.98 ± 0.05 ppm.
Organic P compound classes detected by solution 31P-NMR included phosphonates
from 25 to 7.80 ppm, orthophosphates monoesters at 7.70 to 6.20 ppm and at 5.78 to 3.5
ppm, and orthophosphates diesters between 2.20 and -3.40 ppm (Cade-Menun, 2005; Cade-
Menun et al., 2010). An example spectrum of the monoester region from MP-P35 treatment
at 0 to 5 cm depth is shown in Fig. 4. For all treatments and depths, the 31P-NMR spectra
indicated that orthophosphate monoesters were dominated by stereoisomers of inositol
hexakisphosphate (IP6). Three of these (myo, neo, and scyllo) were detected in all
treatments, whereas D-chiro-IP6 was detected in NT-P0 in one block at 0 to 5 cm depth, and
in NT-P35 at the three depths. The most abundant stereoisomer of IP6, myo-IP6, gave four
characteristics signals in the ratio 1:2:2:1 at 5.48 (± 0.02), 4.51 (± 0.02), 4.11 (± 0.02), and
4.02 (± 0.02) ppm (Turner et al., 2003; Cade-Menun, 2005). The identification of these
peaks as myo-IP6 was confirmed by rerunning two samples after spiking with phytic acid
(McDowell et al., 2007). Two signals at 6.41 (± 0.01) ppm and 4.27 (± 0.01) ppm in 1:4
ratio were assigned to the 4 equatorial/2-axial conformation of neo-IP6 based on Turner et
al. (2012). The peaks for D-chiro IP6 were detected in 2:2:2 ratio in the 2-equatorial/4-axial
conformation at 6.23 ppm, 4.75 (± 0.01) ppm, and 4.34 (± 0.03), ppm (Turner et al., 2012).
The signal from scyllo-IP6 occurred at 3.71 ± 0.02 ppm (Cade-Menun et al., 2010).
Glucose 6-phosphate was detected at 5.12 ± 0.03 ppm, while the 4.88 ± 0.01 ppm and 4.55
± 0.02 ppm peaks were assigned to α-glycerophosphate (α-glyc) and β-glycerophosphate
(β-glyc), respectively in a 1:2 ratio. Choline phosphate (chol-P, 3.85 ± 0.03 ppm) and two
signals of nucleotides (4.33 ± 0.01, 4.16 ± 0.02 ppm) were also detected in every soil
109
sample. Peak locations for β-glyc and chol-P were confirmed by spiking samples with these
two P forms.
Unidentified groups of the orthophosphate monoesters were divided into three
general groups based on their locations (Cade-Menun, 2005; Hill and Cade-Menun, 2009).
The Monoester 1 group contained unidentified peaks between 6.48 and 7 ppm, and included
a peak at 6.51± 0.04 ppm. Peaks in the Monoester 2 region were detected at 5.68 ± 0.02
ppm, 5.26 ± 0.01 ppm, and 4.75 ± 0.03 ppm. The Monoester 3 group included a peak at 3.79
± 0.15 ppm observed in the soil samples containing D-chiro IP6. An unknown peak was
detected at 4.93 ± 0.02 ppm. It currently has not been specifically identified; unpublished
work (Cade-Menun, 2014) suggests that it may be myo-1-IP, but this requires confirmation
with spiking experiments, which were not done for the current manuscript. The DNA peak
was found in most samples at -0.75 ± 0.03 ppm. The remaining orthophosphate diesters
were divided into two groups. Peaks for “Other diester 1”, were detected between 3.34 and
0.41 ppm, and may include diagnostic peaks for phospholipids (Cade-Menun et al., 2010).
The “Other diester 2” group was observed between 1.76 and -3.72 ppm. The unidentified
groups of the orthophosphate monoesters (Monoester 1, 2, and 3) and diester (Other diester
1 and 2), the unknown compound, and D-chiro IP6 that was not detected in all soil samples
were amalgamated and identified as a residual phosphate fraction “Res” of TP.
6.5.3 Distribution of 31P nuclear magnetic resonance phosphorus forms
The statistical results for the relative percentages of soil P forms determined by 31P-
NMR are presented in Table 3. The only P form to show an interaction of tillage, depth and
fertilization was orthophosphate, which was significant at p < 0.1. Orthophosphate was the
prominent fraction of extracted P in all analyzed soil samples (Table 4). The highest
concentration (365 ± 21 mg kg-1), corresponding to the relative percentage of extracted P of
49.7%, was found in the NT-P35 treatment at 0 to 5 cm depth and was significantly different
from the values at 5 to 20 cm (p < 0.01, Fig. 1). In contrast, the lowest concentration (206 ±
26 mg kg-1), which is equivalent to 37.7 % of extracted P, was observed in the deep layer (10-
20 cm) under MP-P0.
Significant effects of the interaction between tillage and depth were observed for
pyrophosphate, scyllo-IP6, nucleotides and DNA, while significant effects of the interaction
110
between tillage and fertilization occurred only for pyrophosphate (Table 3). Significant
differences were observed between the top soil (0-5 cm) and the deep (5-20 cm) layers
under no-till for pyrophosphate, scyllo-IP6 and nucleotides and between 0 to 10 cm and 10
to 20 cm for DNA (Fig. 5). In MP treatment, the DNA content increased significantly from
the top 5 cm layer (9.3 ± 0.4 mg kg-1) to 5 to 10 cm (11.6 ± 0.1 mg kg-1), and 10 to 20 cm
(11.9 ± 0.2 mg kg-1, Fig. 5).
Where there were no interactions with other treatments, the percentage of
phosphonate was significantly higher under the P35 treatment and the percentages of α and
β-glycerophosphate and choline phosphate were significantly higher under the P0 treatment
(Table 4). The percentages of α and β-glycerophosphate were the only P forms affected by
depth without an interaction to another treatment (Table 3), and were significantly higher at
10-20 cm than higher in the soil profile (Table 4). There were no significant differences
between treatments for polyphosphate, myo-IP6, neo-IP6, glucose-6P and the residue
fraction of P (Table 3).
6.6 DISCUSSION
Our results showed that TP significantly varied with soil depth under NT treatment,
where it accumulated in the top unfertilized soil layer (0–5 cm) and in the 0 to 10 cm layer
where P fertilizer was applied. This agrees with Redel et al. (2011) who found a significant
difference between NT and conventional tillage in TP amount in the 0 to 20 soil layer.
Additionally, pH, PM3, Mg, and TC were significantly higher in the topsoil (0–5 cm) of NT
treatment than the deeper layers (5–20 cm), and TN accumulated in the surface 10 cm. The
opposite trend was observed for Al. Conversely, the distributions of TC and TN were
homogenous along the soil profile under MP management where they slowly increased (Fig.
2). This suggests that stratification in NT results from the retention of crop residues at the
soil surface, where decomposition led to the release of soluble P and the accumulation of TC
and TN (Poirier et al., 2009), and to a pH increase (Hussain et al., 1999; Paul et al., 2001),
which produced a decrease in extractable Al (Shang et al., 1992) and an increase in
exchangeable Mg (Hussain et al. 1999). The stratification of PM3 in the 0 to 5 cm layer of
the NT-P0 treatment, in contrast with MP-P0 (Fig. 2), suggests that the build-up of organic
matter and the subsequent leaching of P during 18 years of soil cultivation played an
111
important role in P stratification. This accumulation of PM3 seemed to be accentuated in the
P-fertilized treatment (Fig. 1) due to a lack of mixing of soil with applied fertilizer. The
depletion of available P in lower layers of NT (5-20 cm), due in part to the higher bulk
density, may result in reduced P uptake by plants (Lupwayi et al., 2006). Thus, crop yield
was greater under MP than under zero tillage (Légère et al., 2008; Messiga et al., 2012). The
same trend for PM3 has been reported elsewhere for available P extracted with different
methods, when compared NT and MP systems applied on P fertilized soil ( Cade-Menun et
al., 2010; Duiker and Beegle, 2006; Hussain et al., 1999; Vu et al., 2009). The increase of
PM3 with P fertilization, regardless of soil tillage treatment and depths, was also observed in
samples collected in 2009 from the same research plot (Sheng et al., 2013). Likewise, the
31P-NMR spectra showed concomitant stratification of orthophosphate in NT-P35 (Fig. 1).
Shi et al. (2012) found greater alkaline phosphatase activity, which mineralized more
orthophosphate monoesters at the surface soil under NT management in greater amounts
compared to MP of the same research site in 2009. This could easily explain the
significantly higher amount of orthophosphate on the NT soil surface compared to
conventional tillage, as could stratification of the orthophosphate applied in fertilizer.
Stratification also occurred under NT for scyllo-IP6 and the microbe-associated
compounds: pyrophosphate, nucleotides and DNA (Fig. 5). The origin of scyllo-IP6 remains
unknown, although the fact that it is a stereoisomer of myo-IP6 (Turner et al., 2002) points
to epimerization reactions (L'Annunziata, 1975) or microbial production (Caldwell and
Black, 1958). However, no significant differences between the treatments were found for
myo-IP6 and neo-IP6, which were uniformly distributed across treatments (Table 2). Under
NT treatment, the pyrophosphate and DNA concentrations were greater at the surface (0–5
cm) soil layer than the deeper layers (Fig. 5), whereas DNA significantly accumulated
below the 5 cm layer under MP. Total C and N were similarly affected (Fig. 2), suggesting
that DNA was synthesized in greater amounts under NT owing to the higher organic matter
accumulated at the soil surface (Condron et al., 2005) in comparison to conventional
tillage. It was unlikely to have accumulated due to sorption to the mineral surface, because
the soil pH was higher than the isoelectric point (pH 5.0) of DNA (Condron et al., 2005).
The opposite trend was observed for the nucleotides, which may be derived from the
hydrolysis of RNA during NMR analysis (He et al., 2011), or may be present naturally in
112
soil (Vestergren et al., 2012). Under direct drill, scyllo-IP6 and nucleotides accumulated
significantly in the deeper layers (Fig. 5), possibly due to their preferential movement
through the soil column (Condron et al., 2005). The accumulation of scyllo-IP6 could be
attributed also to the higher amount of Mehlich-3 extractable Al in the 5 to 20 cm soil layer
(Fig. 2). Indeed, Shang et al. (1992) found that inositol hexakisphosphate sorption was
dependent on the contents of Al oxides and the adsorption rate increased when the pH
decreased. Mononucleotides, and the RNA from which they likely originated in the soil
sample, do not sorb strongly to the soil; their increase may reflect higher microbial
biomass, but this was not analyzed for this study.
The α- and β-glycerophosphates and choline-P have been identified as degradation
products of phospholipids of cellular membranes during NMR analysis (Doolette et al.,
2009; He et al., 2011; Young et al., 2013). Both α and β-glycerophosphate increased
significantly in the soil profile regardless of treatment, suggesting that without degradation
phospholipids would also have increased. Little is known about these compounds in soils
(Cade-Menun et al., 2010), so it is difficult to explain the factors and processes controlling
their presence. The percentages of α and β-glycerophosphate and choline-P were reduced
under P fertilization in general. However, phosphonates, which are more stable in soils and
thus less bioavailable (Condron et al., 2005), were detected at a higher percentage in soils
receiving 35 kg P ha-1 than in unfertilized soils.
Overall, it appears that labile inorganic P accumulated at the surface of no-till soil
and decreased with depth, while organic P, such as scyllo-IP6 and nucleotides, increased
deeper in the soil profile. Consequently, NT may increase the potential for soluble
inorganic P loss in surface runoff (Kleinman et al., 2009), and lead to the loss of organic
monoesters draining through different hydrological pathways (Condron et al., 2005). These
hypotheses were further confirmed by Dodd et al. (2014) who showed that NT system
largely increased soluble P concentrations in the surface run-off and dissolved reactive P in
leachate compared with MP.
As knowledge on molecular species increases for the soil P cycle, more elaborate
theories could be built relating inorganic and organic P species into a framework of
balances using isometric log ratios (Filzmoser et al., 2009), similarly to balance designs in
113
biochemistry (Aslam et al., 2013) and ionomics (Parent et al., 2013). An optimum balance
between molecular species and beneficial cultural practices could be defined to maximize
crop yield while minimizing P losses.
6.7 CONCLUSIONS
In this study, we showed that long-term zero tillage resulted in the stratification of
TP, TC, TN, pH, Al, Mg, and PM3 in the soil profile, owing to the build-up of organic
matter at soil surface. The accumulation of soluble inorganic P in the top 5 cm was higher
with P fertilizer application due in part to the lack of mixing of fertilizer. Additionally, no-
till treatment led to the accumulation of 31P-NMR monoesters such as the scyllo-IP6 and the
nucleotides in deeper layers. This stratification may increase the potential of soluble
inorganic P loss by surface runoff, and organic P by leaching and drainage. The relative
percentage of simple orthophosphate monoesters such as the β-glycerophosphate and
choline-P were lower under P fertilization, whereas the percentage of phosphonates was
higher. Further investigations are needed to unravel the interactions between P species and
to understand P changes as perturbed by agricultural management practices.
6.8 ACKNOWLEDGMENTS
This study was funded by The Sustainable Agriculture Environmental Systems
(SAGES) initiative of Agriculture and Agri-Food Canada and by the Natural Sciences and
Engineering Research Council of Canada (NSERC-DG 2254). The authors are very
grateful to S. Côté and S. Michaud for their technical support during sampling operations
and laboratory analysis. The NMR work was conducted at the Stanford Magnetic
Resonance Laboratory at the Stanford University School of Medicine. We gratefully thank
Dr. Corey Liu for his assistance with the P-NMR analysis.
114
6.9 REFERENCES
Ahlgren, J., F. Djodjic, G. Börjesson, and L. Mattsson. 2013. Identification and
quantification of organic phosphorus forms in soils from fertility experiments. Soil
Use Manage. 29:24-35.
Aitchison, J. 1986. The statistical analysis of compositional data. Monographs on Statistics
and Applied Probability. 1sted. London, UK.
Aslam, S., P. Garnier, C. Rumpel, S. E. Parent, and P. Benoit. 2013. Adsorption and
desorption behavior of selected pesticides as influenced by decomposition of maize
mulch. Chemosphere 91: 1447–55.
Cade-Menun, B.J. 2005. Characterizing phosphorus in environmental and agricultural
samples by 31P nuclear magnetic resonance spectroscopy. Talanta 66:359-371.
Cade-Menun, B.J., and C.W. Liu. 2014. Solution 31P-NMR spectroscopy of soils from
2005-2013: A review of sample preparation and experimental parameters. Soil Sci.
Soc. Am. J. 78 (1):19-37.
Cade-Menun, B.J., and C.M. Preston. 1996. A comparison of soil extraction procedures for
31P NMR spectroscopy. Soil Sci. 161:770-785.
Cade-Menun, B.J., M.R. Carter, D.C. James, and C.W. Liu. 2010. Phosphorus forms and
chemistry in the soil profile under long-term conservation tillage: A phosphorus-31
nuclear magnetic resonance study. J. Environ. Qual. 39:1647-1656.
Cade-Menun, B.J., G. Bell, S. Baker-Ismail, Y. Fouli, K. Hodder, D.W. McMartin, C.
Perez-Valdivia, and K. Wu. 2013. Nutrient loss from Saskatchewan cropland and
pasture in spring snowmelt runoff. Can. J. Soil Sci. 93:445-458.
Caldwell, A.G., and C.A. Black. 1958. Inositol hexaphosphate: II. Synthesis by soil
microorganisms. Soil Sci. Soc. Am. J. 22:293-296.
Chichester, F.W., and C.W. Richardson. 1992. Sediment and nutrient loss from clay soils as
affected by tillage. J. Environ. Qual. 21:587-590.
Condron, L.M., B.L. Turner, and B.J. Cade-Menun. 2005. The chemistry and dynamics of
soil organic phosphorus. In: J.T. Sims and A.N. Sharpley, editors, Phosphorus:
agriculture and the environment. Wisconsin, USA. p. 87-121.
Dick, R.P. 1992. A review: long-term effects of agricultural systems on soil biochemical
and microbial parameters. Agric. Ecosyst. Environ. 40:25-36.
Dodd, R.J., R.W. McDowell, and L.M. Condron. 2014. Is tillage an effective method to
decrease phosphorus loss from phosphorus enriched pastoral soils? Soil Tillage Res.
135:1-8.
Doolette, A.L., and R.J. Smernik. 2011. Soil organic phosphorus speciation using
spectroscopic techniques. In: E.K. Bünemann, A. Oberson and E. Frossard editors,
Phosphorus in Action. Springer Heidelberg Dordrecht London, New York. p. 3-36.
Doolette, A.L., R.J. Smernik, and W.J. Dougherty. 2009. Spiking improved solution
phosphorus-31 nuclear magnetic resonance identification of soil phosphorus. Soil
Sci. Soc. Am. J. 73:919-927.
Duiker, S.W., and D.B. Beegle. 2006. Soil fertility distributions in long-term no-till,
chisel/disk and moldboard plow/disk systems. Soil Tillage Res. 88:30-41.
Egozcue, J.J., V. Pawlowsky-Glahn, G. Mateu-Figueras, and C. Barceló-Vidal. 2003.
Isometric logratio transformations for compositional data analysis. Math. Geol.
35:279-300.
115
Elser, J.J., M.E.S. Bracken, E.E. Cleland, D.S. Gruner, W.S. Harpole, H. Hillebrand, J.T.
Ngai, E.W. Seabloom, J.B. Shurin, and J.E. Smith. 2007. Global analysis of
nitrogen and phosphorus limitation of primary producers in freshwater, marine and
terrestrial ecosystems. Ecol. Lett. 10:1135-1142.
Fernández, F.G., S.M. Brouder, C.A. Beyrouty, J.J. Volenec, and R. Hoyum. 2008.
Assessment of plant-available potassium for no-till, Rainfed Soybean. Soil Sci.
Soc. Am. J. 72:1085-1095.
Filzmoser, P., and K. Hron. 2009. Correlation analysis for compositional data. Math.
Geosci. 41:905-919.
Filzmoser, P., K.Hron, and C. Reimann. 2009. Univariate statistical analysis of
environmental (compositional) data: Problems and possibilities. Sci. Total Environ.
407:6100-6108.
Grant, C.A., and G.P. Lafond. 1994. The effects of tillage systems and crop rotations on
soil chemical properties of a Black Chernozemic soil. Can. J. Soil Sci. 74:301-306.
Haygarth, P.M., L. Hepworth, and S.C. Jarvis. 1998. Forms of phosphorus transfer in
hydrological pathways from soil under grazed grassland. Eur. J. Soil Sci. 49:65-72.
He, Z., D.C. Olk, and B.J. Cade-Menun. 2011. Forms and lability of phosphorus in humic
acid fractions of hord silt loam soil. Soil Sci. Soc. Am. J. 75:1712-1722.
Hendershot, W.H., H. Lalande, and M. Duquette. 2008. Soil reaction and exchangeable
acidity. In: Carter MR and EG Gregorich editors, Soil Sampling and Methods of
Analysis. Can Soc Soil Sci, CRC Press Inc., Boca Raton, FL. p 173–178.
Hill, J.E., and B.J. Cade-Menun. 2009. Phosphorus-31 nuclear magnetic resonance
spectroscopy transect study of poultry operations on the Delmarva Peninsula. J.
Environ. Qual. 38:130-138.
Holm, F.A., R.P. Zentner, A.G. Thomas, K. Sapsford, A. Légère, B.D. Gossen, O. Olfert,
and J.Y. Leeson. 2006. Agronomic and economic responses to integrated weed
management systems and fungicide in a wheat-canola-barley-pea rotation. Can. J.
Plant Sci. 86:1281-1295.
Hussain, I., K.R. Olson, and S.A. Ebelhar. 1999. Long-term tillage effects on soil chemical
properties and organic matter fractions. Soil Sci. Soc. Am. J. 63:1335-1341.
Kleinman, P.A., A. Sharpley, L. Saporito, A. Buda, and R. Bryant. 2009. Application of
manure to no-till soils: phosphorus losses by sub-surface and surface pathways.
Nutr. Cycl. Agroecosyst. 84:215-227.
L'Annunziata, M.F. 1975. The origin and transformations of the soil inositol phosphate
isomers. Soil Sci. Soc. Am. J. 39:377-379.
Lafond, G.P., F. Walley, W.E. May, and C.B. Holzapfel. 2011. Long term impact of no-till
on soil properties and crop productivity on the Canadian prairies. Soil Tillage Res.
117:110-123.
Légère, A., F.C. Stevenson., and N. Ziadi. 2008. Contrasting responses of weed
communities and crops to 12 years of tillage and fertilization treatments. Weed
Tech. 22:309-317.
Lupwayi, N.Z., G.W. Clayton, J.T. O’Donovan, K.N. Harker, T.K. Turkington, and Y.K.
Soon. 2006. Soil nutrient stratification and uptake by wheat after seven years of
conventional and zero tillage in the Northern Grain belt of Canada. Can. J. Soil Sci.
86:767-778.
116
Martín-Fernández, J.A., C. Barceló-Vidal, and V. Pawlowsky-Glahn. 2003. Dealing with
zeros and missing values in compositional data sets using nonparametric
imputation. Math. Geol. 35:253-278.
McDowell, R.W., and G.F. Koopmans. 2006. Assessing the bioavailability of dissolved
organic phosphorus in pasture and cultivated soils treated with different rates of
nitrogen fertiliser. Soil Biol. Biochem. 38:61-70.
McDowell, R.W., I. Stewart, and B.J. Cade-Menun. 2006. An examination of spin–lattice
relaxation times for analysis of soil and manure extracts by liquid state phosphorus-
31 nuclear magnetic resonance spectroscopy. J. Environ. Qual. 35:293-302.
McDowell, R.W., B.J. Cade-Menun, and I. Stewart. 2007. Organic phosphorus speciation
and pedogenesis: analysis by solution 31P nuclear magnetic resonance
spectroscopy. Europ. J. Soil Sci. 58:1348-1357.
Mehlich, A. 1984. Mehlich 3 soil test extractant: A modification of Mehlich 2 extractant.
Commun. Soil Sci. Plant Anal. 15:1409-1416.
Messiga, A.J., N. Ziadi, C. Morel, C. Grant, G. Tremblay, G. Lamarre, and L.-E. Parent.
2012. Long term impact of tillage practices and biennial P and N fertilization on
maize and soybean yields and soil P status. Field Crops Res. 133:10-22.
Murphy, J., and J.P. Riley. 1962. A modified single solution method for the determination
of phosphate in natural waters. Anal. Chim. Acta 27:31-36.
Nelson, N.S. 1987. An acid‐persulfate digestion procedure for determination of phosphorus
in sediments. Commun. Soil Sci. Plant Anal. 18:359-369.
Olson, K.R., and S.A. Ebelhar. 2009. Impacts of conservation tillage systems on long-term
crop yields. J. Agron. 8:14-20.
Parent, L. E., S.-É. Parent, V. Hébert-Gentile, K. Naess, and L. Lapointe. 2013. Mineral
balance plasticity of cloudberry (Rubus chamaemorus) in Quebec-Labrador. Am. J.
Plant Sci. 4:1508-1520.
Paul, K.I., A.S. Black, and M.K. Conyers. 2001. Effect of plant residue return on the
development of surface soil pH gradients. Biol. Fertil. Soils 33:75-82.
Poirier, V., D.A. Angers, P. Rochette, M.H. Chantigny, N. Ziadi, G. Tremblay, and J.
Fortin. 2009. Interactive effects of tillage and mineral fertilization on soil carbon
profiles. Soil Sci. Soc. Am. J. 73:255-261.
Redel, Y.D., M. Escudey, M. Alvear, J. Conrad, and F. Borie. 2011. Effects of tillage and
crop rotation on chemical phosphorus forms and some related biological activities
in a Chilean Ultisol. Soil Use Manage. 27:221-228.
SAS Institute. 2001. The SAS system for Windows. Release 8.2. SAS Inst., Cary, NC.
Selles, F., B.G. McConkey, and C.A. Campbell. 1999. Distribution and forms of P under
cultivator- and zero-tillage for continuous- and fallow-wheat cropping systems in
the semi-arid Canadian prairies. Soil Tillage Res. 51:47-59.
Shang, C., J.W.B. Stewart, and P.M. Huang. 1992. pH effect on kinetics of adsorption of
organic and inorganic phosphates by short-range ordered aluminum and iron
precipitates. Geoderma 53:1-14.
Sharpley, A.N. 2003. Soil mixing to decrease surface stratification of phosphorus in
manured soils. J. Environ. Qual. 32:1375-1384.
Sharpley, A.N., and S.J. Smith. 1994. Wheat tillage and water quality in the Southern
plains. Soil Tillage Res. 30:33-48.
117
Sheng, M., R. Lalande, C. Hamel, and N. Ziadi. 2013. Effect of long-term tillage and
mineral phosphorus fertilization on arbuscular mycorrhizal fungi in a humid
continental zone of Eastern Canada. Plant Soil:1-15.
Shepard, R. 2005. Nutrient management planning: Is it the answer to better management? J.
Soil Water Conserv. 60:171-176.
Shi, Y., R. Lalande, N. Ziadi, M. Sheng, and Z. Hu. 2012. An assessment of the soil
microbial status after 17 years of tillage and mineral P fertilization management.
Appli. Soil Ecol. 62:14-23.
Tremblay, G., L. Robert, P. Filion, G. Govaerts, R. Mongeau, J. Filiatrault, J.M. Beausoleil,
G. Moreau, T.S. Tran, 2003. Régies culturales et fertilisations azotée et phosphatée
dans une rotation maïs-soya. Bulletin Technique No. 3.05. CÉROM, Saint-Bruno-
de-Montarville, Quebec City, QC, 8 pp.
Turner, B.L., N. Mahieu, and L.M. Condron. 2003. Quantification of myo-inositol
hexakisphosphate in alkaline soil extracts by solution 31P nmr spectroscopy and
spectral deconvolution. Soil Sci. 168:469-478.
Turner, B.L., M.J. Papházy, P.M. Haygarth, and I.D. McKelvie. 2002. Inositol phosphates
in the environment. Phil Trans Roy Soc London B: 357:449-469.
Turner, B.L., A.W. Cheesman, H.Y. Godage, A.M. Riley, and B.V.L. Potter. 2012.
Determination of neo- and d-chiro-inositol hexakisphosphate in soils by solution
31P NMR spectroscopy. Environ. Sci. Technol. 46:4994-5002.
van den Boogaart, K.G., R. Tolosana-Delgado, and M. Bren. 2011. Compositions:
compositional data analysis, R Package Version 1.10-2.
http://CRAN.Rproject.org/package=compositions.
Vestergren, J., A.G. Vincent, M. Jansson, P. Persson, U. Ilstedt, G. Gröbner, R. Giesler,
and J. Schleucher. 2012. High-resolution characterization of organic phosphorus in
soil extracts using 2D 1H–31P nmr correlation spectroscopy. Environ. Sci. Technol.
46:3950-3956.
Vu, D.T., C. Tang, and R.D. Armstrong. 2009. Tillage system affects phosphorus form and
depth distribution in three contrasting Victorian soils. Soil Res. 47:33-45.
Whitton, B.A., S.L.J. Grainger, G.R.W. Hawley, and J.W. Simon. 1991. Cell-bound and
extracellular phosphatase activities of cyanobacterial isolates. Microbial Ecol.
21:85-98.
Young, E.O., D.S. Ross, B.J. Cade-Menun, and C.W. Liu. 2013. Phosphorus speciation in
riparian soils: A phosphorus-31 nuclear magnetic resonance spectroscopy and
enzyme hydrolysis study. Soil Sci. Soc. Am. J. 77:1636-1647.
Zibilske, L.M., J.M. Bradford, and J.R. Smart. 2002. Conservation tillage induced changes
in organic carbon, total nitrogen and available phosphorus in a semi-arid alkaline
subtropical soil. Soil Tillage Res. 66:153-163.
118
Table 6-1 Analysis of variance for the effects of tillage, P fertilization and depth on clr transformed concentrations of soil total P
(TP), Mehlich-3 extractable P (PM3), aluminium (Al), iron (Fe), calcium (Ca), magnesium (Mg), total carbon (TC) and total
nitrogen (TN), and pH.
Sources of
variation
TP† PM3 Al Fe Ca Mg TC TN pH
Tillage (T) NS‡ NS NS NS NS NS NS NS NS
Phosphorus (P) NS ** NS NS NS NS NS NS NS
Depth (D) NS *** *** ** NS * ** **** ****
T × P NS NS NS NS NS NS NS NS NS
T × D **** ** *** NS NS **** ** * *
P × D NS * NS NS NS NS NS NS NS
T × P × D * * NS NS NS NS NS NS NS
* Significant at the 0.05 probability level.
** Significant at the 0.01 probability level.
*** Significant at the 0.001 probability level.
**** Significant at the 0.10 probability level.
† PM3, Mehlich-3–extractable P; TC, total C; TN, total N; TP, total P.
‡ Nonsignificant.
119
Table 6-2 Chemical shift of P forms detected in the 31P-NMR spectrum of the soil as
affected by tillage and P fertilization management and depth.
Category P form or compound class Chemical shift
ppm
Inorganic P orthophosphate 6.00
pyrophosphate 4.02
polyphosphates 4.23 to 24.76
polyphosphate end group 3.98
Organic P phosphonates 25.0–7.8 (signals at 20.36,
18.76)
Orthophosphate monoesters Myo-IP6 5.48, 4.51, 4.11, 4.02
Neo- IP6 6.41, 4.27
D-chiro IP6 6.23, 4.75, 4.34
Scyllo- IP6 3.71
glucose 6-phosphate 5.12
-glycerophosphate 4.88
-glycerophosphate 4.55
choline phosphate 3.85
nucleotides 4.33, 4.16
monoester 1 7.0–6.48 (signal at 6.51)
monoester 2 5.68, 5.26, 4.75
monoester 3 3.79
unknown 4.93
Orthophosphate diesters DNA 0.75
other diester 1 3.34–0.41
other diester 2 1.76 to 3.72
120
Table 6-3 Analysis of variance for the effects of tillage, P fertilization, and depth on centered log ratio–transformed soil P forms
determined by 31P nuclear magnetic resonance spectroscopy.
Sources of
variation
Inorganic P† Organic P‡
Ortho Pyro Poly Phos myo-IP6 neo-
IP6
scyllo-IP6 Gluc-6P -Glyc -Glyc Chol-
P
Nucl DNA Res
P
Tillage (T) NS§ NS NS NS NS NS NS NS NS NS NS NS NS NS
Phosphorus (P) * NS NS * NS NS NS NS **** **** **** NS NS NS
Depth (D) * **** NS NS NS NS * NS **** **** NS * NS NS
T × P NS **** NS NS NS NS NS NS NS NS NS NS NS NS
T × D NS **** NS NS NS NS **** NS NS NS NS ** ** NS
P × D NS NS NS NS NS NS NS NS NS NS NS NS NS NS
T × P × D **** NS NS NS NS NS NS NS NS NS NS NS NS NS
* Significant at the 0.05 probability level.
** Significant at the 0.01 probability level.
**** Significant at the 0.10 probability level.
† Ortho, orthophosphate; Poly, polyphosphate; Pyro, pyrophosphate.
‡ -Glyc, -glycerophosphate; -Glyc, -glycerophosphate; Chol-P, choline-phosphate; Gluc-6P, glucose-6 phosphate; myo-IP6, myo-inositol
hexakisphosphate; neo-IP6, neo-inositol hexakisphosphate; Nucl, nucleotides; Phos, phosphonate; scyllo-IP6, scyllo-inositol hexakisphosphate; Res P,
residual organic P (unidentified organic P forms).
§ Nonsignificant.
121
Table 6-4 Back-centered log ratio–transformed soil P forms determined by 31P nuclear magnetic resonance spectroscopy as
affected by tillage, P fertilization, and depth.
Treatment
†
Inorganic P‡ Organic P§
Ortho Pyro Poly Phos myo-
IP6
neo-IP6 scyllo-
IP6
Gluc-
6P -
Glyc
-Glyc Nucl Chol-
P
DNA Res P
% total P
Tillage
MP 42.3 ± 1.4a¶ 1.0 ± 0.1a 0.8 ±
0.2a
2.5 ±
0.2a
10.0 ±
0.3a
4.0 ± 0.1a 4.0 ± 0.2a 2.5 ±
0.1a
1.5 ±
0.1a
2.9 ±
0.3a
5.8 ±
0.4a
2.2 ±
0.2a
1.8 ±
0.1a
18.8
± 0.6a
NT 41.2 ± 1.5a 0.9 ± 0.1a 0.6 ±
0.1a
2.9 ±
0.2a
9.8 ±
0.3a
4.0 ± 0.1a 3.7 ± 0.3a 2.2 ±
0.1a
1.8 ±
0.1a
3.6 ±
0.2a
5.2 ±
0.5a
2.0 ±
0.2a
1.7 ±
0.1a
20.3 ±
0.9a
Phosphorus
P0 38.1 ± 1.0b 1.1 ± 0.1a 0.7 ±
0.1a
2.4 ±
0.2b
10.3 ±
0.4a
4.0 ± 0.1a 4.3 ± 0.2a 2.5 ±
0.2a
1.8 ±
0.1a#
3.6 ±
0.2a#
6.2 ±
0.3a
2.5 ±
0.2a
1.9 ±
0.2a
20.9 ±
0.9a
P35 45.6 ± 1.3a 0.9 ± 0.1a 0.8 ±
0.2a
3.0 ±
0.2a
9.5 ±
0.1a
3.9 ± 0.0a 3.5 ± 0.3a 2.2 ±
0.1a
1.5 ±
0.1b#
2.9 ±
0.2b#
4.8 ±
0.5a
1.7 ±
0.1b
1.6 ±
0.1a
18.1 ±
0.5a
Depth
0–5 cm 44.0 ± 1.8a 1.1 ± 0.1a 0.9 ±
0.2a
2.5 ±
0.2a
9.9 ±
0.3a
3.9 ± 0.1a 3.5 ± 0.3b 2.2 ±
0.1a
1.5 ±
0.1b
3.0 ±
0.3b
4.8 ±
0.5b
1.9 ±
0.2a
1.7 ±
0.1a
19.2 ±
1.1a
5–10 cm 41.3 ± 1.8b 0.9 ± 0.1b 0.6 ±
0.2a
3.0 ±
0.3a
9.9 ±
0.4a
3.9 ± 0.1a 4.2 ± 0.3a 2.3 ±
0.2a
1.6 ±
0.1b#
3.1 ±
0.3b#
5.9 ±
0.5a
2.3 ±
0.3a
1.7 ±
0.2a
19.3 ±
0.8a
10–20 cm 40.1 ± 1.9b 0.9 ± 0.2b 0.7 ±
0.1a
2.7 ±
0.3a
10.0 ±
0.4a
4.0 ± 0.1a 4.0 ± 0.3a 2.5 ±
0.1a
1.8 ±
0.2a#
3.6 ±
0.3a#
5.8 ±
0.5a
2.1 ±
0.2a
1.8 ±
0.2a
20.1 ±
0.9a
† MP, moldboard plow; NT, no-till; P0, soil treatment with 0 kg P ha1; P35, soil treatment with 35 kg P ha1.
‡ Ortho, orthophosphate; Poly, polyphosphate; Pyro, pyrophosphate.
§-Glyc, -glycerophosphate; -Glyc, -glycerophosphate; Chol-P, choline-phosphate; Gluc-6P, glucose-6 phosphate; myo-IP6, myo-inositol
hexakisphosphate; neo-IP6, neo-inositol hexakisphosphate; Nucl, nucleotides; Phos, phosphonate; scyllo-IP6, scyllo-inositol hexakisphosphate; Res P,
residual organic P (unidentified organic P forms).
¶ For each treatment, different letters indicate significantly different according to LSD (0.05).
# For each treatment, different letters indicate significantly different according to LSD (0.1).
122
Figure 6-1 Distribution of total P (TP), Mehlich-3 extractable P (PM3) and orthophosphate
concentrations at various soil depths under (a, c, e) mouldboard plow (MP) and (b, d, f) no-
till (NT) treatments. P0 and P35 represent additions of 0 and 35 kg P ha−1, respectively.
Values are means of three replications. For each treatment, different letters indicate
significantly different means among soil depth according to LSD (0.05). † For each
treatment, different letters indicate significantly different means among depth according to
LSD (0.1).
123
Figure 6-2 Distribution of (a) total carbon (TC) and (b) total nitrogen (TN) content, and (c)
Al Mehlich-3 and (d) Mg Mehlich-3 at various soil depths under mouldboard plow (MP)
and no-till (NT) treatments. Values are means of three replicates. For each treatment,
different letters indicate significantly different means among soil depth according to LSD
(0.05).
124
Figure 6-3 Phosphorus-31 nuclear magnetic resonance spectroscopy spectrum showing the
range of P compounds detected at the 0 to 5 cm depth of the mouldboard plow fertilized
treatment (Oth.D1, other diester 1; Oth.D2, other diester 2).
125
Figure 6-4 Phosphorus-31 nuclear magnetic resonance spectroscopy spectrum showing the
P compounds detected in the monoester region at the 0 to 5 cm depth of mouldboard plow
fertilized treatment. (A) neo-IP6; (B) orthophosphate; (C) myo-IP6; (D) glucose-6P; (E)
unknown; (F) α-glycerophosphate; (G) β-glycerophosphate; (H) nucleotides; (I) choline-P;
(J) scyllo-IP6; (M1) monoester 1; (M2) monoester 2.
126
Figure 6-5 Distributions of (a) pyrophosphate, (b) scyllo-IP6, (c) DNA and (d) nucleotides
concentrations at various soil depths under mouldboard plow (MP) and no-till (NT)
treatments. Values are means of three replicates. For each treatment, different letters
indicate significantly different means among depths according to LSD (0.05).
127
CHAPITRE VII: CONCLUSIONS ET RECOMMANDATIONS
Le phosphore est un élément nutritif indispensable pour la production végétale. Il
est présent sous différentes formes qui coexistent par des mécanismes physiques, chimiques
et biologiques dans un cycle fermé. Dans les sols cultivés, les pratiques culturales
perturbent le cycle de P et peuvent affecter la production agricole et/ou contribuer à
l’eutrophisation des eaux de surface. La compréhension de la dynamique et des
changements des formes du P dans les écosystèmes agricoles est un élément clé pour une
meilleure gestion de P. Les méthodes utilisées jusqu’à date pour caractériser les
changements des pools de P se basent sur une description opérationnelle. D’autre part, les
fractions de P sont définies comme étant des données compositionnelles; soient des
données strictement positives comprises entre 0 et une quelconque unité de mesure. Selon
Aitchison (1986), l’application des statistiques conventionnelles sur ces données peut
générer des résultats erronés menant à des interprétations contradictoires. En outre, les
techniques couramment utilisées pour mesurer le P du sol sont laborieuses, relativement
lentes, coûteuses, et nécessitent l’utilisation de plusieurs extractifs chimiques. L’objectif
général de cette thèse a été de mesurer les formes du P du sol et d’étudier leurs
changements selon les pratiques culturales moyennant de nouveaux outils.
Nos objectifs étaient de (i) évaluer le potentiel de la spectroscopie dans le proche
infrarouge (SPIR) à prédire le P total, le P disponible extrait à la solution Mehlich-3 (PM3)
et à l’eau (Cp), (ii) évaluer le potentiel de la SPIR à prédire le P organique, (iii) démontrer
que l’analyse compositionnelle permet d’étudier les formes chimiques du P sans biais, et
(iv) utiliser la spectroscopie de résonance magnétique nucléaire du 31P et l’analyse
compositionnelle pour identifier les espèces ioniques et moléculaires du P et caractériser
leur distribution selon le système du travail du sol et la fertilisation phosphatée.
Ces objectifs ont été atteints et les conclusions suivantes relatives à chacun d’eux
sont tirées :
Les résultats de la première étude de cette thèse (troisième chapitre) ont démontré
que le P total est modérément prédictible par la spectroscopie dans le proche-infrarouge
pour un sol sablo-loameux au Québec. Ce modèle de prédiction pourrait être appliqué dans
des situations où les exigences de précision sont relativement faibles. Néanmoins, la SPIR
128
ne peut pas être une alternative à l’analyse chimique conventionnelle du P disponible à la
plante tel que mesuré par les méthodes de Mehlich-3 (PM3) ou à l’eau (Cp).
Dans la deuxième étude (quatrième chapitre), les résultats ont montré que le PT ne
peut pas être prédictible par la SPIR dans des sols loameux et argileux-loameux,
contrairement au PM3. Ces résultats contradictoires entre les deux études peuvent
s’expliquer en partie par la sensibilité de cette technique spectroscopique à la texture du sol,
à la méthode de référence utilisée et à la variation de la teneur du P dans le sol. De ce fait,
ces modèles de prédiction par la SPIR devraient être validés dans d’autres sites de texture
différentes et teneurs variables en P. Un résultat important dans ce chapitre est que le P
organique est prédictible directement par la SPIR dans ces sols de texture moyennement
grossière à moyennement fine dû probablement à son lien avec la matière organique. Ceci
constitue un progrès dans les techniques analytiques du Po permettant de promouvoir sa
caractérisation et son étude dans les écosystèmes, étant donné que la SPIR est une
technique rapide, directe, économique et durable de point de vue environnemental. La
troisième étude de la thèse a permis de démontrer que l’analyse statistique conventionnelle
des espèces de P déterminées par la RMN-31P est biaisée. En effet, nous avons démontré
que les résultats des analyses de variance et de corrélation des espèces de P mesurées en
proportions ou en concentrations, brutes ou log transformées, sont souvent différents
menant à des interprétations incohérentes. Ceci est le résultat de la redondance, la
dépendance de l’échelle, et la distribution non-normale des données compositionnelles.
L’utilisation des transformations du log ratio centré (clr) ou isométrique (ilr) a permis
d’éviter ce biais statistique et d’avoir des résultats fiables et cohérents. Ainsi, nous avons
révélé dans ce chapitre l’importance de l’analyse compositionnelle pour une étude non
biaisée des formes de P dans le sol, et nous recommandons son utilisation pour l’étude des
formes de P dans l’environnement.
Dans la dernière étude de cette thèse, nous avons caractérisé les espèces ioniques et
moléculaires de P et leur distribution dans des sols collectés dans une rotation maïs soya
moyennant la résonance magnétique nucléaire du 31P. Les résultats ont démontré que
l’accumulation de P dans la couche superficielle du semis direct est principalement due aux
ions orthophosphates, et elle est plus importante dans les sols fertilisés. Cependant, les
formes organiques s’accumulaient en profondeur (5–20 cm) sous forme d’inositols
129
monoesters susceptibles d’atteindre les cours d’eaux adjacents par drainage. Sur la base de
ces résultats, nous recommandons de déterminer les apports de P au sol sous semis direct
selon des analyses de P par profondeur (0-5, 5-10 et 10-20 cm) pour éviter l’accumulation
du P en surface et s’assurer d’apporter les quantités en P disponible nécessaires aux plantes.
130
131
CHAPITRE VIII: ANNEXE
COMPOSITIONAL ANALYSIS OF POOLS IN CANADIAN
MOLLISOLS
D. ABDI1,2, N. ZIADI2 and L-É. PARENT1
1Agriculture and Agri-Food Canada, 2560 Hochelaga Boulevard, Quebec City, Quebec,
Canada, G1V 2J3; 2Département des sols et de génie agroalimentaire, Université Laval,
Quebec City, QC, Canada, G1K 7P4.
Proceeding of 4th International Workshop on
Compositional Data Analysis, Spain, 2011
132
8.1 RÉSUMÉ
Le cycle du phosphore dans les Mollisols des prairies canadiennes est perturbé par
les pratiques culturales. Les modèles utilisés jusqu’à date ne prennent pas en considération
les intéractions entre les différents pools du P. L’analyse compositionnelle utilisant les
coordonnées du log-ratio isométrique (ilr) est appropriée pour modeler ces intéractions.
L’objectif de l’étude était de modéler les changements des pools de P dans des Mollisols en
fonction des pratiques culturales et du temps en utilisant les ilr. Deux bases de données
publiées ont été utilisées. Les résultats ont démontré que la rotation culturale et la
fertilisation changeaient majoritairement la balance entre les pools inorganique et
organique, et celle entre les pools les plus disponibles et les moins disponibles du P
inorganique en augmentant le risque de perte de P. Les résultats ont montré aussi des
changements importants dans l’horizon Ah au cours du temps, dus à la perte en P
organique.
133
8.2 ABSTRACT
During cultivation, the internal phosphorus cycle of Mollisols (Chernozems) of the
Canadian Prairies is perturbed by crop sequences including wheat phases, tillage practices, and
regular applications of fertilizers. To monitor these changes, a proximate sequential phosphorus
(P) fractionation procedure was developed by Hedley et al. (1982) to extract inorganic and
organic P fractions as very labile (resin-P), labile (NaHCO3-P), slowly available (NaOH-P), and
very slowly available (HCl-P) pools. Models used so far to monitor P pools do not address the
interactive behaviour of P fractions constrained to a closed compositional space. Compositional
data analysis using isometric log ratio (ilr) coordinates is appropriate for modelling the
interactive P pools using sequential binary partitions of P pools. Our objective was to model
changes of P pools in Mollisols in response to management and time using ilr coordinates. We
used a dataset with treatments and another where a Mollisol was analyzed at time zero and 4, 65,
and 90 yr after sod breakup. Seven P fractions were assigned to P reactivity groups to compute
six ilr coordinates. The ilr2 contrasting inorganic (geochemical) and organic (biological) P pools
and ilr4 between the most readily available and less P bioavailable pools were the most sensitive
to crop sequence and fertilization. Using composition at time zero as reference, the Aitchison
distance reached a plateau after the 4th year in the Bm horizon compared to continuous change in
the Ah horizon. Time changed the P balance of cultivated Mollisols primarily in the inorganic vs.
organic P pools. The risks of yield loss and environmental damage can be minimized using soil
tests that quantify the rapidly bioavailable inorganic P pools and crop management strategies that
promote biological P pools.
134
8.3 INTRODUCTION
Mollisols (Chernozems) form an important soil group used for large scale grain
production in the Canadian Prairies. They are naturally fertile soils well supplied with plant
nutrients such as phosphorus. These soils have been perturbed by tillage practices and
applications of phosphatic fertilizers and manure P. As a result, soil and crop management
influenced the internal soil P cycling of Mollisols. Phosphorus fractionation procedures can
quantify the P pools likely to change slowly or rapidly in soils under perturbation.
Hedley et al. (1982) proposed a sequential extraction procedure to chemically assess the
availability of soil organic and inorganic P forms. Cross and Schlesinger (1995) classified the
Hedley et al. (1982) interactive P pools into rapidly plant-available (resin-P and NaHCO3-P) and
refractory (NaOH-P, sonic P, HCl-P, and residual P) pools. The oxalate-extractable P (Pox)
estimates the inorganic P accumulation from fast and slow reactions with iron (Feox) and
aluminium (Alox) hydroxides (Lookman et al., 1996). Inositol phosphates that may account for
more than 50% of soil organic P may also react with Fe and Al compounds in soils (Shang et al.,
1990, 1992; Celi et al., 1999). Refractory and residual P pools contribute little to soil P cycling at
time scale required for soil management. Since organic P usually declines in soils following
cultivation (Stevenson, 1986) chemical fractionation can assess long-term change in P pools in
response to land use or soil management (Frossard et al., 2000).
Raw soil P fractions have been used to describe P distribution and model P dynamics in
soils in state-space (Shuai and Yost, 2004), path (Tiessen et al., 1983), variance, regression and
correlation analyses (e.g. McKenzie et al., 1992; Tiessen et al., 1984). Compositional data such
as raw P fractions have severe limitations for linear modelling since they are constrained to a
close space of strictly positive data spoiled by redundancy and spurious correlations. Gaussian
laws cannot be applied to those data since it is impossible to obtain analytical data less than zero
or more than 100%. For these reasons, P pools bear relative information about pool exchange
processes. Raw P concentration data must be log ratio transformed before analysis by linear
statistical procedures conceived for the real space made of both negative and positive values
(Aitchison, 1986).
135
Isometric log ratio (ilr) transformation with orthonormal bases are sequential binary
partitions (SBP) of compositional data (Egozcue and Pawlowski-Glahn, 2005). For D
phosphorus fractions, there are (D-1) SBP’s. The conceptual model of Tiessen et al. (1984)
shows mass exchange among P pools. A conceptual framework for binary partitions includes
partitions between rapidly and slowly bioavailable P pools (Hedley et al., 1982) and between
geochemical (inorganic) and biological (organic) P pools (Cross and Schlesinger, 1995). The
Aitchison distance can be computed across ilr coordinates as a distance between perturbed and
reference compositions.
The objective of this study is to present the conceptual model that describes relationships
among soil P pools and to decompose these relationships into sequential binary partitions and ilr
coordinates. Time and treatment variations in P pools are analyzed using datasets on the effect of
crop sequence and fertilization on P pools and on time change in P pools in two Mollisols of the
Canadian Prairies.
8.4 MATERIALS AND METHODS
McKenzie et al. (1992) fractionated soil P in dryland grain crop sequences (continuous
wheat, wheat-fallow and wheat-wheat-fallow) on a Lethbridge sandy clay loam (Calcareous Dark
Brown Chernozem) given nitrogen and/or phosphorus fertilizers for 14 to 19 years. On the other
hand, time change in P pools balance was modelled by Tiessen et al. (1983) in a Blaine Lake silt
loam (Orthic Black Chernozemic) following chronosequence starting with native prairie (time
zero) over 4-, 60- and 90-years of crop sequences.
The P pools were quantified using a modified Hedley et al. (1982) procedure. Sonic
P pools that account for a small fraction of total P pools were amalgamated with their
respective P pools. Residue inorganic and organic P pools that are undefined P pools were
amalgamated into a single residue P pool. A modified Tiessen et al. (1984) conceptual
model relating P pools is presented in Figure 6.1. There geochemical and biological P pools
(Cross and Schlesingere, 1995) can be partitioned into slowly and rapidly bioavailable P
pools (Hedley et al., 1982). Residue P represents the slowly bioavailable P pools as
illustrated in Tiessen et al. (1984). The main P pools for partitions were thus the
geochemical, biological, and residue P pools. The first sequential binary partition (SBP)
136
was between residue P and other P fractions (Table 6.1). The second partition was between
the geochemical and biological P pools. In the geochemical pool, there was only one
unidirectional transformation, the one from primary minerals to solution P, leading to the
third partition. Other P pools are bidirectional between rapid or slow geochemical or
biological species (Figure 6.1).
8.4.1 Isometric log ratio transformation and the Aitchison distance
A D-part composition can be described by its parts as follows
(Aitchison, 1986):
(1)
Where is the closure operator to unit . The isometric log ratio coordinate is
computed from SBP’s as follows (Egozcue and Pawlowski-Glahn, 2005):
)(
)(ln*
xg
xg
sr
rsxi (2)
Where is the geometric mean of P fractions in group and is the
geometric mean of P fractions in group . The ilr sign indicates in what direction P pools
change in response to treatment or over time. The Aitchison distance between two
compositions is computed as follows across D-1 compositional dimensions (Egozcue and
Pawlowsky-Glanh, 2006):
(3)
(4)
137
Where is the reference composition.
8.4.2 The Mackenzie et al. (1992) dataset
The response of Mollisol P pools to perturbation by cultural practices and by nitrogen (N)
and phosphorus (P) fertilization regimes investigated by McKenzie et al. (1992) is presented in
Table 6.2. In absolute terms, the P fertilization influenced more markedly the resin and NaHCO3,
and NaOH pools compared to other pools.
The effect sizes of treatments on P pools are reflected by ilr values (Table 6.3). The effect
size of added P varied with fertilization and crop sequence. The effect of added P was most
prominent in ilr2, ilr3 and ilr4 for continuous wheat and the wheat-wheat-fallow sequence while
treatment effects were much smaller in the wheat-fallow sequence (Table 3). The dominance of
wheat in the sequence affected markedly the balance between geochemical and biological pools
and that between P in primary minerals and other inorganic pools.
A clearer picture of P pools change is given by subtracting from the effect of treatments
that of no cultivation (Table 6.4). As shown by the Aitchison distance (Table 6.4), check and
added N were closest to uncultivated conditions across crop sequences, especially where wheat
was dominant. Cultivation, crop sequence and fertilization increased the inorganic and organic
pools compared to residue P, indicating more P bioavailability. The geochemical pool increased
over the biological one across crop sequences (ilr2). The slowly and rapidly bioavailable
inorganic P pools largely increased compared to P from primary minerals (ilr3) under continuous
wheat and wheat-wheat-fallow compared to the wheat-fallow sequences. In general, the balance
between slowly and rapidly bioavailable inorganic P pools (ilr5) slightly decreased compared to
the uncultivated reference composition. Differences were also small in the balance between
slowly and rapidly bioavailable organic P pools (ilr6) and the uncultivated reference
composition.
The ilr4 is the most important balance for plant nutrition and the environment. Where no
P was added, the wheat-fallow sequence showed the highest P bioavailability but also the highest
risk for eutrophication of surface waters by dissolved P. Where P was added, crop sequences
showed similar ilr4 values and Aitchison distances. These results indicate that routine soil tests
representative of ilr4 are useful tools to address both agronomic and environmental issues.
138
However, the biological dimension of the system (ilr2) can be manipulated on the long run by
crop production systems that can decrease ilr4 while maintaining the balance between organic
pools (ilr6).
8.4.3 The Tiessen et al. (1983) dataset
Time change of P pools in horizons Ah and Bm has been studied by Tiessen et al. (1983)
in a Mollisol after breakup of the natural prairie ecosystem followed by cultivation for 90 years
(Table 6.5). The most prominent change was a decrease in the biological pools in both soil
horizons.
After computing ilr coordinates at each time step, the degree of change in P balances can
be measured as changes in each ilr coordinate and globally as the Aitchison distance (Figure 2).
The slowly and rapidly bioavailable P pools decreased in both horizons compared to residue P
(ilr1) and primary minerals P (ilr3) just after sod breakup and varied chaotically thereafter. The
inorganic P pools decreased then increased (ilr2) probably due to humification following sod
breakup and to cultivation thereafter. Resin P increased as a result of mineralization of root
organic matter in the Bm horizon and decreased generally as a result of crop P uptake. Rapidly
bioavailable inorganic (ilr5) and organic (ilr6) P pools decreased compared to slowly
bioavailable ones probably as a result or lesser microbial turnover following cultivation.
Above all, the Aitchison distance across ilr coordinates increased more rapidly in the Bm
than the Ah horizons due to more rapid change in the most readily and the most slowly available
P pools (Table 6.5). Thereafter, P balances remained stationary in the Bm horizon while there
was considerable depletion of organic and residue P pools in the Ah horizon. Hence, few years
after sod breakup, major changes in P pools occurred in the Ah horizon of this Mollisol where
soil conservation practices must have largest effect on plant nutrition and environmental quality.
8.5 CONCLUSION
Using isometric log ratios to model P dynamics, the effect of crop sequences and
fertilization on Mollisol P pools was primarily related to the dominance of wheat in the sequence
and to P fertilization. More crops and less fallow in the sequence maintained soil P pools closer
to uncultivated conditions.
139
The change of P pools over time was shown by an early change in balance where slowly
to rapidly bioavailable P forms generally decreased compared to the recalcitrant residue and
primary minerals P. Overall, the organic P pools decreased in the long run in both Ah and Bm
horizons. Time change started rapidly then stabilized in the Bm horizon. The soil monotonically
moved away from native conditions in the Ah horizon primarily due to loss of organic P.
140
8.6 REFERENCE
Aitchison, J. (1986). The Statistical Analysis of Compositional Data. Monographs on Statistics and
Applied Probability. Chapman & Hall Ltd., London (UK). 416 p.
Celi, L., S. Lamacchia, F. A. Marsan, and E. Barberis (1999). Interaction of inositol hexaphosphate on
clays: adsorption and charging phenomena. Soil Science 164, 574-484.
Cross, A. F. and W. H. Schlesinger (1995). A literature review and evaluation of the Hedley
fractionation: Applications to the biogeochemical cycle of soil phosphorus in natural ecosystems.
Geoderma 64, 197-214.
Egozcue, J. J. and V. Pawlowski-Glahn (2006). Simplicial geometry for compositional data. p. 145-
159 in A Buccianti, G Mateu-Figueras, V Pawlowski-Glahn (eds) Compositional data analysis in the
geosciences: from theory to practice. Geol. Soc., London, Spec. Publ. 264.
Egozcue, J. J. and V. Pawlowski-Glahn (2005). Groups of parts and their balances in compositional
data analysis. Mathematical Geology 37, 795-828.
Frossard, E., L. M. Condron, A. Oberson, S. Sinaj, and J. C. Fardeau (2000). Processes governing
phosphorus availability in temperate soils. Journal of Environmental Quality 29, 15-23.
Hedley, M. J., J. W. B. Stewart, and B. S. Chauhan (1982). Changes in inorganic and organic soil
phosphorus fractions by cultivation practices and by laboratory incubations. Soil Science Society of
America Journal 46, 970-976.
Lookman, R., K. Jansen, R. Merckx, and K. Vlassak. (1996). Relationship between soil properties and
phosphate saturation parameters a transect study in northern Belgium. Geoderma 69 (3-4), 265-274.
McKenzie, R. H., J. W. B. Stewart, J. F. Dormaar, and G. B. Schaalje (1992). Long-term crop rotation
and fertilizer effects on phosphorus transformations: I. In a Chernozemic soil. Canadian Journal of
Soil Science 72, 569-579.
Shang, C., J. W. B. Stewart, and P. M. Huang (1992). pH effect on kinetics of adsorption of organic
and inorganic phosphates by short-range ordered aluminum and iron precipitates. Geoderma 53, 1-
14.
Shang, C., P. M. Huang, and J. W. B. Stewart (1990). Kinetics of adsorption of organic and inorganic
phosphates by short-range ordered precipitate of aluminium. Canadian Journal of Soil Science 70,
461-470.
Shuai, X. and R. S. Yost (2004). State-space modeling to simplify soil phosphorus fractionation. Soil
Science Society of America Journal 68, 1437-1444.
Stevenson, F. J. (1986). Cycles of soils. Carbon, nitrogen, phosphorus, sulfur, micronutrients. Wiley-
Interscience, NY. 380 p.
Tiessen, H., J. W. B. Stweart, and C. V. Cole (1984). Pathways of Phosphorus Transformations in
Soils of Differing Pedogenesis. Soil Science Society of America Journal 48, 853-858.
Tiessen, H., J. W. B. Stewart, and J. O. Moir (1983). Changes in organic and inorganic phosphorus
composition of two grassland soils and their particle size fractions during 60-90 years of cultivation.
Journal of Soil Science 34, 815-823.
141
Table 8-1 Sequential binary partitions of soil P fractions (r is number of P fractions with plus
sign and s is number of P fractions with minus sign).
ilr Inorganic P Organic P Residue r s Balance
coefficient Resin NaHCO3 NaOH H2SO4 NaHCO3 NaOH
1 1 1 1 1 1 1 -1 6 1 0.926
2 1 1 1 1 -1 -1 0 4 2 1.155
3 1 1 1 -1 0 0 0 3 1 0.866
4 1 -1 -1 0 0 0 0 1 2 0.816
5 0 1 -1 0 0 0 0 1 1 0.707
6 0 0 0 0 1 -1 0 1 1 0.707
142
Table 8-2 Mollisol P fractions following crop sequence and NP fertilization (data from
McKenzie et al., 1992).
Treatment Inorganic P fractions Organic P fractions Residue P
Resin NaHCO3 NaOH HCl NaHCO3 NaOH
mg kg-1
Uncultivated control
None 8 6 12 218 7.5 73 254
Continuous wheat
Check (no N nor P) 19 8 20 201 4.4 52 219
Added N 15 8 22 195 8.9 65 215
Added N and P 78 26 44 215 9.5 66 220
Added P 73 19 36 212 6.3 62 223
Wheat-wheat-fallow sequence
Check (no N nor P) 19 8 23 199 5.0 53 214
Added N 15 7 21 203 5.1 62 218
Added N and P 51 16 32 216 4.9 57 222
Added P 61 15 30 213 4.4 50 220
Wheat -fallow sequence
Check (no N nor P) 43 14 36 217 7.8 59 210
Added N 37 13 33 212 8.4 61 213
Added N and P 59 18 38 221 9.0 64 213
Added P 68 17 39 226 6.8 57 215
143
Table 8-3 Ilr coordinates of P pools following crop sequence and fertilization (data from
McKenzie et al., 1992).
Treatment Ilr1 Ilr2 Ilr3 Ilr4 Ilr5 Ilr6
Uncultivated control
None -2.342 -0.251 -2.828 -0.048 -0.490 -1.609
Continuous wheat
Check (no N nor P) -2.095 0.709 -2.278 0.332 -0.648 -1.746
Added N -1.961 0.124 -2.292 0.100 -0.715 -1.406
Added N and P -1.412 1.122 -1.361 0.682 -0.372 -1.371
Added P -1.589 1.224 -1.516 0.838 -0.452 -1.617
Wheat-wheat-fallow sequence
Check (no N nor P) -2.031 0.662 -2.229 0.275 -0.747 -1.669
Added N -2.089 0.433 -2.379 0.174 -0.777 -1.766
Added N and P -1.734 1.236 -1.719 0.664 -0.490 -1.735
Added P -1.757 1.384 -1.693 0.862 -0.490 -1.719
Wheat -fallow sequence
Check (no N nor P) -1.634 0.895 -1.777 0.531 -0.668 -1.431
Added N -1.682 0.736 -1.847 0.474 -0.659 -1.402
Added N and P -1.513 0.950 -1.613 0.664 -0.528 -1.387
Added P -1.563 1.217 -1.601 0.793 -0.587 -1.503
144
Table 8-4 Ilr differences in P pools between treatments and uncultivated check (data from
McKenzie et al., 1992).
Treatment Ilr1 Ilr2 Ilr3 Ilr4 Ilr5 Ilr6
Aitchison
distance
Distance from uncultivated control
Continuous wheat
Check (no N nor P) 0.247 0.961 0.551 0.380 -0.158 -0.137 1.475
Added N 0.381 0.376 0.536 0.148 -0.225 0.203 0.687
Added N and P 0.930 1.373 1.468 0.730 0.118 0.238 5.510
Added P 0.753 1.475 1.312 0.886 0.038 -0.008 5.251
Wheat-wheat-fallow sequence
Check (no N nor P) 0.311 0.913 0.600 0.323 -0.257 -0.060 1.464
Added N 0.253 0.684 0.449 0.222 -0.287 -0.157 0.890
Added N and P 0.608 1.487 1.109 0.712 0.000 -0.126 4.333
Added P 0.585 1.635 1.136 0.911 0.000 -0.110 5.146
Wheat-fallow sequence
Check (no N nor P) 0.708 1.146 1.051 0.579 -0.178 0.178 3.319
Added N 0.660 0.988 0.981 0.522 -0.169 0.207 2.718
Added N and P 0.829 1.201 1.215 0.712 -0.038 0.222 4.164
Added P 0.779 1.468 1.227 0.841 -0.097 0.106 4.998
145
Table 8-5 Effect of time on P pools in a Mollisol (data from Tiessen et al., 1984).
Time Inorganic P Organic P Residue P
Resin NaHCO3 NaOH H2SO4 NaHCO3 NaOH
year mg kg-1
Ah horizon
0 25.8 13.5 31.0 174 49.5 167 337
4 22.5 13.4 32.7 177 52.0 177 416
65 14.8 10.4 32.6 200 31.7 124 318
90 21.6 11.2 32.0 196 19.5 75 273
Bm horizon
0 9.1 5.3 18.4 199 16.6 63 236
4 8.0 3.3 18.4 190 16.5 72 290
65 6.3 4.2 22.8 221 15.1 66 242
90 7.1 3.9 17.9 228 9.7 47 254
146
Figure 8-1 Conceptual relational model between P pools in Mollisols (modified from
Tiessen et al., 1984).
147
0.000
0.200
0.400
0.600
0.800
1.000
1.200
0 20 40 60 80 100
Time elapsed until breakup (year)
Ait
chis
on d
ista
nce
Ah
Bm
Figure 8-2 Time change in P balance distances from initial conditions in a Blaine lake soil
(data from Tiessen et al., 1983).