1

Une description statistique multi-variable Une description statistique multi-variable des nuages au dessus de l’océan tropical à des nuages au dessus de l’océan tropical à partir des observations de jour de l’A-train partir des observations de jour de l’A-train en haute résolution spatiale pour évaluer la en haute résolution spatiale pour évaluer la paramétrisation des processus nuageuses paramétrisation des processus nuageuses

dans les modèles climatiques. dans les modèles climatiques.

D. Konsta, H. Chepfer, J-L. Dufresne,

S.Bony, G. CesanaLaboratoire de Météorologie Dynamique LMD

Institut Pierre Simon Laplace IPSL, Paris

Seminaire INTRO, 30 September 2010

2

Climate ModelsObservations

Evaluation of clouds in climate models

Data processing (starting from level1) : Lidar CALIPSO (Cloud cover:

330m,Vertical structure: 30m) Radiometer PARASOL (reflectance: 6km)Radiometer MODIS (reflectance: 250m-

500m-1km)

CFMIP-OBS: observational datasets consistent with the simulator

CALIPSO – GOCCPPARASOL- reflectance in 1constant

direction (θv=30°, φv=320°)

COSP Simulator: - Subgrid cloud simulator-SCOPS- Lidar simulator- PARASOL simulator

Simulated DatasetsCALIPSO-likePARASOL-like

consistency

LMDZ5LMDZ-New Physics

3

Zonal Mean Cloud Fraction – monthly mean

Latitude

Pre

ssur

e (h

Pa)

Latitude

CALIPSO-GOCCP OBS LMDZ New Physics +SIM

LMDZ5 LMDZ New Physics

0

0.3

LMDZ5Overestimate:- High cloudsUnderestimate:- Tropical low/mid clouds- Congestus- Mid level mid lat

LMDZ New PhysicsBetter representation of clouds

0

0.3

LMDZ5+SIM

4

Cloud Cover and Cloud Vertical Distribution in circulation regimes- Monthly mean

ω500 (hPa/day) ω500 (hPa/day)ω500 (hPa/day)

Pre

ssur

e

CF CALIPSO-GOCCP CF LMDZ5+SIM CF LMDZ new+SIM

OBSERVATIONS:- Subsidence regimes → Strong presence of low stratiform clouds- Convective regimes → clouds at high troposphere + mid level clouds

Tropical ocean

LMDZ5:-underestimation of low level clouds-no mid level clouds-overestimation of high convective clouds

LMDZ New Physics:-representation of boundary layer clouds in all regimes-overestimation of mid level clouds in one single layer-fewer high clouds

5

Clouds Optical depth (or Reflectance)

Radiometer PARASOL: directional reflectances, selection of one constant single direction (θv=30°, φs- φv=320°)

Reflectance is a proxy of optical thickness

Spherical Particles

Non Spherical Particles

(for θs=30°)

Optical ThicknessR

efle

ctan

ce

50 0

0.9

0PARASOLReflectance 1constant direction

6

Cloud Cover and All Sky Reflectance – monthly mean

Error compensations between Total Cloud Cover and Optical depth (vertically integrated value within the lat x lon grid box) → Need to evaluate the cloud parameterizations in climate models

ALL SKY REFLECTANCE

CLOUD FRACTION

LMDZ New Physics+parasol simulator

LMDZ5+lidar simulator

PARASOL 1con.dir.OBS

CALIPSO-GOCCPOBS

LMDZ New Physics+lidar simulator

LMDZ5+parasol simulator

7

Cloud Cover and Cloud Optical Depth in circulation regimes- Monthly mean

CLOUDY REFLECTANCE CLOUD FRACTION

ω500 (hPa/day) ω500 (hPa/day)

OBS

LMDZ5+SIM

LMDZ new+SIM

Tropical ocean

-Subsidence regimes: strong underestimation of cloud fraction but strong overestimation of cloud optical depth (less from LMDZ New Physics)-Convective regimes: underestimation of cloud cover and cloud optical depth

→ Need to evaluate the cloud parameterizations in climate models

• To evaluate the cloud parameterizations in climate models:

Monthly mean observations are not sufficient

We need to use high resolution (spatial and temporal ) observations

8

9

A case study: low tropical boundary layer clouds- high resolution obs -

Impact of the spatial resolution of the sensorsNeed a clean separation clear/cloudyNeed colocated and simultaneous observations

CALIPSO Level 1

CALIPSO-GOCCP

CLOUDSAT

Reflectance MODIS 1km

Reflectance MODIS 500m

Reflectance MODIS 250m

Reflectance CALIPSO 125m

CF MODIS 5km

CF PARASOL 18.5km

Alti

tude

(km

)

Latitude

Latitude

Long

itude

Clo

ud F

ract

ion

0

1 Reflectance

0

0.2MODIS

CALIPSO

PARASOL

A methodology: from the case study to global statistics using high spatial resolution data

Reflectance MODIS 250m

CDF

PDF

1-CF

1°

1°

All Sky Refl=0.04

Cloudy Refl=0.07

Clear Refl=0.02

=0.4

=0.6

Same methodology for simulator’s outputsIn each grid box (obs/mod): Cloud Fraction and Cloudy Refl

11

Cloud Optical DepthEvaluation of the model at high resolution

→ Overestimation of low values of All-Sky Reflectance and underestimation of high values. BUT for cloudy reflectance (no clear sky contribution):→ More optically thick clouds and less optically thin clouds simulated.

OBS- PARASOL

Cloudy Reflectance0 0.80.2 0.4 0.6

All SKy Reflectance0 0.80.2 0.4 0.6

LMDZ5+SIM

LMDZ new+SIM

PDFOptical thickness (spherical particles and θs=30°)

0 40.53.4 8.1 16.5

Corresponding CDF: 50% of the cloud: Obs optical depth = 2.6 Models cloud optical depth = 4.8

Tropical ocean

12

Relationship between Cloud Cover and Cloud Optical Depth

OBSERVATIONS Tropical ocean

Cloud Fraction0 1

All

Sky

Ref

lect

ance

0

0.6

Clo

udy

Ref

lect

ance

0

0.6

Obs monthly

Obsdaily

Obsdaily

→ The relation between optical depth and Cloud Fraction changes with the scale of averaging changes in time (monthly.vs. daily) and in space (all sky .vs. cloudy)

13

Relationship between Cloud Cover and Cloud Optical Depth

Tropical ocean

Cloud Fraction0 1

All

Sky

Ref

lect

ance

0

0.6

Clo

udy

Ref

lect

ance

0

0.6

Obs monthly

Obsdaily

Obsdaily

Cloud Fraction0 1

Cloud Fraction0 1

LMDZ5daily

LMDZ newdaily

LMDZ5daily

LMDZ newdaily

LMDZ5monthly

LMDZ newmonthly

→ Model has difficulties to reproduce the ‘instantaneous’ relationship

=> Here after, we use « High Resolution » : Cloudy Refl, Daily

14

Relationship between Cloud Cover and Cloud Optical Depth for high and low tropical oceanic clouds

Clo

udy

Ref

lect

ance

Clo

udy

Ref

lect

ance

1

0

1

0

Cloud Fraction0 1

Cloud Fraction0 1

Cloud Fraction0 1

Highclouds

Low clouds

OBS

OBS

LMDZ5

LMDZ5

LMDZ- new

LMDZ- new

• Error compensation between optically thin high clouds and very thick boundary layer clouds• Underestimation of the Cloud Fraction

Tropical ocean

15

Cloud Cover versus Vertical distribution versus Cloud ReflectanceP

ress

ure

Cloud Fraction (CF(p)/CFtot) Cloud Fraction (CF(p)/CFtot)

OPTICALLY THIN CLOUDS OPTICALLY THICK CLOUDS

OBS- PARASOL

LMDZ5+SIM

LMDZ new+SIM

0 0.6 0 0.6

Tropical ocean

16

Focus on low level boundary layer clouds:Relationship between the Cloud Top Pressure and the Cloudy Reflectance

Tropical ocean

Cloudy ReflectanceOr Optical depth

Cloudy Reflectance Cloudy Reflectance0.05 0.3

Pto

p

OBSERVATIONS LMDZ5+SIM LMDZ new+SIM

0.2 0.9 0.2 0.9

• OBS: The cloud optical depth increases with the cloud top altitude (and with the cloud cover) → the cloud grows vertically (and horizontally)

• Difficulties of the model to reproduce the relationship

17

ConclusionsA statistical view of clouds with A-train observations: • simultaneous and independent observations of multiple cloud parameters at high

resolution→ assess cloud process parameterization in climate models • the spatial resolution of different sensors and the temporal resolution of the statistical

analysis are critical• study of cloud properties only (excluding ‘Clear sky’ contribution)• link between Cloud Cover, Vertical Structure and Cloud Optical Depth• low clouds: cloudy reflectance increase with the cloud top altitude

LMDZ model evaluation: • Error’s compensations between

- underestimation of low tropical clouds/ few medium clouds and overestimation of high clouds

- underestimation of the total Cloud Cover and overestimation of the Cloud Optical Depth (mainly in regions of subsidence)

- Optically thinner high clouds and optically thicker boundary layer clouds • Better representation of clouds from LMDZ New Physics

Perspectives:• Similar analysis based on “high resolution” A-train observations to evaluate other

climate models • Analysis of the subgrid variability (observations and models)

18

Thank you!