Towards retrieving 3-D cloud fractions using Infrared Radiances from multiple sensors Dongmei Xu JCSDA summer colloquium, July 27 2015- August 7 2015.

Towards retrieving 3-D cloud fractions using Infrared Radiances from multiple sensors

Dongmei Xu

JCSDA summer colloquium, July 27 2015- August 7 2015

• 2015.1— Postdoc Researcher in MMM Laboratory in NCAR

• 2010.9—2014.12 PhD Candidate of Meteorology, Nanjing University of Information Science & Technology (NUIST)

• 2007.9—2010.6 Master of Meteorology, NUIST

• 2003.9—2007.6 Bachelor of Information and Computation Science (Department of Maths), NUIST

PhD thesis:

Data assimilation and synthetically retrieving clouds with satellite infrared radiance observations

Data assimilation Retrieving clouds

cloud detection

clear-sky radiance

Data assimilation

Verification using multi-

sensor IR radiance

Multivariate and Minimum Residual

(MMR) method

Multi-sensor Advection-

Diffusion nowCast

system

Implemented channel dependent cloud detection scheme and the Metop-2 IASI radiance DA facility inWRFDAEvaluated the impact of data assimilate Metop-2 IASIIn WRFDA on the forecasts on typhoon and hurricane casesIn both hybrid and 3dvar frameworks.

Why?• Cloud parameters, such as cloud top pressure

and effective cloud fraction, are useful for cloud initialization in numerical weather prediction

(NWP), to understand their impact to the earth’s climate change, to estimate incoming and outgoing thermal radiation

budget …

• It is crucial to develop a fast and efficient algorithm to estimate real-time global cloud information in NWP studies to achieve fresh cloud analysis products.

MMRMMRCRTM(clear)CRTM(clear)

T, Q

Ts, εs

Cloud fractions

obs

Clear Simulated Tb

k_top ： the cloud top level

k_topc 0.01

How? MMR (Multivariate and Minimum Residual Method)

WRFWRFDA

ν_clearR

ν_cloud 0 1 2 n ν_clear1

( , , , ..., ) 0 _ k

n

kk

R c c c c c R c R

ν_cloud obs

clear

1( , )

2

2

_0

_

R RJ c

Rc

ν _ obsR

ν _ kR :the radiance calculated for overcast black cloud at level k.

ν_cloudR

1 2 nc c , c , ..., c

0c : the fraction of clear sky

:the array of vertical effective cloud fractions for K model levels

:the observed radiance

:the modeled cloudy radiance

:the radiance calculated in clear sky at the wavenumber v

(1)

(2)

The formulations:

0

1

0 1, 0

1,

[ , ]

k

nk

k

c k k

c c

with

DATA

the nadir

AIRS IASI

IASI

before after

AIRS

afterbefore

Main results

Inter-comparisons among cloud retrievals from different sensors

Cloud mask

Cloud top pressure

Cloud profile Cloud retrievals from Multi-sensor Advection-

Diffusion nowCast system

Cloud mask

AIRS

GOES-Sounder

IASI

MODISGOES-Imager

GOESproductsas reference

(NASA-Langley cloud and radiation

products)

1900 UTC 03 June 2012

IASI MODIS

AIRS GOES-Sounder

GOES-Imager

Cloud mask

0800UTC

0900UTC

1100UTC

AIRS MODIS

Cloud top

Cloud profile

CloudSat AIRS MODIS

0800UTC

0900UTC

1100UTC

Cloud profile

Date/height AIRS (ets/corr)

MODIS (ets/corr)

0800 UTC/9 km0800 UTC/5.5 km0800 UTC/3 km0800 UTC/2.5 km

0.32/0.600.36/0.610.45/0.640.34/0.56

0.28/0.510.25/0.550.44/0.630.41/0.60

0900 UTC/9 km0900 UTC/5.5 km0900 UTC/3 km0900 UTC/2.5 km

0.27/0.490.43/0.600.62/0.780.56/0.70

0.31/0.500.40/0.540.47/0.650.39/0.55

1100 UTC/9 km1100 UTC/5.5 km1100 UTC/3 km1100 UTC/2.5 km

0.10/0.150.40/0.620.30/0.460.26/0.39

0.11/0.190.41/0.590.38/0.610.41/0.63

Cloud profile

Multi-sensor Advection-Diffusion nowCast system

AIRS IASI

MODISGOES-Sounder

GOES-Imager

（ age_index ）

2012060300-2012060323

AIRS IASI

MODIS GOES-Sounder

GOES-Imager

• Investigating a new retrieval prototype based on the Particle Filter (PF) algorithm in the framework of GSI (Gridpoint Statistical Interpolation system)

• Conducting comparisons between the MMR and PF methods.

Ongoing work

…..

…..

100%

90%

10%

.

.

.

…….

PF

APF

CloudSat MMR AIRS

PF APF

Weights for different particles

Conclusions• The MMR method is proved to be robust in

retrieving the quantitative cloud mask, using radiances from multiple satellites.

• MMR produced realistic cloud top pressures, with an accuracy varying with the sensors’ spectral resolutions.

• The accuracy of the MMR scheme in detecting mid-level clouds was found to be higher than for higher and lower clouds.

• The development of a new prototype of cloud retrieval scheme base on particle filter is underway.

Thanks for your time !

Channels’ weight functionpeak

clrv

kv

R

Rindexability _

Channels’ abilities to identify clouds

Smaller values: sensitive to clouds

close to 1: hard to identify clouds