Hyperspectral Cloud-Clearing

Hyperspectral Cloud-Clearing

Allen Huang, Jun Li, Chian-Yi Liu, Kevin Baggett, Li Guan, & Xuebao Wu

SSEC/CIMSS, University of Wisconsin-Madison

•AIRS/AMSU V3.5 &V4.0 Cloud-Clearing Characteristic

•AIRS/AMSU Additional C.C. QC Using MODIS

•AIRS/MODIS Over Sampling Single-FOV N* C.C.

•AIRS/MODIS 2-FOV Vs. AIRS/AMSU 9-FOV C.C. Comparison

5th Workshop on Hyperspectral Science of UW-Madison MURI, Airborne, LEO, and GEO Activities

The Pyle Center

University of WisconsinMadison

7 June 2005

Case Granule Dataset Used

•4 Granules of CollocatedAIRS & MODIS Data•MODIS 1-km Cloud Mask•AIRS C.M. (from MODIS)•No ancillary data used

Australia Granule

Wisconsin Granule

South Africa Granule Hurricane Isabel Granule

2 Sep. 2002AIRS Focus Day

17 Sep. 2003

Why cloud_clearing?• AIRS clear footprints are less

than 5% globally.

Why use MODIS for AIRS cloud-clearing?

• Many AIRS cloudy footprints contain clear MODIS pixels; it is effective for N* calculation and Quality Control for CC radiances

• Cloud-clearing can be achieved on a single footprint basis (hence maintaining the spatial gradient information);

• There is a direct relationship between MODIS and AIRS radiances because they see the same spectra region.

AIRS All Observations

AIRS Clear Only Obs.

Case Global Dataset Used

•240 Granules of CollocatedAIRS & MODIS Data•MODIS 1-km Cloud Mask•AIRS C.M. (from MODIS)•No ancillary data used

AIRS Global Window Channel Brightness Temp. ImagesAscending Daytime Passes

(Upper Left)Descending Nighttime Passes

(Lower Right)

AIRS/AMSU (3 by 3 AIRS FOV) V3.5 Vs V4.0 C.C. Comparison

X-axis:V3.5 C.C. Bt.- Blue circleV4.0 C.C. Bt. – Red crossY-axis: Clear MODIS Bt.

Without Q.C.

MODIS Band 223.95 microns



X-axis:V3.5 C.C. Bt.- Blue circleV4.0 C.C. Bt. – Red CrossY-axis: Clear MODIS Bt.

With Q.C.


Q.C. filtered most of the unreliable data as well as some good data.


X-axis:V3.5 C.C. Bt.- Blue CircleV4.0 C.C. Bt. – Red CrossY-axis: Clear MODIS Bt.

Without Q.C.





With Q.C.





Without Q.C.





With Q.C.



Estimated AIRS/AMSU C.C. Bias and RMSEWithout (Green) and With (Blue) MODIS as Q.C.Australia Granule

Use MODIS data as AIRS/AMSU C.C. Q.C. can enhance yields & performance

Estimated AIRS/AMSU C.C. Bias and RMSEWithout (Green) and With (Blue) MODIS as Q.C.South Africa Granule


Estimated AIRS/AMSU C.C. Bias and RMSEWithout (Green) and With (Blue) MODIS as Q.C.Wisconsin Granule


AIRS/MODIS Synergistic C.C. can Supplement AIRS/AMSU C.C.Especially over Desert Region

AIRS/AMSU C.C.(3 by 3 AIRS FOV)V3.5 – BlueV4.0 – Green

AIRS/MODIS C.C.(1 by 2 AIRS FOV)Red

Aqua MODIS IR SRF Overlay on AIRS Spectrum

Direct spectral relationship between IR MODIS and AIRS provides unique application of MODIS in AIRS cloud_clearing !

Threshold for AIRS Pair C.C. Retrieval:

Each AIRS footprint within the C.C. pair (2 by 1 ) must have at least 15 MODIS confident clear (P=99%) pixel (partly cloudy)

MODIS Bands

Used in C.C.

MODIS Bands Used in Q.C.

Multi-band N* 22 24 25 28 30 31 32 33 34

22 24 25 28 30 31 32 33 34

Single-band N* 31or 22 22 24 25 28 30 31 32 33 34

MODIS/AIRS Synergistic N* Cloud Clearing

MODIS/AIRS Synergistic Single-Channel N* Cloud-ClearingGeneral Principal

After Smith

Or Q.C.

MODIS/AIRS Synergistic Multi-Channel N* Cloud ClearingGeneral Principal

2)]([(1

*)( cci

clrM

i i

RfRNJi

2*

*21

)]1

([(1

*)(N

NRRfRNJ i

clrM

i ii

0)(

*

*

N

NJ

iii

clrMi

i

iii

clrMi

i

RfRfRRf

RfRfRRf

N

i

i

)]()(][)([1

)]()(][)([1

2122

2112

*

Li et al, 2005, IEEE-GRS

Step 1: Get cloud-cleared AIRS radiances for Principal and supplementary footprints.Step 2: Find best Rcc by comparing with MODIS clear radiances observations in the principal footprint.

(1) Principal footprint has to be partly cloudy, while the supplementary footprint can be either partly cloudy or full cloudy.

(2) The 3 by 3 box moves by single AIRS footprint, therefore each partly cloudy footprint has chance to be cloud-cleared

85

4

3

2

1 6

7

1ccR

2ccR

3ccR

4ccR

5ccR

6ccR

7ccR

8ccR

MODIS/AIRS Synergistic N* Cloud ClearingAIRS Two-FOVs (Pair) Strategy

June issue of IEEE Trans. on Geoscience and Remote Sensing, 2005

1 2 3

4

567

8

8 possible AIRS pairs (2 FOVs)

1

MODIS/AIRS Synergistic N* Cloud ClearingOver Sampling Strategy

1

Pseudo Single AIRS FOV

2 3

4

567

8


1 2


2


1

2 3

4

567

8


1 2 3


3


1

Single-Channel Vs. Multi-Channel N* C.C. Error Comparison

South Africa Granule


Wisconsin Granule


Hurricane Isabel Granules

3.96 um 4.52 um

6.7 um12.0 um


Single-Channel N* Channel Selection (22 Vs. 31) C.C. Error Comparison

Australia Granule

Single-Channel N* Channel Selection (22 Vs. 31) C.C. Error Comparison

Wisconsin Granule

Multi-Channel N* Desert vs. Land C.C. Error Comparison

Single-Channel N* Different Q.C. (0.5K vs. 1.0K) C.C. Yield & Error Comparison



Wisconsin Granule


Hurricane Isabel Granules

Wisconsin Granule

AIRS/AMSU C.C. (3 by 3 AIRS FOV) V4.0 - BlueAIRS/MODIS C.C. (1 by 2 AIRS FOV)Multi-Ch. - BlackSingle-Ch.:Band 31 – Green; Band 22 - Red


AIRS/AMSU C.C.(3 by 3 AIRS FOV)V4.0 - Blue

AIRS/MODIS C.C.(1 by 2 AIRS FOV)Multi-Ch. - BlackSingle-Ch.:Band 31 – GreenBand 22 - Red

Australia Granule


AIRS/AMSU C.C.(3 by 3 AIRS FOV)V4.0 - Blue

AIRS/MODIS C.C.(1 by 2 AIRS FOV)Multi-Ch. - BlackSingle-Ch.:Band 31 – GreenBand 22 - Red


AIRS BT(All)

AIRS BT(Clear Only)

AIRS BT(Cloud-Cleared)

AIRS BT(All)

AIRS BT(Clear Only) AIRS BT

(Cloud-Cleared)

Cloud-Cleared Vs. Cloud-Contaminated RetrievalDetails see Wu/Li Retrieval presentation

Synergistic AIRS/MODIS C.C. Summary

• Synergistic AIRS/MODIS C.C. could provide cloud-cleared radiances over non-oceanic scenes with good yield and performance at high spatial resolution (pseudo single AIRS FOV)

• AIRS/MODIS C.C. is one of the promising GOES-R risk reduction research for future HES cloudy sounding processing

• Since no Geo-microwave sensor is been planned, synergistic HES/ABI C.C. might become one of the baseline processing approach that worthy of further study

Hyperspectral Cloud-Clearing

Documents

Transcript of Hyperspectral Cloud-Clearing