Utilization of Land Surface Emissivity
for Precipitation Retrieval – An
Obvious Linkage between ITWG and
IPWG – and Implications for GPM-era
Algorithms
Ralph Ferraro, Sid Boukabara, Fuzhong Weng
NOAA/NESDIS
College Park and Camp Springs, MD USA
Nai-Yu Wang
ESSIC, College Park, MD
AND
PMM Science Team Land Surface Characterization Working Group & Affiliates –
including Gail Skofronick-Jackson, Christa Peters-Lidard, Catherine Prigent, Fatima
Karbou, Chuntao Liu, Joe Turk, Li Li, Sarah Finn, Ken Harrison…
17th International TOVS Study Conference Pacific Grove, CA April 2010
1
Motivation for this talk…
• We need accurate Є retrievals to improve the
development of physically based precipitation
retrievals over land
– Q signal (low freq.) over land < Є variability
– IWP signal (high freq. ) over land < Є variability
– GPM mission focus…international constellation!
• Enhance collaborations between IPWG and ITWG
– Build off of progress made at 2009 Toulouse workshop
and gear towards IPWG-5 (Hamburg, October 2010)
– IPWG and GPM need Є under cloudy/precipitating
conditions from 10 – 190 GHz
17th International TOVS Study Conference Pacific Grove, CA April 2010
2
Outline
• Scientific need
• PMM Science Team emissivity experiment
– What we are doing
– Preliminary results
• Summary and next steps
– Intercomparison study
– ITWG/IPWG Synergy
17th International TOVS Study Conference Pacific Grove, CA April 2010
3
How we can use Є information to
improve precipitation retrievals?
Clear sky
T/Q Profiles
Surface
Emissivity
Atmosphere
Emission
scattering
Precipitation
Microphysical
Information
Emissivity
Model
Land Surface
Type
Radiative Transfer
Model
Constrain Bayesian
through a priori
databases
Simulated
TB
Observed
TB
AMSU, SSMIS
Slide courtesy of N-Y. Wang
17th International TOVS Study Conference Pacific Grove, CA April 2010
4
Sensitivity of Surface to Q and V
Ts = 280 K, RAIN (Q=0.40), no-scattering
Dry: V = 10 mm
Moist: V = 60 mm
17th International TOVS Study Conference Pacific Grove, CA April 2010
5
Sensitivity on Retrievals
Results courtesy of B. Johnson/G. Skofronick-Jackson
• Examination of lake
effect snow bands from
C3VP
• Impacts of incorrect Є
– 5-10 K @89 GHz
– 1 – 3 K @ 183+7 GHz
• This translates up to
100% error in retrieved
snowfall rates (0-2
mm/hr)
17th International TOVS Study Conference Pacific Grove, CA April 2010
6
Study
Domain
A diverse set of targets were selected:
•C3VP – 44 N, 80 W
•Amazon(2) – 7 S, 70 W and 2 N, 55 W
•Open Ocean(3) – 0 N, 150 W; 35 N, 30 W; 45 S, 35 W
•Desert – 22 N, 29 E
•SGP – 35 N, 97 W
•Inland Water – 48 N, 87 W
•SE US (HMT-E) - 34 N, 81 W
•Wetland surface - 18 S, 57 W
•Finland – 60 N, 25 E
17th International TOVS Study Conference Pacific Grove, CA April 2010
7
GPM Land Surface Working Group (LSWG) Study Sites
(Some obtained from ITWG Land Working Group)
Finland
Inland Water
C3VP
SGP
Ocean-2
HMT-E
Desert
Amazon-2
Ocean-1
Amazon-1
Wetland
Ocean-3
17th International TOVS Study Conference Pacific Grove, CA April 2010
8
Study Parameters
• 12 Targets/9 types of surfaces
• 1 Year: 1 July 06 – 30 June 07
• Assemble data sets:
– Satellite
• AMSR-E, SSMI, SSMIS, TMI, AMSU, WindSat
– Ancillary satellite
• ISCCP, PR/VIRS, CloudSat, CMORPH
– Model
• GDAS, LSM, JCSDA Emissivity
• Participants generate Є “their way” but:
– Must use only the data sets supplied
– Make results freely accessible by others (post on web)
• Results to be stratified by site, cloud mask, rain, etc.
17th International TOVS Study Conference Pacific Grove, CA April 2010
9
Study Web Page
http://cics.umd.edu/~rferraro/LSWG.html
17th International TOVS Study Conference Pacific Grove, CA April 2010
10
Emissivity Estimates Received
Algorithm
Sensor(s)
Targets
Algorithm Type
NESDIS (Boukabara)
AMSR-E;
AMSU/MHS
All
1DVAR
GSFC (Wang &
Skofronick-Jackson)
AMSU/MHS
SSMIS
C3VP
Physical (remove atm.)
CICS (Wang/Gopalan)
AMSR-E; TMI
Amazon
Physical (remove atm.)
LSM/CRTM (PetersLidard/Harrison)
AMSR-E; AMSU
SSM/I
All
Physical – surface from
LSM, GDAS and CRTM
NRL (Turk/Li)
WindSat
All
Physical – surface from
satellite, MLE
CNRS (Prigent)
AMSU; SSM/I
All
Physical (remove atm.)
Meteo-France (Karbou) AMSU; SSM/I
All
Physical (remove atm.)
17th International TOVS Study Conference Pacific Grove, CA April 2010
11
Some results…caveats
• Not truly an “apples to apples” comparison
– Not all estimates used exactly the same set of satellites or
our input data
– Some estimates are 1 degree areal average some are
center of the box
– Some give exact observation time, some 12 hour
increments, …
• Did daily averages for all sensor types
• My inefficiencies in computer programming and
database management
• Goal - NOT trying to determine which estimate is the
best, rather, where they are similar and where they
are not
17th International TOVS Study Conference Pacific Grove, CA April 2010
12
HMT-E Site – SSM/I
e19V
HMT (ALL)
MEAN 0.955
HMT (ALL)
STD
0.013
HMT (No Rain) MEAN 0.954
HMT (No Rain) STD
0.013
e19H
0.935
0.013
0.934
0.013
PRIGENT
e22V e37V e37H
0.965 0.941 0.924
0.020 0.015 0.014
0.966 0.939 0.923
0.019 0.014 0.014
e85V
0.929
0.049
0.932
0.021
e85H
0.913
0.054
0.915
0.022
e19V
0.956
0.012
0.956
0.010
e19H
0.938
0.019
0.938
0.017
e22V
0.957
0.014
0.958
0.012
KARBOU
e37V e37H
0.949 0.935
0.013 0.022
0.949 0.934
0.012 0.021
e85V
0.941
0.023
0.941
0.018
e85H
0.925
0.037
0.924
0.029
Prigent - SSMI - HMT
Emissivity
• This is a site with annual
cycle in vegetation, some
winter snowfall
• All data
– Similarity 19VH & 37VH
– Prigent saturation @ 22V
– Response to rain evident
1
0.98
0.96
0.94
0.92
0.9
0.88
0.86
0.84
0.82
0.8
e19V
e37V
e22V
e85V
1
15
29
43
54
65
77
89 101 117 128 138 155 167 178 205 223 238 254 338
Day
Karbou vs. Prigent - HMT - Rain Free
1.000
• Rain free data
– Data are in closer
agreement, mean & Std.
– Still see daily variations up to
5%
Emissivity
0.980
K - 22V
0.960
K - 85V
0.940
P - 22V
P - 85V
0.920
0.900
0.880
6
41
73
89
130
157
178
230
276
Day
17th International TOVS Study Conference Pacific Grove, CA April 2010
13
HMT-E Site – AMSU
e23
HMT (ALL)
MEAN 0.942
HMT (ALL)
STD
0.012
HMT (No Rain) MEAN 0.941
HMT (No Rain) STD
0.012
KARBOU
e31
e50
e89
0.943 0.943 0.944
0.011 0.016 0.028
0.942 0.945 0.951
0.011 0.013 0.011
PETERS-LIDARD
e23 e31
e50
e89
0.942 0.948 0.958 0.950
0.012 0.010 0.008 0.003
0.944 0.950 0.959 0.950
0.011 0.010 0.007 0.001
• All data
e23
0.930
0.010
0.931
0.010
BOUKABARA
e31
e50
0.935 0.930
0.010 0.011
0.936 0.931
0.009 0.011
e89
0.935
0.011
0.936
0.011
AMSU- BOUKABARA- HMT-E
Emissivity
– K and P-L similar at 23
and 31 GHz
– B lower than K and P-L
– Other differences noted
1.000
0.980
0.960
23GHz
0.940
0.920
31GHz
89GHz
0.900
0
25
50 75 100 125 150 175 200 225 250 275 300 325 350
Day(2007- 2006)
• Rain free data
AMSU CH2 (31.4 GHz) Emissivities - HMT-E
1.000
Em issivity
– Not much difference in
P-L and B (and K, accept
at 89 GHz)
0.980
0.960
KARBOU
0.940
PETERS-LIDARD
BOUKABARA
0.920
0.900
0
25
50
75 100 125 150 175 200 225 250 275 300 325 350
Day (2007 - 2006)
17th International TOVS Study Conference Pacific Grove, CA April 2010
14
Soil
Moisture
Land Sfc
Temperature
10V 10H
Emissivity
HMT‐E Site
34N 81W
1 Jun 2006‐1 Jul 2007
Red impulses indicate 3‐hourly CMORPH accumulated precipitation average
18V 18H
Emissivity
23V 23H
Emissivity
37V 37H
Emissivity
17th International TOVS Study Conference Pacific Grove, CA April 2010
15
C3VP Site – SSM/I
PRIGENT
e22V e37V e37H
0.932 0.912 0.870
0.030 0.022 0.031
0.939 0.914 0.867
0.034 0.031 0.039
• This is a site with long
periods of snowcover and
annual vegetation changes
• Snow cover evident in 37
and 85 GHz by P and K
• K tends to be lower and
more variable – Great
Lakes influence in my
processing; needs to be
redone correctly…
e85V
0.916
0.026
0.906
0.049
e85H
0.879
0.044
0.868
0.060
e19V
0.893
0.062
0.891
0.058
e19H
0.830
0.113
0.809
0.101
e22V
0.901
0.062
0.896
0.059
KARBOU
e37V
0.903
0.057
0.893
0.059
e37H
0.848
0.122
0.825
0.113
e85V
0.912
0.053
0.897
0.066
e85H
0.862
0.118
0.837
0.117
Prigent - SSMI - C3VP
1.000
0.980
0.960
0.940
Emissivity
e19H
0.861
0.027
0.861
0.032
0.920
e19V
0.900
e37V
0.880
e22V
0.860
e85V
0.840
0.820
0.800
2
19
29
40
51
61
72
87
109
132
160
179
250
313
Day
Prigent vs. Karbou - SSMI - C3VP
Rain Free
1.000
0.980
0.960
0.940
Emissivity
e19V
C3VP(ALL)
MEAN 0.914
C3VP (ALL)
STD
0.020
C3VP (No Rain) MEAN 0.924
C3VP (No Rain) STD
0.031
0.920
P -19V
0.900
P - 85V
K - 19V
0.880
K - 85V
0.860
0.840
0.820
0.800
2
19
29
40
51
61
72
87
109
132
160
179
250
313
Day
17th International TOVS Study Conference Pacific Grove, CA April 2010
16
C3VP Site – AMSU
e23
0.830
0.060
0.813
0.059
• Large variations
between estimates
– View angle handling?
– B is N18 only
– Water/Ice contamination
in K
• B shows least variability
• Some issue with P-L 89
GHz
PETERS-LIDARD
e23 e31
e50
e89
0.946 0.952 0.959 0.944
0.017 0.015 0.010 0.013
0.955 0.954 0.943 0.920
0.023 0.015 0.012 0.022
e23
0.932
0.008
0.931
0.007
BOUKABARA
e31
e50
0.936 0.921
0.008 0.008
0.935 0.921
0.006 0.007
e89
0.915
0.008
0.916
0.008
Peters-Lidard - AMSU Emissivities - C3VP
Emissivity
MEAN
STD
MEAN
STD
1.000
0.990
0.980
0.970
0.960
0.950
0.940
0.930
0.920
0.910
0.900
23.8 GHz
31.4 GHz
89 GHz
0
25
50
75 100 125 150 175 200 225 250 275 300 325 350
Day (2007 - 2006)
AMSU CH2 (31.4 GHz) Emissivities - C3VP
0.990
0.970
Emissivity
C3VP(ALL)
C3VP (ALL)
C3VP (No Rain)
C3VP (No Rain)
KARBOU
e31
e50
e89
0.839 0.866 0.888
0.058 0.049 0.042
0.821 0.842 0.847
0.057 0.051 0.057
0.950
0.930
KARBOU
0.910
PETERS-LIDARD
0.890
BOUKABARA
0.870
0.850
0
25
50
75 100 125 150 175 200 225 250 275 300 325 350
Day (2007 - 2006)
17th International TOVS Study Conference Pacific Grove, CA April 2010
17
Soil
Moisture
Land Sfc
Temperature
10V 10H
Emissivity
C3VP Site
44N 80W
1 Jun 2006‐1 Jul 2007
Red impulses indicate 3‐hourly CMORPH accumulated precipitation average
18V 18H
Emissivity
23V 23H
Emissivity
37V 37H
Emissivity
17th International TOVS Study Conference Pacific Grove, CA April 2010
18
Summary and Future
• The advancement of precipitation over land in GPM-era must
consider
– Accurate surface characterization 6 – 200 GHz under all weather
conditions
– Understanding of the sensitivity of the retrievals over all surface types
and frequency range
• PMM Science Team intercomparison study
– Preliminary results indicate similarities and differences between the
different methods
• Early hope that they will all be “about the same” is likely not true – we
have a lot of work to do!
• Future
– A lot more work to do - we need your help!
– How can ITWG work with IPWG and GPM communities?
• Let’s try to develop a joint emissivity model for 6 – 200 GHz under all
weather conditions
• Come to IPWG#5 - Hamburg, Germany, 11-15 October 2010
• Come to GPM algorithm team meeting – May 2010, College Park, MD
17th International TOVS Study Conference Pacific Grove, CA April 2010
19
Backup Slides
Further results not shown due to time constraints
Credits – Turk (JPL), Peters-Lidard (NASA), Boukabara
(NOAA), J. Wang (NASA), N. Wang (ESSIC)
17th International TOVS Study Conference Pacific Grove, CA April 2010
20
MIRS Results – AMSR-E
Results courtesy of S. Boukabara/W. Chen
•6.9V: relatively small interannual
Variability EXCEPT:
•Finland in winter
•C3VP in winter
•SGP due to vegetation/lack of it
•Wetlands in fall
•89V: highly variable for most sites
17th International TOVS Study Conference Pacific Grove, CA April 2010
21
MIRS Results – AMSU vs. AMSR-E
Results courtesy of S. Boukabara/W. Chen
•AMSR-E ~50 deg incidence angle, AMSU shown here between 40-60 deg.
•AMSU is also mixed polarization
•Only 5 targets of AMSU data were made available to date
•Similarities:
•Large annual cycle changes at C3VP and wetland (and magnitudes)
•Finland is clearer lower than other targets
•Differences:
TOVS Study Conference •Water targets (mixed polarization?) 17th International
Pacific Grove, CA April 2010
22
Amazon Comparison
July 2006 - Am azon (19 V)
1.000
0.970
0.960
AMSR-E (CICS)
0.950
TMI (CICS)
0.940
0.930
AMSR-E (MIRS)
0.920
0.910
20
8
20
6
20
2
19
9
19
7
Julian Day
July 2006 - Am azon (10 V)
1.000
0.990
0.980
0.970
0.960
0.950
AMSR-E (CICS)
0.940
AMSR-E (MIRS)
TMI (CICS)
0.930
0.920
0.910
20
8
20
6
20
2
19
9
19
7
19
3
19
1
0.900
18
4
Means:
AMSR-E (CICS)=0.934
TMI (CICS)=0.938
AMSR-E (MIRS)=0.946
Emissivity
19
3
19
1
0.900
18
4
Emissivity
0.990
0.980
Means:
AMSR-E (CICS)=0.942
TMI (CICS)=0.942
AMSR-E (MIRS)=0.952
Julian Day
17th International TOVS Study Conference Pacific Grove, CA April 2010
23
Clear Sky
Cloudy
CICS - Amazon Target
AMSR-E
Results courtesy of N-Y. Wang/K. Gopalan
TMI
10V
10V
1
1
0.8
0.6
240
0.8
245
250
255
260
265
270
0.6
240
275
18V
1
245
250
255
260
265
270
275
0.6
240
250
255
260
265
270
275
37V
surface
245
ε
surface
ε
265
270
275
245
250
255
260
265
270
275
260
265
270
275
260
265
270
275
260
265
270
275
21V
1
1
0.8
0.6
240
245
250
255
37V
1
1
0.8
0.8
245
250
255
260
265
270
275
89V
0.6
240
245
250
255
85V
1
1
0.8
0.6
240
260
1
0.8
0.6
240
255
0.8
23V
0.6
240
250
19V
0.8
0.6
240
245
0.8
245
250
255
260
Julian day, 2006
265
270
275
0.6
240
245
250
255
Julian day, 2006
•37 GHz or less
•Reasonable stability of Є over vegetated target
•Cloud affects minimal
•89 GHz
•Cloud and precipitation affects dramatic
•Similar values during clear conditions
17th International TOVS Study Conference Pacific Grove, CA April 2010
24
MIRS vs. CICS - Amazon
Results courtesy of S. Boukabara/W. Chen
10V
1
0.8
0.6
240
245
250
255
260
265
270
275
260
265
270
275
260
265
270
275
260
265
270
275
260
265
270
275
18V
1
0.8
0.6
240
245
250
255
ε
surface
23V
1
0.8
0.6
240
245
250
255
37V
1
0.8
0.6
240
245
250
255
89V
1
0.8
0.6
240
245
250
255
Julian day, 2006
•Very limited data to compare at this time….
•Seems like reasonable agreement in ranges of values – clear sky
17th International TOVS Study Conference Pacific Grove, CA April 2010
25
NASA – AMSU-B at C3VP
Results courtesy of J. Wang/G. Skofronick-Jackson
Є89
Є150
17th International TOVS Study Conference Pacific Grove, CA April 2010
26
C3VP Comparison
0.95
0.93
0.91
0.89
0.87
0.85
0.83
0.81
0.79
0.77
0.75
MIRS 92V
Means:
MIRS=0.913
GSFC=0.911
177
161
143
127
110
93
76
60
44
28
12
361
339
320
303
286
268
252
236
215
199
GSFC 92V
182
Emissivity
C3VP - MIRS vs. GSFC - SSMIS 92 V GHz
Julian Day
1
0.98
0.96
0.94
0.92
0.9
0.88
0.86
0.84
0.82
0.8
MIRS 150H
Means:
MIRS=0.883
GSFC=0.930
177
161
143
127
110
93
76
60
44
28
12
361
339
320
303
286
268
252
236
215
199
GSFC 150H
182
Emissivity
C3VP - MIRS vs. GSFC - SSMIS 150 H GHz
Julian Day
17th International TOVS Study Conference Pacific Grove, CA April 2010
27
C3VP Forward Modeling
AMSU-A 31.4 GHz
Slide courtesy of
Peters-Lidard, Harrison & Mocko
NASA/GSFC
LIS‐Catchment‐RTM Tb simulation
Std‐error (linear regression)
Std‐error (linear regression)‐‐no snow
Std‐error (linear regression)‐‐snow
Correlation
Correlation‐‐no snow
Correlation‐‐snow
26-29 October 2009
6.00K
6.75K
6.69K
LIS‐Noah‐CRTM Tb simulation
Std‐error (linear regression)
Std‐error (linear regression)‐‐no snow
Std‐error (linear regression)‐‐snow
0.86
0.86
0.38
Correlation
Correlation‐‐no snow
Correlation‐‐snow
5.68K
6.49K
5.68K
0.87
0.86
‐0.70
17th International TOVS Study Conference -
2009 PMM SciencePacific
TeamGrove,
Meeting
Salt 2010
Lake City, UT
CA - April
28
C3VP Forward Modeling
AMSU-A 31.4 GHz
LIS-Noah-CRTM Time Series
Slide courtesy of
Peters-Lidard, Harrison & Mocko
NASA/GSFC
26-29 October 2009
17th International TOVS Study Conference -
2009 PMM SciencePacific
TeamGrove,
Meeting
Salt 2010
Lake City, UT
CA - April
29
C3VP Forward Modeling
AMSU-B 89 GHz
Slide courtesy of
Peters-Lidard, Harrison & Mocko
NASA/GSFC
LIS‐Catchment‐CRTM Tb simulation
Std‐error (linear regression)
Std‐error (linear regression)‐‐no snow
Std‐error (linear regression)‐‐snow
Correlation
Correlation‐‐no snow
Correlation‐‐snow
26-29 October 2009
8.16K
2.58K
13.58K
0.82
0.92
0.75
LIS‐Noah‐CRTM Tb simulation
Std‐error (linear regression)
7.02K
Std‐error (linear regression)‐‐no snow 3.15K
Std‐error (linear regression)‐‐snow
10.73K
Correlation
Correlation‐‐no snow
Correlation‐‐snow
0.87
0.94
0.80
17th International TOVS Study Conference -
2009 PMM SciencePacific
TeamGrove,
Meeting
Salt 2010
Lake City, UT
CA - April
30
C3VP Forward Modeling
AMSU-B 89 GHz
LIS-Noah-CRTM Time Series
Slide courtesy of
Peters-Lidard, Harrison & Mocko
NASA/GSFC
17th International TOVS Study Conference Pacific Grove, CA April 2010
31