Title: Microwave Integrated Retrieval System MIRS Performances Summary
1Microwave Integrated Retrieval System (MIRS)
Performances Summary
June 15th 2008
- S.-A. Boukabara, K. Garrett, C. Kongoli, W. Chen,
F. Iturbide-Sanchez
2Context
- MIRS Fourth Release scheduled June 15th 2008
- Several updates introduced in the MIRS algorithm
- Extension of retrieval in precipitating
conditions - Extension to multiple sensors (NOAA-18, Metop-A
and DMSP SSMIS) - Covariance Matrix (cross-correlations added)
- Bias Fine Tuning
- Regression-based First Guess
- QC
- Extension to other products (T,Q, Sfc Properties,
cloud/rain parameters, etc) - Need to keep track of performances
- Some depend on season
- Some depend on vertical layer
- Some depend on ground truth used
- Some are only qualitative (lack of ground truth)
- Credit due to the Integrated Collocation Data
Base (ICDB), provided by Tony Reale and Co
3List of products (Official)
- Metop-A and NOAA-18
- Temperature profile (ocean)
- Moisture (ocean and non-costal land)
- Total Precipitable Water (TPW) (ocean and
non-costal land) - Land Surface Temperature (LST)
- Emissivity Spectrum (All surfaces)
- Surface Type (sea, land, snow, sea-ice)
- Emissivity-based Snow Water Equivalent (SWE)
- Emissivity-based Snow Cover Extent (SCE)
- Emissivity-based Sea Ice Concentration (SIC)
- Vertically Integrated Non-precipitating Cloud
Liquid Water (CLW) - Vertically Integrated Ice Water Path (IWP)
- Vertically Integrated Rain Water Path (RWP)
- DMSP F16 SSMIS
- Temperature profile (ocean)
- Moisture (ocean and non-costal land)
- Total Precipitable Water (TPW) (ocean and
non-costal land) - Land Surface Temperature (LST)
Note The hydrometeor profiles dropped from
official list (lack of information content in
radiances, see next slide)
Total 30 products
4List of unofficial products (Delivered For
Testing purposes)
Note Cloud profile made available for testing
purposes (see next slide)
- Metop-A and NOAA-18
- Cloud Liquid Water Profile (CLWP) over ocean
- Surface Temperature (skin) of snow-covered land
- Sea Surface Temperature (SST)
- Effective grain size of snow (over snow-covered
land surface) - Multi-Year (MY) Type Sea Ice concentration
- First-Year (FY) Type Sea Ice Concentration
- DMSP F16 SSMIS
- Extended Total Precipitable Water over
non-coastal Land - Emissivity-based Snow Water Equivalent (SWE)
- Emissivity-based Snow Cover Extent (SCE)
- Emissivity-based Sea Ice Concentration (SIC)
- Surface Temperature (skin) of snow-covered land
- Sea Surface Temperature (SST)
- Effective grain size of snow (over snow-covered
land surface) - Multi-Year (MY) Type Sea Ice concentration
- First-Year (FY) Type Sea Ice Concentration
Total 21 test products
5Important Note
- This summary is a snapshot of the performances.
- For more details, look at the STAR MIRS web site
mirs.nesdis.noaa.gov
6Assessing performances
- Go through all products performances, for all
sensors (30 official products and 21 test
products 51 total from 3 sensors) - Performances assessments depend on product
sometimes qualitative. - Keep in mind these performances are impacted by
errors in ground truth data themselves, by
time/space collocation, by representativeness
errors, etc. - Intra-truth variability assessed by comparing to
multiple sources - These performances are assessed in cloudy
conditions so greater spatial coverage of
applicability of these assessments
7Note about Hydrometeor Vertical Distribution
(Atmospheric Profile) -1st Case-
- Profiles of Hydrometeors retrieved by MIRS Used
to get integrated values (CLW, IWP, RWP) - Vertical distribution sometimes not very
accurate due to lack of information content in
Radiances. See examples - Sometimes vertical distribution is accurate (need
assessment by users if it fits requirements - Suggested to make available for testing purposes
-
Convergence reached (in two iterations)
Temp
Iter0
Iter2
Iter1
Iter0
Iter1
Iter2
WV
Iter0
Iter1
Iter2
Truth
Background
Cloud
Retrieval
8Note about Hydrometeor Vertical Distribution
(Atmospheric Profile) -2nd case-
A case where convergence was reached but the
vertical distribution of retrieved is
significantly different from truth
9NOAA-18 AMSU/MHS
10Temperature Profile (1/4)(over open water ocean,
against GDAS)
MIRS
GDAS
MIRS GDAS Diff
- The temperature is officially delivered over
ocean only. But over non-ocean (land, snow, sea
ice), temperature is still valid. - Validation is performed by comparing to
- GDAS
- ECMWF
- RAOB
MIRS GDAS Diff
N18
11Temperature Profile (2/4)(over open water ocean,
against ECMWF)
Angle dependence taken care of very well, without
any limb correction
MIRS
ECMWF
Note Retrieval is done over all surface
backgrounds but also in all weather conditions
(clear, cloudy, rainy, ice)
MIRS ECMWF Diff
MIRS ECMWF Diff
N18
12Temperature Profile (3/4)(over open water ocean,
against RAOBs, COSMIC, ATOVS, Forecast)
Stdev Ocean
Bias Ocean
Original 100 layers resolution
Bias of roughly 1 K noticed at the surface
Stdev Ocean
Vertical Averaging (IORD-II requirements (moving
window of 1, 1.5 and 5 km)
Collocation criteria (COSMIC, ATOVS, SSMIS,
RAOB) /- 5 hours, /- 100 Kms Data spanning 42
days
N18
13Temperature Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for temperature in
cloudy - Uncertainty (surface to 700 mb 2.5K per 1km
layer, 700 mb to 300 mb 1.5K per 1 km layer, 300
to 30 mb 1.5K per 3km layer, 30 to 1mb 1.5K per
5km layer)
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
N18
14Moisture Profile (1/4)(over open water and land,
against GDAS)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
MIRS
GDAS
land
Bias
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
Sea
Stdev
N18
15Moisture Profile (2/4)(over open water and land,
against ECMWF)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
MIRS
ECMWF
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
land
Bias
Sea
Stdev
When assessing, keep in mind all ground truths
(wrt GDAS, ECMWF, RAOB)
N18
16Moisture Profile (3/4)(over open water and land,
against RAOB, COSMIC, Forecast, ATOVS)
N18
Stdev is found very good over land and ocean
Ocean Stdev
Ocean Bias
MIRS is compared to Raob (along with COSMIC,
ATOVS and Forecast)
Land Stdev
Land Bias
Bias wrt RAOB (over land) not consistent with
ECMWF and GDAS
17Moisture Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for Water Vapor
Mixing Ratio (in g/Kg), for cloudy - Uncertainty (surface to 600 mb greater of 20 or
0.2 g/Kg, 600 mb to 100 mb greater of 40 or 0.1
g/Kg) - expressed as percent error of average mixing
ratio in 2km layers - - No measurement precision
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
N18
18Total Precipitable Water (TPW) (1/3) (over open
water and land)
ECMWF
MIRS extends TPW over other surface types (for
the first time for an operational algorithm)
MIRS
- Validation done by comparison to
- GDAS
- MSPPS
- ECMWF
- Radiosondes
GDAS
MSPPS
N18
19Total Precipitable Water (TPW) (2/3) (over open
water and land)
Global
Global
Global
MIRS vs RAOB
MIRS vs GDAS
MIRS vs ECMWF
Ocean
Ocean
Ocean
MIRS vs RAOB
MIRS vs GDAS
MIRS vs ECMWF
N18
20Total Precipitable Water (TPW) (3/3) (over open
water and land)
IORD-II Requirements - Accuracy greater of 2
mm or 10 - Precision 1mm
N18
21Land Surface Temperature (1/3)Comparison against
GDAS
Ascending, MIRS
Ascending, GDAS
Bias 0.08K Std 6.9K
Bias -1.4K Std 5.7K
N18
Descending, MIRS
Descending, GDAS
22Land Surface Temperature (2/3)Comparison against
ECMWF
Ascending, MIRS
Ascending, ECMWF
Bias -1.5K Std 5.5 K
Bias -0.7K Std 5.8K
N18
Descending, MIRS
Descending, ECMWF
23Land Surface Temperature (3/3)Comparison against
Other Sources and Summary
N18
24Emissivity Spectrum (1/3)(All surfaces)
- Validation (qualitative and quantitative) of
emissivity done by comparing to - MSPPS emissivity (23.8, 31.4 and 50.3 GHz
channels) not available for all channels - Analytical emissivities (over land, sea ice and
snow) - Ocean emissivity model applied to GDAS wind
fields - Ocean emissivity model applied to ECMWF wind
fields - Analytical emissivities (all sfcs) using RAOB
profiles
Heavily dependent on validity of raob models
fields (wind speed, skin temperature, water vapor
and temperature profiles)
N18
Analytical emissivity not valid when specular
assumption not applicable (high topography)
25Emissivity Spectrum (2/3)(All surfaces)
MSPPS
MIRS
MIRS vs ECMWF
ECMWF
MIRS vs MSPPS
N18
26Emissivity Spectrum (3/3)(All surfaces)
31.4 GHz
50.3 GHz
23.8 GHz
No IORD-II requirement for emissivity
Large uncertainty over snow (bias large) when
comparing to both GDAS and ECMWF analytical
emissivity. But large uncertainties exist for the
GDAS and ECMWF Tskin over snow, which served to
determine the analytical emissivity
N18
27Surface Type Classification
Types detected ocean, sea ice, land and snow
TB-based Pre-classification
Emissivity-based Post-classification
- Both made available to users
- When two coincide, high confidence
- When two differ, ambiguity possible
N18
28Snow Water Equivalent (SWE)(Emissivity-based)
MSPPS
MIRS
N18
29Example of MIRS, MSPPS, AMSR-E and IMS snow
cover retrievals. MIRS overall shows more
consistency with IMS. Noted is Central Asia,
with MSPPS and AMSR-E showing more false alarm
than MIRS
Snow Cover Extent (SCE) (1/2)(Emissivity-based)
MIRS Snow Cover
IMS Snow Cover
AMSR-E Snow Cover
MSPPS Snow Cover
30Snow Cover Extent (SCE) (2/2)(Emissivity-based)
N18
31Example of MIRS, MSPPS and IMS Sea Ice Conc.
Retrievals over the Northern Hemisphere MIRS
shows more consistency with IMS
Sea Ice Concentration (SIC) (1/4)(Emissivity-base
d, Comparison to IMS, MSPPS)
MIRS SIC on April 2, 2007
MSPPS SIC on April 2, 2007
IMS SIC on April 2, 2007
N18
32Example of on-line MIRS and MSPPS Sea Ice Conc.
Inter-comparisons and Stats MIRS shows a
higher mean sea ice concentration than MSPPS,
consistent with statistical results with respect
to IMS
Sea Ice Concentration (SIC) (2/4)(Emissivity-base
d, Comparison to MSPPS)
N18
33Sea Ice Concentration (SIC) (3/4)(Emissivity-base
d, Comparison to IMS)
Current version MIRS N18 AMSU-MHS Sea Ice
Concentration Performance Assessment over
Northern Hemisphere 4-km IMS as Reference
N18
34Current and DAP MIRS Sea Ice Concentration
Performance over high-concentration areas (100
ice all-year around) case study
-------------------------------------------------
--------------------------------------------Latit
ude 84-86 N Longitude 135-137 W
Year 2006
Sea Ice Concentration (SIC) (4/4)(Emissivity-base
d, Assessment of SIC in 100 ice-covered sea
areas)
35Cloud Liquid Water (CLW) (1/4)(Non-Precipitating,
in simulation)
Mean 0.006 mm Stdev 0.043 mm CorrFact
0.991
Experiment set up
ECMWF Data (cloudy)
MIRS Retrievals (CLW)
Performances of cloud retrieval are very good in
simulation
CRTM
MIRS
When comparing to real data, most errors are
believed due to assumptions in other cloud
parameters (vertical distribution, density, size
distribution, etc) as well as due to collocation
errors (time, space and representativeness)
Synthetic TBs (cloudy)
N18
36Cloud Liquid Water (CLW) (2/4)(Non-Precipitating,
over open water ocean)
- Based on the Cloud Liquid Water Profile retrieval
(just like TPW is based on the Q profile) - Much higher correlation between MIRS and MSPPS
noticed in descending than in descending nodes
Note MSPPS uses Ts and wind speed from forecast
data as inputs - MIRS CLW is for non-precipitating clouds. MSPPS
is for all clouds. - MSPPS is the reference is microwave cloud amounts
- MIRS relies on physical constraints to retrieve
cloud, rain and ice Jacobians and
forward-simulated radiances based on Community
Radiative Transfer Model (CRTM)
Descending
Ascending
MIRS vs MSPPS
MIRS vs MSPPS
Corr Factor 0.63
Corr Factor 0.47
N18
37Cloud Liquid Water (CLW) (3/4)(Non-Precipitating,
over open water ocean)
MSPPS (official)
MIRS
Note the different scales MIRS 0-0.3 MSPPS
0-0.7 mm
Higher small amounts of clouds noticed from MIRS
(as opposed to MSPPS), due to use of high
frequencies?
38Cloud Liquid Water (CLW) (4/4)(Comparison to
ECMWF analysis field of clouds)
Wrt-MSPPS uncertainty (best estimate) 0.15 mm
IORD-II Requirements Uncertainty of 0.25 mm
over ocean
Histograms comparisons (MIRS vs ECMWF)
Corr Factor 0.50
MIRS vs ECMWF
39Ice Water Path (IWP)(over open water ocean and
land)
- Based on the Frozen Water Profile retrieval (just
like TPW is based on the Q profile) - The Ice Water Path is retrieved globally (over
ocean and land)
MSPPS
MIRS
Note the different scales
N18
40Ice Water Path (IWP)(over open water ocean and
land)
Ascending
Descending
MIRS vs MSPPS
MIRS vs MSPPS
Noticed Systematic bias of 0.1 mm between MSPPS
and MIRS (MSPPS higher)
IORD-II requirement Precision greater of 0.05
mm or 10, Accuracy greater of 0.1 mm or 25
Important Note(s) 1) There is a large
uncertainty on the actual amounts of cloud, ice,
liquid. MIRS is using physical constraints
(fitting the radiances with a forward model that
has Scattering included, use of Jacobians wrt
ice, liquid, as well from CRTM, etc) 2)
Measurement information is dependent on (a)
number of particles, (b) their sizes, and (c)
their densities IORD-II text. We rely on CRTM
for the assumptions made regarding number, size,
density of ice particles. 3) MIRS retrievals are
at AMSU footprint resolution while MSPPS are at
MHS resolution. Due to this, it is expected that
the MIRS retrievals are smoothed lower.
Retrievals is relative to amounts within
footprint.
41Ice Water Path (IWP)(Impact of spatial
resolution)
MIRS IWP Retrieval _at_ AMSU resolution
Field of ice gets smoothed out because of
footprint size
MIRS IWP Retrieval _at_ MHS resolution
Case of Katrina (very heavy precipitation) -August
2005-
42Rain Water Path (RWP) 1/2(over open water ocean
and land)
- Based on the Liquid (Rain) Water Profile
retrieval (just like TPW is based on the Q
profile) - RWP is retrieved globally
- Qualitative comparison with MSPPS Rainfall Rate
(RR)
MSPPS Rainfall rate
MIRS Rain Water Path
N18
Some areas are missing Retrieved as cloud in
MIRS (see next slide)
43Rain Water Path (RWP) 2/2(over open water ocean
and land)
MIRS Liquid Water Path (Precipnonprecip)
MSPPS Rainfall rate
MIRS Rain Water Path
N18
44METOP-A AMSU/MHS
45Temperature Profile (1/4)(over open water ocean
and other, against GDAS)
MIRS
GDAS
MIRS GDAS Diff
- The temperature is officially delivered over
ocean only. But over non-ocean (land, snow, sea
ice), temperature is still valid. - Validation is performed by comparing to
- GDAS
- ECMWF
- RAOB
MIRS GDAS Diff
METOP-A
46Temperature Profile (2/4)(over open water ocean)
Angle dependence taken care of very well, without
any limb correction
MIRS
ECMWF
Note Retrieval is done over all surface
backgrounds but also in all weather conditions
(clear, cloudy, rainy, ice)
MIRS ECMWF Diff
MIRS ECMWF Diff
METOP-A
47Temperature Profile (3/4)(over open water ocean,
against RAOBs, COSMIC, ATOVS, Forecast)
Bias Ocean
Stdev Ocean
Original 100 layers resolution
Stdev Ocean
Bias of roughly 1 K noticed at the surface
Vertical Averaging (IORD-II requirements (moving
window of 1, 1.5 and 5 km)
Collocation criteria (COSMIC, ATOVS, SSMIS,
RAOB) /- 5 hours, /- 100 Kms Data spanning 42
days
METOP-A
48Temperature Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for temperature in
cloudy - Uncertainty (surface to 700 mb 2.5K per 1km
layer, 700 mb to 300 mb 1.5K per 1 km layer, 300
to 30 mb 1.5K per 3km layer, 30 to 1mb 1.5K per
5km layer)
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
METOP-A
49Moisture Profile (1/4)(over open water and land,
against GDAS)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
MIRS
GDAS
Bias
land
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
Sea
Stdev
METOP-A
50Moisture Profile (2/4)(over open water and land,
against ECMWF)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
MIRS
ECMWF
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
land
Bias
Sea
Stdev
When assessing, keep in mind all ground truths
(wrt GDAS, ECMWF, RAOB)
METOP-A
51Moisture Profile (3/4)(over open water and land,
against RAOB, COSMIC, Forecast, ATOVS)
METOP-A
Stdev is found very good over land and ocean
(even better than N18 and COSMIC)
Ocean Stdev
Ocean Bias
MIRS is compared to Raob (along with COSMIC,
ATOVS and Forecast)
Land Bias
Land Stdev
Bias wrt RAOB (over land) not consistent with
ECMWF and GDAS
52Moisture Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for Water Vapor
Mixing Ratio (in g/Kg), for cloudy - Uncertainty (surface to 600 mb greater of 20 or
0.2 g/Kg, 600 mb to 100 mb greater of 40 or 0.1
g/Kg) - expressed as percent error of average mixing
ratio in 2km layers - - No measurement precision
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
METOP-A
53Total Precipitable Water (TPW) (1/2)(over open
water and land)
MIRS extends TPW over other surface types (for
the first time for an operational algorithm)
MIRS/METOP
GDAS
- Validation done by comparison to
- GDAS
- MSPPS
- ECMWF
- Radiosondes
MSPPS
RAOB/OCEAN
RAOB/Land
METOP-A
54Total Precipitable Water (TPW) (2/2) (over open
water and land)
IORD-II Requirements - Accuracy greater of 2
mm or 10 - Precision 1mm
METOP-A
55Land Surface Temperature (1/3)Comparison against
GDAS
Ascending, GDAS
Ascending, MIRS
Bias -0.8K Std 5.3K
Bias -0.6K Std 5.3K
Descending, GDAS
Descending, MIRS
METOP-A
56Land Surface Temperature (2/3)Comparison against
ECMWF
Ascending, MIRS
Ascending, ECMWF
Bias -1.5K Std 5.9 K
Descending, ECMWF
Bias -2.5K Std 5.3K
Descending, MIRS
METOP-A
57Land Surface Temperature (3/3)Comparison against
Other Sources and Summary
Larger bias noticed with HIRS than when
comparing to ECMWF/GDAS
METOP-A
58Emissivity Spectrum (1/3)(All surfaces)
- Validation (qualitative and quantitative) of
emissivity done by comparing to - MSPPS emissivity (23.8, 31.4 and 50.3 GHz
channels) not available for all channels - Analytical emissivities (over land, sea ice and
snow) - Ocean emissivity model applied to GDAS wind
fields - Ocean emissivity model applied to ECMWF wind
fields - Analytical emissivities (all sfcs) using RAOB
profiles
Heavily dependent on validity of raob models
fields (wind speed, skin temperature, water vapor
and temperature profiles)
Analytical emissivity not valid when specular
assumption not applicable (high topography)
METOP-A
59Emissivity Spectrum (2/3)(All surfaces)
MSPPS
MIRS
MIRS vs ECMWF
ECMWF
MIRS vs MSPPS
METOP-A
60Emissivity Spectrum (3/3)(All surfaces)
31.4 GHz
50.3 GHz
23.8 GHz
No IORD-II requirement for emissivity
Large uncertainty over snow (bias large) when
comparing to ECMWF analytical emissivity. But
large uncertainties exist for the GDAS and ECMWF
Tskin over snow, which served to determine the
analytical emissivity
METOP-A
61Surface Type Classification
Types detected ocean, sea ice, land and snow
- Both made available to users
- When two coincide, high confidence
- When two differ, ambiguity possible
METOP-A
62Snow Water Equivalent (SWE)(Emissivity-based)
METOP-A
63Example of MIRS, MSPPS, AMSR-E and IMS snow
cover retrievals. MIRS overall shows more
consistency with IMS. Noted is Central Asia,
with MSPPS and AMSR-E showing more false alarm
than MIRS
Snow Cover Extent (SCE)(Emissivity-based)
IMS Snow Cover
MIRS Snow Cover
AMSR-E Snow Cover
MSPPS Snow Cover
METOP-A
64Snow Cover Extent (SCE)(Emissivity-based)
Statistical Assessment of current MIRS METOP-A
AMSU-MHS Snow coverover North America in 2007
4-km IMS As Reference
METOP-A
65Example of on-line METOP-A MIRS and MSPPS Sea Ice
Conc. Inter-comparisons MIRS shows higher sea
ice concentrations, consistent with statistical
results with respect to IMS(see next slide)
Sea Ice Concentration (SIC) (1/2)(Emissivity-base
d)
METOP-A
66Sea Ice Concentration (SIC) (2/2)(Performances)
Current version Metop-A AMSU-MHS Sea Ice
Concentration Performance Assessment over
Northern Hemisphere collocated 4-km IMS as
Reference - 2007 ---------------------------------
--------------------------------------------------
-------------
Mean Sea Ice Concentrations in
METOP-A
67Cloud Liquid Water (CLW) (1/2)(Non-Precipitating,
over open water ocean)
- Based on the Cloud Liquid Water Profile retrieval
(just like TPW is based on the Q profile) - Much higher correlation between MIRS and MSPPS
noticed in ascending than in descending nodes
Note MSPPS uses Ts and wind speed from forecast
data as inputs - MIRS CLW is for non-precipitating clouds. MSPPS
is for all clouds. - MSPPS is the reference is microwave cloud amounts
- MIRS relies on physical constraints to retrieve
cloud, rain and ice Jacobians and
forward-simulated radiances based on Community
Radiative Transfer Model (CRTM)
Wrt-MSPPS uncertainty 0.15 mm
Descending
Ascending
MIRS vs MSPPS
MIRS vs MSPPS
Corr Factor 0.64
Corr Factor 0.69
METOP-A
68Cloud Liquid Water (CLW) (2/2)(Comparison to
ECMWF analysis field of clouds)
Histograms comparisons (MIRS vs ECMWF)
Corr Factor 0.46
MIRS vs ECMWF
METOP-A
69Ice Water Path (IWP)(over open water ocean and
land)
- Based on the Frozen Water Profile retrieval (just
like TPW is based on the Q profile) - The Ice Water Path is retrieved globally (over
ocean and land)
MSPPS
MIRS
Note the different scales
METOP-A
70Ice Water Path (IWP)(over open water ocean and
land)
Ascending
Descending
MIRS vs MSPPS
MIRS vs MSPPS
Noticed Systematic bias of 0.1 mm between MSPPS
and MIRS (MSPPS higher)
IORD-II requirement Precision greater of 0.05
mm or 10, Accuracy greater of 0.1 mm or 25
Important Note(s) 1) There is a large
uncertainty on the actual amounts of cloud, ice,
liquid. MIRS is using physical constraints
(fitting the radiances with a forward model that
has Scattering included, use of Jacobians wrt
ice, liquid, as well from CRTM, etc) 2)
Measurement information is dependent on (a)
number of particles, (b) their sizes, and (c)
their densities IORD-II text. We rely on CRTM
for the assumptions made regarding number, size,
density of ice particles. 3) MIRS retrievals are
at AMSU footprint resolution while MSPPS are at
MHS resolution. Due to this, it is expected that
the MIRS retrievals are smoothed lower.
Retrievals is relative to amounts within
footprint.
METOP-A
71Rain Water Path (RWP) 1/2(over open water ocean
and land)
- Based on the Liquid (Rain) Water Profile
retrieval (just like TPW is based on the Q
profile) - RWP is retrieved globally
- Qualitative comparison with MSPPS Rainfall Rate
(RR)
MSPPS Rainfall rate
MIRS Rain Water Path
Some areas are missing Retrieved as cloud in
MIRS (see next slide)
METOP-A
72Rain Water Path (RWP) 2/2(over open water ocean
and land)
MIRS Liquid Water Path (Precipnonprecip)
Thanks to CRTM, MIRS is able to distinguish
precipitating from non-precipitating clouds
liquid water
MSPPS Rainfall rate
MIRS Rain Water Path
METOP-A
73DMSP-F16 SSMI/S
74Temperature Profile (1/4)(over open water ocean,
against GDAS)
MIRS
GDAS
MIRS GDAS Diff
- The temperature is officially delivered over
ocean only. But over non-ocean (land, snow, sea
ice), temperature is still valid. - Validation is performed by comparing to
- GDAS
- ECMWF
- RAOB
MIRS GDAS Diff
DMSP SSMIS
75Temperature Profile (2/4)(over open water ocean)
Angle dependence taken care of very well, without
any limb correction
MIRS
ECMWF
Note Retrieval is done over all surface
backgrounds but also in all weather conditions
(clear, cloudy, rainy, ice)
MIRS ECMWF Diff
MIRS ECMWF Diff
DMSP SSMIS
76Temperature Profile (3/4)(over open water ocean,
against RAOBs, COSMIC, ATOVS, Forecast)
Bias Ocean
Stdev Ocean
Original 100 layers resolution
Stdev Ocean
Collocation criteria (COSMIC, ATOVS, SSMIS,
RAOB) /- 5 hours, /- 100 Kms Data spanning 42
days
Vertical Averaging (IORD-II requirements (moving
window of 1, 1.5 and 5 km)
DMSP SSMIS
77Temperature Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for temperature in
cloudy - Uncertainty (surface to 700 mb 2.5K per 1km
layer, 700 mb to 300 mb 1.5K per 1 km layer, 300
to 30 mb 1.5K per 3km layer, 30 to 1mb 1.5K per
5km layer)
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
DMSP SSMIS
78Moisture Profile (1/4)(over open water, against
GDAS)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
MIRS
GDAS
Bias
land
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
Sea
Stdev
DMSP SSMIS
79Moisture Profile (2/4)(over open water, against
ECMWF)
- Validation of WV done by comparing to
- GDAS
- ECMWF
- RAOB
- Retrieval done over all surfaces in all weather
conditions
ECMWF
MIRS
- Assessment includes
- Angle dependence
- Statistics profiles
- Difference maps
Bias
land
Sea
Stdev
DMSP SSMIS
80Moisture Profile (3/4)(over open water and land,
against RAOB, COSMIC, Forecast)
DMSP SSMIS
Perfs of SSMIS sounding not bad, given
calibration anomalies
Ocean Stdev
Ocean Bias
MIRS is compared to Raob (along with COSMIC, and
Forecast)
Land Stdev
Land Bias
Bias wrt RAOB (over ocean and land) not
consistent with ECMWF and GDAS
81Moisture Profile (4/4)(Performances)
Ocean
Land
- Note IORD-II requirements for Water Vapor
Mixing Ratio (in g/Kg), for cloudy - Uncertainty (surface to 600 mb greater of 20 or
0.2 g/Kg, 600 mb to 100 mb greater of 40 or 0.1
g/Kg) - expressed as percent error of average mixing
ratio in 2km layers - - No measurement precision
These requirements are for CrIS and ATMS, which
have more channels and higher sensing skills in
general than AMSU, MHS or SSMIS
DMSP SSMIS
82Total Precipitable Water (TPW) (1/2)(over open
water)
TPW over other surface types available (but not
officially operational in this phase)
MIRS Vs RAOB
GDAS
MIRS
MIRS Vs ECMWF
MSPPS
ECMWF
MIRS Vs GDAS
- Validation done by comparison to
- GDAS, MSPPS, ECMWF, Radiosondes
DMSP SSMIS
83Total Precipitable Water (TPW) (2/2) (over open
water)
IORD-II Requirements - Accuracy greater of 2
mm or 10 - Precision 1mm
DMSP SSMIS
84Land Surface Temperature (1/4)Comparison against
GDAS
Ascending, GDAS
Ascending, MIRS
Bias 1.8K Std 5.3K
Bias 1.3K Std 4.7K
Descending, GDAS
Descending, MIRS
DMSP SSMIS
85Land Surface Temperature (2/4)Comparison against
ECMWF
Ascending, MIRS
Ascending, ECMWF
Bias 1.3K Std 5.4 K
Descending, ECMWF
Bias 0.5K Std 4.4K
Descending, MIRS
DMSP SSMIS
86Land Surface Temperature (3/4)Comparison against
GOES
87Land Surface Temperature (3/3)Comparison against
Other Sources (HIRS) Summary
Larger bias noticed with HIRS than when
comparing to ECMWF/GDAS
Much more consistent performances When comparing
to GOES Tskin.
DMSP SSMIS
88Emissivity Spectrum (1/3)(All surfaces)
MIRS
GDAS
- Validation (qualitative and quantitative) of
emissivity done by comparing to - Analytical emissivities (over land, sea ice and
snow) - Ocean emissivity model applied to GDAS wind
fields - Ocean emissivity model applied to ECMWF wind
fields - Analytical emissivities (all sfcs) using RAOB
profiles
Heavily dependent on validity of raob models
fields (wind speed, skin temperature, water vapor
and temperature profiles)
Analytical emissivity not valid when specular
assumption not applicable (high topography)
SSMIS
89Emissivity Spectrum (2/3)(All surfaces)
MIRS
GDAS
MIRS vs ECMWF
ECMWF
ECMWF
MIRS vs GDAS
DMSP SSMIS
90Emissivity Spectrum (3/3)(All surfaces)
19 V
19H
37 V
No IORD-II requirement for emissivity
DMSP SSMIS
91Surface Type Classification
MIRS/METOP-A
MIRS/N18
MIRS/SSMIS
DMSP SSMIS
92Conclusions Talking Points
93Discussion (1/4)
- Performances were assessed using different
sources. Sometimes results are different,
reflecting inter-truth variability. - When consistent behavior is noticed, assumed that
MIRS is the likely reason - SSMIS is found, as expected from radiances
noisiness, to have slightly more degraded
performances (EDRS TPW) - N18 and Metop-A running at AMSUA resolution
- SSMIS running at UAS resolution
- Both resolutions could be augmented (computer
time cost)
94Discussion (2/4)
- TPW is extended to all surfaces Ocean, Land, Sea
ice and Snow operationally for NOAA-18 and
Metop-A, for the first time. - Also for first time, retrieval is performed (and
convergence reached) in cloudy, rainy,
ice-impacted scenes - False alarm / Undetection possible with the
surface type product (always the case with this
type of products). We make available two products
(TB-based and EM-based) for a higher confidence. - For SSMIS, non-convergence noticed over coastal
areas (70 kms footprint more likely to be mixed
and therefore more difficult to fit TBs) - Products from Metop-A, NOAA-18 and DMSP are not
exactly identical everywhere but are
statistically consistent.
95Discussion (3/4)
- Very small bias found for N18 and Metop in Tskin
retrieval when compared to GDAS and ECMWF. Larger
bias found when comparing to HIRS (from ICDB).
Inconsistency. - MIRS seems to pick smaller amounts of cloud
liquid water, likely due to the use of high
frequency channels in MIRS. - Systematic Bias in IWP wrt MSPPS found (0.1 mm)
but with a high correlation (0.68)
96Discussion (4/4)
- Convergence is reached everywhere all surfaces,
all weather conditions including precipitating,
icy conditions - This is a major achievement a radiometric
solution is found even when precip/ice present.
With CRTM physical constraints.
Previous version (non convergence when precip/ice
present)
Current version
97Whats Next
- Continue validation of cloud products using
CLOUDSAT/TRMM data - Code checking using Valgrind and ForCheck
(potential code changes) - Potential improvement for T(p)
98For Next Phase
- Extension of products to include rain rate and
SSMIS emissivity-based sfc properties as well as
cloud/hydrometeors properties - Start extension of sensors to include NPP/ATMS
(for all products)