Assimilation of surface sensitive infrared
channels over land at Environment Canada
L. Garand, S.K. Dutta, and S. Heilliette
DAOS Meeting
Montreal, Qc
August 15-16, 2014
Outline
• Context & motivation
• Approach
• Assimilation results
________________________
• Also: Update on PCW mission
DRAFT – Page 2 – July 31, 2017
Context & motivation
Env. Can. moving to ensemble-variational system with:
•Flow-dependent background errors including surface skin
temperature correlations with other variables
•Analysis grid at 50 km, increments interpolated to model
25 km grid (15 km in 2015)
•~140 AIRS and IASI channels assimilated: many sensitive
to low level T, q, and Ts. RTM is RTTOV-10.
Favorable context to attempt assimilating surface-sensitive IR
channels over land and sea-ice vs earlier work with GOES (Garand et
al. JAM, 2004) with analysis grid at 150 km and no hyperspectral IR.
DRAFT – Page 3 – July 31, 2017
Ensemble spread of Ts (Feb-Mar 2011)
00 UTC
06 UTC
Maximum ~15h local
In SH (summer)
Maximum at night
Tibetan area (winter)
12 UTC
18 UTC
Ts background error
over land in 4Dvar is
Constant: 3 K.
In EnVar, B is
0.5(BENKF+ BNMC)
DRAFT – Page 4 – July 31, 2017
Ensemble T(~964 hpa) spread
(Jan-Feb 2011)
00 UTC
06 UTC
Only minor
variations with
local time
Values < 0.5 K in SH
seem underestimated
12 UTC
18 UTC
DRAFT – Page 5 – July 31, 2017
Key factors to consider
•
•
•
•
•
Reliable cloud mask
Spectral emissivity definition (CERES, U-Wisconsin)
Highly variable topography
Radiance bias correction
Background and observation error definition
DRAFT – Page 6 – July 31, 2017
Approach guided by prudence
Assimilate under these restrictive conditions over land and sea ice:
•Estimate of cloud fraction < 0.01
•High surface emissivity (> 0.97)
•Relatively flat terrain (local height STD < 100 m)
•Diff between background Ts and retrieval
based on inverting RTE limited to 4K
Radiance bias correction approach:
•For channels flagged as being surface sensitive, use only ocean
data to update bias coefficients
DRAFT – Page 7 – July 31, 2017
Assigned observation error to radiances
Assigned = f std(O-P)
15.0mm
f typically in range 1.6-2.0
Desroziers = <(O-P)(O-A)>
4.13mm
15.3mm
DRAFT – Page 8 – July 31, 2017
4.50mm
Limitation linked to topography
Criterion used: local STD of topography < 50 m (on 3X3 ~50 km areas)
RED: accepted, white std > 100 m, blue 100>std>50 m
DRAFT – Page 9 – July 31, 2017
Limitation linked to surface emissivity
Surface emissivity
AIRS ch 787
Emissivity based on
CERES land types
(~15 km)
+ weighted average
for water/ice/snow
fraction.
Accept only emissivity > 0.97 ; Bare soil, open shrub regions excluded.
DRAFT – Page 10 – July 31, 2017
Examples of emissivity from CERES
10.7 micron
3.9 micron
water: 0.990
water: 0.977
A large portion of land masses have surface emissivity > 0.97
DRAFT – Page 11 – July 31, 2017
First attempt: negative impact in
region 60-90 N/S
00 hr
72hr
Possible cause: cloud contamination ?
DRAFT – Page 12 – July 31, 2017
Second attempt: added constraints
• Cut latitudes > 60 deg
• Local gradient of topography < 50 m (was 100 m)
Strong impact on (O-B) stats for surface channels
DRAFT – Page 13 – July 31, 2017
T std error diff. (CNTL-EXP)
NH-Extratropics 6 Feb to 31March 2011
vs ERA Interim
vs own analysis
Consistent positive impact vs ERA Interim and own
DRAFT – Page 14 – July 31, 2017
T std diff (CNTL-EXP) vs ERA-Interim
Tropics
SH Extratropics
DRAFT – Page 15 – July 31, 2017
Zonal T STD difference (CNTL-EXP)
vs ERA-Interim
72-h
120-h
DRAFT – Page 16 – July 31, 2017
Time series of T std diff at 850 hPa
(a)
NH extratropics
(b)
Tropics
EXP superior to CNTL 54-66 % of cases at days 3-5
DRAFT – Page 17 – July 31, 2017
CNTL
EXP
850 hPa TT anomaly cor.
NH extratropics
Tropics
DRAFT – Page 18 – July 31, 2017
CNTL
EXP
120-h N-Extr vs ERA-Interim
CNTL
EXP
U
HR
GZ
DRAFT – Page 19 – July 31, 2017
T
Validation vs raobs 120-h
Feb-Mar 2011 (59 days)
SH-extratropics
CNTL
EXP
NH-extratropics
U
V
U
GZ
T
GZ
DRAFT – Page 20 – July 31, 2017
V
T
Validation vs raobs 120-h
FEB-Mar 2011 (59 days)
North America
Europe
CNTL
EXP
U
V
U
V
GZ
T
GZ
T
DRAFT – Page 21 – July 31, 2017
Added yield: about 15%
(for surface sensitive channels)
CNTL
EXP
Number of radiances assimilated for surface channel AIRS 787
CNTL: ~1400/6h EXP: ~1600/6h
Region: world, EXP excludes surface-sensitive channels at latitudes > 60
Radiance thinning is at 150 km
DRAFT – Page 22 – July 31, 2017
Std/bias of (O-P) and (O-A), AIRS 787
CNTL EXP
O-P
O-A
No major impact on bias. Strong assimilation over land,
with std (O-P) DRAFT
of ~1.7
(O-A)
– PageK,
23 –std
July 31,
2017 of 0.40 K (not shown)
Conclusion
• Very encouraging results, especially in NH where most
of new data are assimilated
Next steps:
• Relax limitations on emissivity, use MODIS-derived atlas
• Run a summer cycle
• Seek operational implementation
Longer term
• Evaluate problems specific to high latitudes
DRAFT – Page 24 – July 31, 2017
Update on Polar Communications
and Weather (PCW) mission
Goal:
Fill communication and meteorological
observation gaps in the Arctic
Most likely configuration:
2 satellites in highly elliptical orbits
Status:
Seeking approval from Government
in early 2015. Could start
operating in 2021.
Meteorological instrument:
Similar to ABI or FCI
A possible 2-sat HEO system
DRAFT – Page 25 – July 31, 2017
Weather:
High-Level Requirements
Capabilities
• ‘GEO-like’ continuous imaging of Arctic circumpolar region
• ‘GEO-like’ spatial (1-3 km) and temporal (15 min) resolution
• ‘Next-generation’ meteorological imager (ABI, FCI, 16+
channels)
• Near-real time processing to L1c for delivery to Environment
Canada
• Compatibility with GEO imagers as part of WMO Global
Observing System.
DRAFT – Page 26 – July 31, 2017
Orbit Comparison
(2-sat Constellation)
Molniya (12-h)
Apogee: 39,800 km
TAP (16-h)
Apogee: 43,500 km
Tundra (24-h)
Apogee: 48,300 km
DRAFT – Page 27 – July 31, 2017
Science Studies: Uniqueness of HEO
System vs LEO
23 LEO satellites needed to get 15 min
Imagery at 60 N, versus 2 HEO
Trichchenko and Garand,
Can. J. Remote Sensing, 2012
Also, capability to get image triplets for AMV vastly superior to LEO
DRAFT – Page 28 – July 31, 2017
Science Studies:
Impact on NWP
OSSE showing significant impact of PCW
AMVs filling gap in polar areas
Garand et al., JAMC, 2013
Positive impact (blue) at 120-h
In both polar regions from 4 satellites
DRAFT – Page 29 – July 31, 2017
Example of PCW Geometry Matching
[16-hr TAP Orbit] for intercalibration
VIIRS/SNPP
GEO
DRAFT – Page 30 – July 31, 2017
Trishchenko and Garand, GSICS Newsletter, 2012