Digital Earth Emissivity Information System (DEEIS)
Impact Assessments: CRTM
Comparison of NASA AIRS v5 emissivity and STAR GrELS emissivity in CRTM simulations of GOES-12 Imager
- CRTM's current default IR emissivity database (NPOESS) was compared to two other emissivities: NASA AIRS v5 standard emissivity product and STAR's Greenness-adjusted Emissivity for Land Surface (GrELS)
- NASA AIRS v5 standard emissivity product (Level 2), at 45 hinge point frequencies, was composited for the month of August 2008 on 0.5 degree grid
- Screened for cloud fraction and emissivity standard deviation > 0.5
- Ascending (day) and descending (night) orbits were composited separately.
- Emissivity composite were averaged spatially for GFS surface types (13 land surface classes used in GDAS/GFS). This yields an emissivity spectrum for each surface type. Emissivity at channel wavelength can be interpolated between hingepoints.
- Greenness-adjusted Emissivity for Land Surface (GrELS) is based on combinations of reflectance measurements from the JPL Spectral Library. The combinations are specifically chosen for the 13 GFS surface types. Green vegetation fraction (GVF) is used to temporally adjust the emissivity (uses weekly climatology of GVF from 25 years of AVHRR data - from Jiang et al., 2008).
- CRTM was run for simulating GOES-12 Imager for July 21 2008, 0845 UTC (nighttime), for Channel 2 (3.9 um) and Channel 4 (10.7 um)
- Control run: CRTM used default NPOESS database as emissivity input
- First Test run: CRTM used NASA AIRS v5 standard emissivity product, processed as described above, as emissivity input
- Second Test run: CRTM used GrELS as emissivity input
- CRTM inputs were GDAS 6-hour atmospheric profiles (64 veritical layers) and surface temperature (improved Ts from NCEP/EMC/Land Group). Only cloud-free and land surface pixels were used.
Figure: AIRS monthly composite emissivity for August 2008 at hingepoint frequencies of 3.9 um and 10.6 um
Figure: Reflectance/Emissivity spectrum from GrELS (top row) and from NASA AIRS v5 (bottom two rows). AIRS spectrum derived by spatially averaging AIRS monthly composite for GFS Surface Types. Shown are two surface types: Broadleaf Tree with Groundcover (Savanna) and Bare Soil. Nighttime AIRS has lower reflectance (and higher emissivity) than daytime AIRS in mid-infrared region.
Figure: Brightness temperature difference (K), CRTM simulation minus GOES observation, for 3.9um channel, July 21 2008, 0845 UTC (nighttime). When CRTM was run with GrELS, differences are smaller.
Figure: Brightness temperature difference (K), CRTM simulation minus GOES observation, for 10.7um channel, July 21 2008, 0845 UTC (nighttime). When CRTM was run with GrELS, differences are smaller.
Figure: Brightness temperature bias (K), CRTM minus GOES, for 5 of the 13 GFS Surface Types. Best bias in red. GrELS database produces best bias for CRTM simulations of GOES Imager.
Figure: Brightness temperature RMSE (K), for 5 of the 13 GFS Surface Types. Best RMSE in red. GrELS database produces best RMSE for CRTM simulations of GOES Imager.
- STAR GrELS emissivity produces CRTM simulations of GOES-12 Imager brightness temperature with the lowest bias and RMSE compared to NASA AIRS v5 emissivity and NPOESS emissivity, for the GOES Imager 3.9um and 10.7um surface window channels.
- NASA AIRS v5 emissivity day-night differences imply that either daytime or nighttime retrievals may be less accurate.