Impact of AIRS-v7 Calibration on AIRS-CrIS SNOs
Wed, Nov 7, 20181 Location
- Code location:
/home/strow/Work/Sno/Airs_v7/pagano_with_snos.m - Document location:
/home/strow/Git/aslhugo/content/strowpages/pagano_v7
2 Approach
Tom Pagano (JPL) has produced a new AIRS v7 calibration (present calibration is deemed v5). These calibration changes have some scene B(T) dependence. We provide Tom with the mean B(T) values for the AIRS-CrIS SNOs so he could create the mean v7 - v5 calibration differences appropriate for the SNO data.
Tom's calibration differences were converted to the CrIS NSR spectral response functions and then compared to the AIRS-CrIS SNOs to determine if the new calibration fits the AIRS-CrIS differences better than the existing AIRS v5 calibration.
This was done by taking a mean AIRS spectrum and converting it to the CrIS NSR ILS. We then take the same AIRS spectrum and add in the v7 - v5 calibration differences and convert that to the CrIS NSR ILS. The difference between these two spectra is then the v7 - v5 differences but translated to the CrIS NSR ILS.
For comparisons, we add the (v7 - v5) differences to the (CrIS - AIRS) SNOs to form (CrIS - AIRS_v7) SNOs.
3 Results and Discussion
The first figure shows the (CrIS - AIRS_v5) SNOs in color, with different colors denoting different AIRS detector arrays. The (CrIS - AIRS_v7) SNOs are shown in dark grey for all detector arrays. The following figures are zooms of the first figure to more easily view these data.
Figure 1: (CrIS - AIRS_v5) SNOs (in color) with (CrIS - AIRS_v7) SNOs in dark gray. Colors for (CrIS - AIRS_v5) denote the different AIRS detector arrays.
It is difficult to see differences between AIRS_v5 and AIRS_v7 at the scale used in Fig. 1, although AIRS_v7 appears to be a significant improvement for modules M-04a and M-04b between 1460 and 1613 cm-1. The improvement shows up in both the mean and the reduction of the scatter of individual channels.
3.1 Longwave
Figure 2 shows the full longwave region. The window channels have not changed much between AIRS v5 and v7, although the AIRS_v5 SNO's are closer to zero this is not significant since the CrIS radiometry is also uncertain at this level. Past 1000 cm-1 there is little to no improvement with AIRS_v7. These two modules, M-06 and especially M-05, have long been suspected to have problems. Again, although it is difficult to ascribe overall differences in these SNOs to AIRS or CrIS, the high variability, especially in M-05, is very likely due to AIRS calibration errors.
Figure 2: Longwave zoom of Fig. 1.
Figure 3 is a zoom of the extreme longwave channels. Here we see some more dramatic improvements with AIRS_v7, especially for modules M-11 and M-10 between 690 and 780 cm-1. The AIRS v7 calibration smooths out the SNO differences among channels and between modules. The sharp discontinuity between thse modules near 730 cm-1 is now gone! The variability among channels in module M-12 is also lessened with the v7 calibration. The large spike in M-12 at 667 cm-1 may contain some scene dependent calibration issues since the observed radiance differences between channels in that region are extremely large.
Overall the v7 calibration appears to either be an improvement over v5, or no change.
Figure 3: Extreme longwave zoom of Fig. 1.
3.2 Midwave
The AIRS v7 calibration does not change the observed CrIS SNOs much for the 1200 to 1450 cm-1 region, differences are on the 0.05K scale at most and there is still considerable scatter channel-to-channel. As mentioned above, the v7 calibration does appear to improve the M-05 and M-06 scatter and offsets from the other modules in the midwave. It does appear that M-05 and M-06 continue to have a slight offset from the other modules and we suspect that this is due to AIRS, not CrIS, although this is difficult to prove. Comparisons to IASI might help to make this statement more definitive.
Figure 4: Midwave zoom of Fig. 1.
3.3 Shortwave
The AIRS v7 calibration made only very small changes to the SNO differences. There seems to be a small systematic increase in the AIRS_v7 SNO differences for modules M02-b and M-01b from 2180 to 2422 cm-1 of ~0.05K, but no change in channel variability.
Figure 5: Shortwave zoom of Fig. 1.
3.4 Module Summary Statistics
Here we present some module averaged statistics to provide more of an overview of the AIRS_v5 and AIRS_v7 differences relative to CrIS. Figure 6 is the module averaged mean SNO differences for AIRS_v5 and AIRS_v7. These show that the overall differences between v5 and v7 vary with module and do not exhibit anything definitive. However, given that these differences are at most 0.15K it is hard to know which sensor is more in error, and likely both sensors have errors at this level. SNPP CrIS and NOAA-20 CrIS have calibration differences of ~0.1K in the extreme longwave, which could mean that CrIS is mostly responsible for these SNO differences in that region.
Figure 6: (CrIS - AIRS) SNO mean differences by module for both AIRS_v5 and AIRS_v7.
More interesting is Fig. 7 where we show the standard deviation (among channels) in the SNO values for both AIRS_v5 and AIRS_v7. As expected from the visual examination of the SNOs provided earlier, we see singnificant reductions in the standard deviations in the extreme longwave of about 30%. This figure also shows a slight reduction of about 15% in channel-to-channel variability in the M-05 and M-06 modules past 1450 cm-1
4 Summary
We show clear evidence that the AIRS v7 calibration is an improvement over the AIRS v5 calibration for AIRS modules M-12, M-11, M-10 in the extreme longwave and modules M-05 and M-06 in the higher wavenumber region of the midwave. We believe this is an improvement primarily not because of any better agreement with CrIS on average, but because AIRS_v7 reduces the channel-to-channel variability within each module.
Figure 7: (CrIS - AIRS) SNO standard deviation of the mean differences by module for both AIRS_v5 and AIRS_v7.