Probability distributions (PDFs) of AIRS and CrIS scenes have been computed for the mean brightness temperature (BT) of several channels near 900 cm^-1. All scenes are included, but with random equal area sampling. This means that all scan angles are included, so that mean time differences between observations is minimized. Equal area sampling reduces the total size of the data to about 37% of all observations.

The goal of this work is to understand the the differential effects of (a) radiometric calibration offsets between CrIS and AIRS and (b) spatial sensitivity differences between these CrIS and AIRS that can appear as radiometric offsets.

The 900 cm^-1 spectral region has one of the widest range of BTs due to its sensitivity to both surface temperature and deep convective clouds, providing a wider range of PDFs than other (non-window) spectral regions.

Make a not here (or later) about how well AIRS and CrIS agree at very high BT temperatures in the shortwave that are created by glint. I believe their excellent agreement for shortwave glint is due to the rather broad (and uniform) spatial characteristics of glint compared to the same for high surface temperatures.

Figure 1 shows the PDFs for CrIS and AIRS for land and ocean separately. The PDF scale is set to integrate to unity. Note the higher temperatures seen over land, and the very sharp slope of the ocean PDFs that drop close to zero near 300K. The ocean surface temperature PDF extends to more like 305K?, which is masked for these channels by water vapor continuum absorption in the tropics. Note that at this scale you cannot see any differences in the PDFs for CrIS versus AIRS.

Figure 2 is the same PDF plot but using a log scale. Here you can see slight differences between CrIS and AIRS at the high BT end, and if you look closely at the low BT end for land only.

Figure 3 shows the absolute difference between PDFs for each BT bin. The largest percentage difference is for ocean scenes near 294K and corresponds to a difference of about 1.4% in the PDF for that bin, which is located very close to the PDF maximum.

The land PDF differences are more random when viewed in an absolute sense.

Figure 5 shows a representation of the BT contribution to the mean, as a function of the observed BT, for land and ocean. (This plot is not perfectly accurate since the mean of BT differences is not the same as the mean of radiances differences, but they are fairly similar.) The main observation is again that the clear scene BT difference is dominated by the peak of the PDF.

Here we view the PDF differences in terms of percent difference per BT bin. Only bins have been plotted that have a statistical signal-to-noise ratio of greater than 10. The ocean PDFs are very similar with the small dip near 295K mentioned above, and then some larger oscillations at higher BT bins, but where there are very very few observations. We suspect that this noisier data is polluted with differences in coastline avoidance for the CrIS and AIRS dataset. It does appear that over ocean CrIS sees slightly more deep convective clouds (DCCs).

Over land AIRS sees several percent more DCCs, with increases differences as you go from 220K to 190K. Although there are significant differences in DCCs over land versus ocean, it is difficult to know if these differences in PDFs are radiometric in origin, or due to spatial differences between AIRS and CrIS.

CrIS see significantly more hot scenes over land than AIRS, especially past 325K. We postulate that this is due to the small amount of scene smearing by AIRS since the scan mirror is moving while the scene is being recorded by AIRS. Simulations (not discussed here for now) suggest this is a likely cause for the higher number of hot CrIS scenes. Note that the small dip in the difference at 325K is likely where AIRS records the hotter scenes past 325K recorded by by CrIS at these hotter temperatures, thus the AIRS PDF goes up here.

Figure 5 shows the contribution of each BT bin to the mean BT for these channels, showing that the 295K bin is indeed causing the main difference in the global mean BT for ocean scenes. The mean BTs are summarized in the table below

The first three rows of the table below shows the mean BT differences between AIRS and CrIS these observations. These were done by summing the PDF times the radiance value for each BT bin to get the mean radiance, which is then converted to Bt.

We then modify the PDF bins to either (a) minimize the RMS difference between the CrIS and AIRS PDFs, or (b) force the mean BT difference to zero. This is done by applying a set of constant BT shift to all BT bins until these criteria are met. The last two rows of this table indicate how much the mean BT changes if we minimize the PDF RMS differencesd. We see that minimizing the RMS PDF difference changes the mean BT very little, -0.003K for land and +0.018K for ocean, significantly lower than the mean BT differences for land and ocean.

This suggests that the radiometric calibration is quite similar for CrIS and AIRS for these channels and that the main contributor to the mean BT difference for land versus ocean averaged scenes is the shape of the PDF, which we believe is dominated by difference in spatial averaging between AIRS and CrIS.

Figures 6 and 7 show how the land and ocean PDF differences between CrIS and AIRS change if we adjust the radiance scale by the constant BT offset required to make the mean BT values agree for both instruments. The blue curve is the PDF difference without adjustment. The orange curve adjusts the land/ocean BT scale offset derived from forcing the land/ocean mean BT value to be the same. The yellow curve adjusts the land/ocean BT scale offset derived from forcing the ocean/land mean BT value to be the same, ie we use the BT offset for ocean to adjust the land scale, and vice versa.