This work has been in support of the determination of inter-sensor bias using RTP collections.

It is a part of work documented in the SNO Intercalibration paper.

Particular attention is drawn to the discrepency of the CrIS Clear RTP subsets.

Analysis is rather simple: Collect sufficient RTP data, being about a month worth. Compute population densities. Translate AIRS spectra to the CrIS (mostly the mid-resolution, but does not matter). Subset further if required, typically I use tropical ocean night (TON) but it does not matter much (except in the SW for day scenes). Correct the mean radiance for each channel for each sensor to a common view angle, for most accurate estimation of inter-sensor bias. This step does not affect the finding of the CrIS clear RTPs. Finally any post processing stats.

Data sources are defeined and configured in the scripts (listed next), but primarly:
/asl/rtp/rtp_cris_ccast_{hires,lowres}/{random,clear}/
/asl/rtp/rtp_airicrad_v{6,672}/{random,clear}/
/asl/rtp_iasi1/{random,clear}/

Most of this work has concerned months selected from 2019.

There are two sets of scripts, depending whether to compare AIRS vs CrIS or AIRS vs IASI.
/home/chepplew/projects/intercal/{airs_cris_rtp_bias.m, airs_iasi_rtp_bias.m}
Each script has configurable settings to select date range, choose the random or clear rtp sets, and any further subsetting.

The two parent scripts call child scripts in the same directory:
load_some_airs_rtp_data.m
load_some_cris_rtp_data.m
load_some_iasi_rtp_data.m
and
correct_mean_satzen_bias.m
All other dependency scripts from /asl/matlib/ are referenced therein.

For now please refer to plots posted on slack:sno_intercal. plot1

The primary observation is that the CrIS clear subset observations, when comparing radiances in the LW window around 900 cm-1 do not resemble those from the AIRS or IASI as they should - this is especially obvious using the Tropical Ocean night subset.

Whereas sensor bias from all other RTP data is consistent with the SNO bias, those using the clear rtp are not.

Please check my work and see if you agree to the observation of discrepent CrIS clear RTP subsets.