sergio66/JGR_July2025: v1.03 Release this should be master branch

This repo has two directories A) PLOTTER : for reproducing the figures in the JGR paper B) CODE : for processing the tiled AIRS L1C data using Optimal Estimation

=============== PLOTTER : This is a data/source code directory to make figures for JGR 2025JD043501R July 2025 Revision

We have cut and paste from the analysis source code(s) so as to make the matlab plotting source codes (and matfiles), arranged in the order of the 14 figures in main text and 3 figures in the appendices.

You cn either run the code individually in the Matlab worskpace eg clear all; make_jgr_fig7.m or you can loop over them all by running clear all; driver_make_all.m

In either case, we suggest opening 4 matlab plot winows and arrange them in a 2x2 tile fashion Fig 1 | Fig 2 ----- |------- Fig 3 | Fig 4

After making the individual panels for any one figure, we print them as png or pdf and then use "imagemagick" on a Mac to combine the panels for submission to JGR eg gm montage -trim -geometry 1200x950+10+10 -tile 2x1 fig3a.pdf fig3b.pdf fig3_merged.pdf

============= CODE : This spectral trend climate analysis code assumes you have

A : Spectral trends and uncertainties (a) tiled AIRS radiance trends and uncertainties and lag1 corrections 4608 tiles x 2645 AIRS L1C channels for each of the above (b) you can also generate spectral trends from eg ERA5, MERRA2, AIRS v7 L3, CLIMCAPS L3 monthly data matched to the tiles

B : Accurate Jacobians (a) jacobians need to be available (for T/WV/O3/surf temp/trace gas) for each tile T(z), WV(z), O3(z) jacobians are 100 layers x 2645 channels for AIRS L1C, for each tile SKT jacs are 1 x 2645 for each tile Trace Gas Column jacs (Co2,N2O,CH4,CFC11,CFC12) are also 1 x 2645 for each tile

C : Running the code Code is designed to run on the UMBC HPC cluster One tile can be run using the main driver code clust_run_retrieval_setlatbin_AIRS_loop_lonbin.m The code tkaes any one of the 64 latitude bins, and loops over the 72 longitude bins Inside that code, there are many default setting you can adjust using "topts," such as day or night number of channels which quantile you are using (Q01-Q05) do you wish to stick to AIRS 100 layers, or combine/sum jacobians for eg 50 layers or 33 layers tweak the covariance matrices (Tikhonov or exponential decay, diagonal etc)

This is fast, less than 30 seconds per tile, or about 30 mins per 72 lonbins (one complete latitude bin).

D : output files go to Output or Output_Day (depending on whether you chose day trends or night trends to be analyzed)

We are not providing eg AIRS L1C tiled quantile trends since those are huge .mat files and can be generated by downloading the L1C data and tiling it by 16 day intervals.

Questions can be directed to Sergio DeSouza-Machado (sergio@umbc.edu)