Computing a correlation length scale from MFLL-OCO2 CO2 differences, and accounting for correlated errors when assimilating OCO-2 data

This dataset contains code and data used in  'A new exponentially-decaying error correlation model for assimilating OCO-2 column-average CO2 data, using a length scale computed from airborne lidar measurements'  by David F. Baker, Emily Bell, Kenneth J. Davis, Joel F. Campbell, Bing Lin, and Jeremy Dobler, submitted to Geoscientific Model Development.

In particular, the MATLAB script used to compute the autocorrelation spectrum of  Multi-functional Fiber Laser LiDAR (MFLL) and Orbiting Carbon Observatory (OCO-2) column CO2 differences (in Section 2.2 of the paper) is given here as file

comp_MFLL_OCO2_autocorrl_spectrum.m

along with the needed MFLL and OCO-2 CO2 data for each of the six flights analyzed (as described in Section 2.1 of the paper) in files:

20160727_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20160805_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20170215_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20170308_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20171022_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20171027_mfll_averaged_L1_RA_GMAO_ACTadj.h5
20160727_oco_averaged_B9_GMAO_ACTadj.h5
20160805_oco_averaged_B9_GMAO_ACTadj.h5
20170215_oco_averaged_B9_GMAO_ACTadj.h5
20170308_oco_averaged_B9_GMAO_ACTadj.h5
20171022_oco_averaged_B9_GMAO_ACTadj.h5
20171027_oco_averaged_B9_GMAO_ACTadj.h5

The MFLL data given here was downloaded in late 2018 in the form of L1b files (calibrated radiances), as described in Bell et al (2020).

In the second part of the paper, different error correlation models are presented and applied to the averaging of OCO-2 column CO2 data.  The original bias-corrected OCO-2 data, in the form of daily OCO-2 version 10 "Lite" files, have been from obtained from NASA's GES DISC data repository, here:
https://disc.gsfc.nasa.gov/datasets/OCO2_L2_Lite_FP_10r/summary?keywords=OCO2_L2_Lite_FP

The bias-corrected column CO2 retrievals, their uncertainties, and other parameters needed for this analysis were extracted from these
"Lite" files and saved to daily files, which have been packaged up in the following compressed tarball:
OCO2_XCO2_2014_2020.tar.gz

These daily files are read in and averaged across 2-second and 10-second spans (as described in Section 3.5 of the paper), using the different error correlation models outlined in the paper.  The code that implements these averages is given in the following two FORTRAN programs:

Make_OCO2_2sec_averages.f90
Make_OCO2_10sec_averages.f90

which need the following list of days having good OCO-2 data:

OCO2_dates.txt

Program "Make_OCO2_2sec_averages.f90" averages the OCO-2 data across a 2-second (~13.5 km long) span along the groundtrack, collapsing the relatively thin data swath into a one-dimensional data record, upon which the one-dimensional averaging models describe in Sections 3.1 and 3.2 of the paper may be applied.  Program "Make_OCO2_10sec_averages.f90" implements these averaging models, which average the 2-second averages across longer, 10-second (~67.5 km) spans.  Please see the manuscript for more information on the data and methods provided here.