FAIR USE POLICY:

If the data are obtained for potential use in a publication or presentation, 
kindly inform Penn State personnel (co2data@meteo.psu.edu) of the nature of 
this work.  We reserve the right to make corrections to the data based on 
scientific grounds, e.g., recalibration of standard gases or discovery of 
operational issues not known at the time of the release.  

Thanks!
_______________________

This dataset includes the model and observational data corresponding to the 
results described in:

Nathan, B., T. Lauvaux, J. Turnbull, S. Richardson, N. Miles, and K. Gurney: 
Source Sector Attribution of CO2 Emissions Using an Urban CO/CO2 Bayesian 
Inversion System. JGR: Atmospheres, 123, 13,611-13,621. 
https://doi.org/10.1029/2018JD029231


The files included that would be relevant to someone trying to recreate the 
results presented in the above manuscript are all of type .mat, created with 
MATLAB version R2015b.  They can be read in with some equivalent or compatible 
software.

The relevant .mat files are either for the pseudodata analysis or the real-data 
analysis, corresponding to subdirectories
"/eddy/s0/bjn5178/SectorInversions/5km/0.5RMS_Flux/DualSector_NoRand_New/" and "/eddy/s0/bjn5178/SectorInversions/5km/0.5RMS_Flux/DualSector_TowerInversion_NoRand_New/",
respectively.  In the first case, the files will have the naming convention of
"NSscalingmatA_2e5janapr2015_B_meanobserr.mat", and in the second
"NSscalingmatA_2e5janapr2015_B_meanobserr_windBT.mat".  Here "A" denotes a 
number between 1 and 3, corresponding to the ?scaling matrix? used when 
scaling the emissions, and "B" refers to the three scenarios of prior 
emissions, where they are scaled "low", "lesslow", or "high" compared to the 
truth.

Each .mat file contains multiple variables useful for the inversion analysis. 
I will use the placeholder ?X? to represent the fact that each variable has 
either a "1" or a "2" at the end to mark whether the variable corresponds to 
CO2 or CO, respectively.

flux_priX is the prior emission flux for the given species

flux_trueX is the ?true? emission flux for a given species, based on 
multiplying the prior by the transport operator H

flux_posX is the posterior flux for the given species

Asec1_colsumX is a variable with the total sums of the columns in the 
posterior uncertainty matrix A, here for sector 1 (the "combustion engine" 
sector).  This was saved along with the diagonal of this part of the A matrix 
so that A could be reconstructed without having to save the enormous A matrix

Asec1_diagX is the corresponding diagonal for the sector 1 portion of the A 
matrix Asec2_colsumX is like Asec1_colsumX, but for the second (?other?) sector

Asec2_diagX is like Asec1_diag, but for the second sector

Acov_colsumX is the corresponding column sum for the covariance sub-matrices of A

Acov_diagX is the corresponding diagonal for the covariance sub-matrices of A

Bsec1_colsumX is like Asec1_colsumX, but for the prior uncertainty matrix, B

Bsec1_diagX is like Asec1_diagX, but for the prior uncertainty matrix, B

Bsec2_colsumX is like Asec2_colsumX, but for the prior uncertainty matrix, B

Bsec2_diagX is like Asec2_diagX, but for the prior uncertainty matrix, B

Bcov_colsumX is like Acov_colsumX, but for the prior uncertainty matrix, B

Bcov_diagX is like Acov_diagX, but for the prior uncertainty matrix, B

err_redX is the error reduction map for species X gainX is the gain for 
species X after inversion

The usage of these variables in the creation of the figures of the manuscript 
can be found in the MATLAB workspace file 
/home/meteo/bjn5178/workspace20180309.m.