The 2019 methane budget and uncertainties at 1° resolution and each country through Bayesian integration Of GOSAT total column methane data and a priori inventory estimates

There are four sets of data corresponding to the  paper “The 2019 Methane Budget And Uncertainties At 1 Degree Resolution And Each Country Through Bayesian Integration Of GOSAT Total Column Methane Data And A Priori Inventory Estimates”

These include a summary product file A) TopDownEmissions_GOSAT_GEOS_CHEM_2019.nc, B) a set of netcdf files describing the a priori inputs used for the sectoral attribution, C) a set of netcdf files describing the a priori covariances and emissions (or results) of the sectoral attribution, and D) the results from the top-down methane flux inversion that is based on GOSAT data for 2019 and the GEOS-Chem model.

Following is a description of each of the data sets. This description can also be found in the ReadMe file.

A) Summary Product File:

The file TopDownEmissions_GOSAT_GEOS_CHEM_2019.nc contains the prior and posterior emissions and their uncertainties (squared) by sector at 1 degree resolution. 

An example of the variable name is “prior_wetland” which describes the prior emissions used for this analysis. “prior_unc_wetland” is the projected uncertainty at 1 degree resolution. Similarly, “post_wetland” and “post_unc_wetland” are the posterior (or estimated) emissions informed by the GOSAT total column methane data and its uncertainties respectively.

Similar variables are also available for coal, oil, gas, livestock, waste, rice, fires, and geological.

Some notes the user should be aware of that are discussed in the manuscript:

1) Livestock prior emissions are too low by ~20% relative to the source of the emissions (Wolf et al. 2017)

2) Livestock emissions for pixels near Sudan are zero so that posterior livestock emissions are projected to nearby areas in Ethiopia. One should therefore combine emissions and uncertainties for these regions

3) The prior emissions for waste are only for landfills and do not include liquid waste water. 

These issues do not change the conclusions about integrated fossil and integrated livestock / rice / waste emissions described in the paper. They are addressed in a subsequent paper “Verifying Methane Inventories and Trends With Atmospheric Methane Data”, AGU Advances, 2023.

Note that the uncertainties are only appropriate at the grid cell  (i.e. 1 degree) resolution. As the uncertainties are correlated between grid cells, projecting the emission uncertainties to a coarser scale results in a too small uncertainty. 

An alternative, ad-hoc approach for calculating uncertainties at coarser scales could be to 1) linearly sum the uncertainties for the selected region (this assumes the uncertainties are 100% correlated for the selected region, then 2) linearly sum the square of the uncertainties (the value provided by the corresponding matrix in this file) and take the square root. 3) Take the average between these values. 

To properly project to this coarser resolution you should use the prior or posterior covariances provided in this data repository and which are described next.

B) Prior covariances 

Prior covariances are provided in a zip file by sector (e.g. CoalPrior.zip). Within each zip file are netcdf files related to 8 regions. We subdivide the globe into 8 regions as the large size of the covariances may become problematic for downloading and use.  However, a consequence of this regional division is that the calculated uncertainties for the emissions of the different regions are larger. The label might be for an entire continent (e.g. N. America) or for the country that contains the largest emissions within a region (e.g. India, Russia). The emission sector is given as the first part of the name (e.g. Coal). However two names were shortened or renamed to reflect other inventories: Live = Livestock and Farm = Rice

Each a priori covariance contains five variables. While each file is for a specific region (e.g. N. America), the longitude and latitude arrays are for the whole globe.

Here are the lon/lat ranges for each region  the ranges are lon1,lon2,lat1,lat2

N. America ,-175,-40,25,80

S. America: -130,-30,-65,25

Europe/N. Africa:  -24,60,20,80

Africa: -22,60,-40,20

Russia: 60,179,50,90

India: 60,90,5,50

Asia 90,179,5,50

Australia/Indonesia 90,179,-45,5,

The variables contained in each file are:

1) ch4_emissions contains a vector of emissions with units of Tg CH4/ year 

2) The lonlatindex is the index corresponding to ch4_emissions and it is where ch4_emissions are larger than zero

A value of 0 in the lonlatindex refers to the zeroth element in the longitude and latitude array or -179.5, -89.5.  A value of 1 would be -178.5,-89.5, A value of 360 corresponds to -179.5,-88.5

3 and 4) The a priori covariances contain latitude and longitude matrices and are the same for all files and contain the lat/lon values for the whole globe and are generated for user convenience. 

5) SA is the prior covariance corresponding to ch4_emissions. Units are Tg CH4/yr  (squared) or Tg CH4/yr * Tg CH4/yr

C) Posterior Covariances

Posterior covariances have a slightly different naming e.g.   global_gosat_2010_2018_regionemissiontype.nc. The variables in the posterior covariance have slightly different names but correspond to the variables in the prior covariances.

1) posterior_emission contains a vector of emissions with units of Tg CH4/ year (can be compared to ch4_emissions in the prior covariance)

2) The lonlatindex is the index corresponding to posterior_emission

A value of 0 in the lonlatindex refers to the zeroth element in the longitude and latitude array or -179.5, -89.5.  A value of 1 would be -178.5,-89.5, A value of 360 corresponds to -179.5,-88.5

3) posterior_covariance is posterior prior covariance corresponding to posterior_emission. Units are Tg CH4/yr  (squared) or Tg CH4/yr * Tg CH4/yr

D) (Files describing top-down methane flux inversion) These next set of files represent the integrated fluxes from the top-down GOSAT based inversion that are used to quantify the sector based emissions. They are described 

1) CH4_emissions.nc: prior fluxes (integrated emissions within a grid cell) for GOSAT 2019 Inversion (from Qu et al. ACP 2021)

2) Post_SF_emissions.nc: posterior scaling factors that when applied to prior fluxes yield posterior fluxes

3) cluster_map.pkl: mapping from flux state vector element index to lat/lon coordinate

4) S_poscov.pkl: posterior scaling factor covariance matrix that when applied to prior error covariance matrix yields posterior flux covariance matrix