This repository contains code to reproduce the results in the paper "From Many to One: Consensus Inference in a MIP". It implements the SUPE-ANOVA framework for the specific case of analysing the results of the OCO-2 MIPv7.
This repository uses make to orchestrate the workflow and it is expected that it is run from a command line. It has been tested on macOS and ought to work on any UNIX-like system. The code in this repository is written in the R programming language and any version of R newer than 4.0 is expected to work.
The software dependencies are encoded using the renv package. If you have a new enough version of R installed, you can run the following command from the repository root directory:
make bootstrap
This will use renv to install all the required R packages. There may be some system dependencies like netcdf4 that you need to install; these cases will need to be worked through individually if the above command fails. For example, the sf package will be installed, and it has several dependencies documented at https://r-spatial.github.io/sf/index.html.
You need to create a folder named data in the root directory and add the following files to it:
- The file
L4-regionfluxes_2018-07-30.tar.gz, available from the "Level 4 product download" section of the OCO-2 MIPv7 website. This tar file should be expanded into the data directory, resulting in a directorydata/ct/andy/OCO-2/l4_regionfluxeswhich contains a series of NetCDF files. - The OCO-2 MIPv7 region mask file,
oco2_regions_l4mip_v7.nc, also available from the OCO-2 MIPv7 website. The link is currently on the homepage under "38 regions mask file (region definitions)". This file should also be placed in thedatadirectory.
The workflow can be run simply with
make -j4
You can change the -j4 to match how many cores are available in your system. If this command completes successfully, the figures folder will contain all the figures and tables from the paper.