Carbon data harmonization for `moirai`

Description:

This folder contains a number of scripts which are used to generate 18 raster files (6 for soil carbon, 6 for above ground carbon, 6 for below ground carbon) that are used by moirai. The outputs are rasters where the carbon corresponding to individual land classes are harmonized to moirai’s land cover classes from the ESA land classes.

Input carbon rasters (and the shell scripts used to generate them)

1026 raster (.bil) files containing carbon by state and land ESA land class at a 5 arcmin resolution. These are generated using the bash scripts in the folder bash_scripts. To run the bash scripts, the user would need following,

The ESA land cover data (300 m resolution) that is available at https://www.esa-landcover-cci.org/?q=node/164
Soil carbon density data (250 m resolution) that is available at https://data.isric.org/geonetwork/srv/eng/catalog.search#/metadata/ea80098c-bb18-44d8-84dc-a8a1fbadc061
Above and below ground vegetation carbon data from Spawn et al. that is available at https://www.nature.com/articles/s41597-020-0444-4

Note that the user should run the above ground script (Above_ground_carbon.sh) first since this script will generate land cover files that are used by the other two scripts.The total run time of the three scripts is 7.5 hours (2.5 hours per script).

For the sake of convenience the final output files from the above are made available as a zip file (Carbon_rasters.zip)

Other inputs

These are the files in the folder input_files

1. basin_to_country_mapping.csv: Mapping file used to map GLU IDs to basin names.
1. ESA_classes.csv: 19 ESA land cover classes for the carbon accounting
1. ESA_moirai_classes.csv: Rules for the harmonization i.e. which land classes to map to the moirai cover with a prioritization scheme.
1. gcam_glu_boundaries_moirai_land_cells_3p1_0p5arcmin.tif: This is the GLU mapping raster that is generated by diagnostics/generate_moirai_spatial.R
1. refveg_area_carbon.bil: Carbon area for the reference carbon year set in moirai.h. Currently set to 2010. This file is copied from the base outputs of moirai
1. refveg_carbon_thematic.bil: Land cover classes from moirai in the reference carbon year (currently set to 2010). This file is also copied from the base outputs of moirai.

Main harmonization script

moirai_carbon_harmonization.R which performs the harmonization for three types of carbon (soil, above ground biomass, below ground biomass) for 6 states (weighted_average, min, max, q1, q3, median). Each type of carbon requires around 20 mins to run through (bringing the total runtime of the script to 1 hour).

Description of harmonization process

The figure below describes the harmonization process from start to finish.We begin with the raw carbon data and the ESA land classes. Carbon is calculated for each ESA class. This carbon is then resampled to get the 5arcmin map for the same class for each of the 6 states. Finally the R script is used to harmonize the data to the moirai land cover in the reference year.

Figure 1: Description of the harmonization process

As a part of the final harmonization we also use a nearest neighbor interpolation to increase coverage. The current nearest neighbor interpolation is based on 10 nearest neighbors within the same land class. The table below shows the results of NODATA cells remaining by land class after interpolation (both for vegetation and soil). As indicated by the table, even after the interpolation no data is found for the Polar desert and rock ice and Tundra land types.