Más informada CASA (MiCASA)

The Más informada Carnegie-Ames-Stanford Approach (MiCASA) is a diagnostic terrestrial biospheric model. It is a derivative of the Global Fire Emissions Dataset version 3 (GFED3) CASA variant (CASA-GFED 3) produced and maintained at NASA's Goddard Space Flight Center (GSFC). CASA-GFED 3 uses reflectance data from NASA's Moderate Resolution Imaging Spectrometer (MODIS) aboard the Terra and Aqua spacecraft to determine atmosphere-biosphere carbon fluxes due to primary production, respiration, and biomass burning. MiCASA extends and improves upon CASA-GFED 3 in several ways. This includes the ability to run at resolutions finer than 0.5 degrees, generalized and updated meteorology and soil property inputs, and brand new code to process native MODIS and Visible Infrared Imaging Radiometer Suite (VIIRS) tiles into land cover, vegetation, and burned area driver data (i.e., más informada).

MiCASA is also a data product that this code repository can generate. More information about the MiCASA code and data product can be found in the README file for the version 1 data release. A peer-reviewed description and validation paper is in progress.

Recipe

Most of the work involved in getting a run going revolves around building inputs and spinning up, as described in the following two subsections. Once this is done, daily runs and side experiments are fairly straightforward, as described in the final subsection.

Building MODIS/VIIRS inputs

1. Install the MiCASA Python package in editable mode following the instructions in requirements.txt.
2. Run modvir --help to get an example of how to generate the MODIS/VIIRS files.
3. Use the utility src/CASA/utils/prep.sh to batch generate inputs.

NOTE: MiCASA was originally designed to run on NASA high-performance computing assets. Mirroring the entire MCD12Q1, MOD44B, MCD43A4, and MCD64A1 collections takes a lot of time and space, especially the MCD43A4 collection. The next version of MiCASA will support running on limited domains and using services like OPeNDAP and the Earthdata Cloud to make this process easier. We also plan to archive the inputs we use so that the user need not reproduce this step. Nevertheless, these files will be about 4GB per year, so about 100GB total for the entire 24+ year record.

Run names

Running MiCASA requies setting the variable runname. This variable determines which settings to use in defineConstants.m. While any choice of setting should be possible for a given runname, we leave some generic processing options open for ease of use. In general:

• MODIS/VIIRS driver data is read from data/runname/drivers.
• Factorial runs can be denoted with a slash, e.g., runname/XYZ will use the driver data, maps, and meteorology from runname.

Spinning up

1. Build the climatological and annual inputs needed for spin-up. Change into the src/CASA directory and run the following in Matlab/Octave:

   runname = 'v1A/spinup';
   makeCASAmapsClim;
   makeCASAmapsAnnual;
   

2. Run monthly CASA to spin up. Run the following in Matlab/Octave:

   runname = 'v1A/spinup';
   CASA;
   

Daily runs (with LoFI)

Once inputs are built and the monthly spin-up is done, you can run daily CASA in Matlab/Octave by doing the following:

runname = 'v1A';
CASA;
convertOutput;
lofi.make_sink;
lofi.make_3hrly_meteo;
lofi.make_3hrly_land;

If desired, run the post-processing

./utils/post/batch.sh process.sh

This will end once the MODIS/VIIRS inputs end. To continue, simply update the MODIS/VIIRS inputs using the modvir command as above. The remaining inputs do not need to be updated: they are either climatological or specifically for the monthly spin-up.

NRT runs (with LoFI)

More information coming. This uses NRT QFED to extend the biomass burning into NRT. Unless that's publicly available somewhere, we're probably documenting this for our own purposes.

Maintaining local configurations

MiCASA configuration is still pretty rough, but I've made a lot of progress. The main configuration files are src/CASA/defineConstants.m and bin/post/setup.sh. The former defines the CASA model settings along with input and output directories and the latter post processes. It is not necessary to run the post processing for personal installs.

You may make changes to these configuration files that you won't want a git pull to overwrite. In such a case, try

git stash
git pull
git stash pop

If you don't wish to keep changes to files, simply delete them and then pull.

Changes from version 1 to 1A

• Half-weighted magnitude inversion (QC = 1) NBAR values
• Fixed daily burned area disaggregation
• Fixed fractional tree/herb cover for crop sink and SOC
• Moved to GEOS IT meteorology