This is the repository for the study:
Taylor, S.D. and Browning, D.M., 2021. Multi-scale assessment of a grassland productivity model. Biogeosciences, 18(6), pp.2213-2220. https://doi.org/10.5194/bg-18-2213-2021
The following R scripts, run in order, will generate all the appropriate data used in this analysis.
data_prep/generate_site_list.R - make site_list.csv file from the critera specified.
data_prep/generate_model_fitting_sets.R - populate the model_fitting_set_info folder with information on the different model fitting scales.
data_prep/download_phenocam_data.R - download the raw phenocam data to data/phenocam/
data_prep/download_daymet_values.R - download daymet climate data for each site to daymet_data.csv
data_prep/process_phenocam_data.R - preprocess the downloaded phenocam data to the file data/processed_phenocam_data.csv
data_prep/process_soil_values.R - extract the soil variables from the rasters in data/soil_rasters, with results going in to site_list.csv
data/ - This holds the model input data (phenocam, daymet, soil data). The two model output files (full_model_predictions.csv, and cv_model_predictions.csv) are available in the zenodo repo. https://doi.org/10.5281/zenodo.3897319
The following python scripts fit all iterations of the models. The _cv version fit and predict a small subset of the models in a leave 1 out cross validation scheme. See the paper section 2.4 Model Evaluation.
The files populate the results folder from which the analysis scripts read from.
Note the fitting procedures are designed to run on an HPC using the python library dask.
The actual PhenoGrass code is in the python package GrasslandModels located at https://github.com/sdtaylor/GrasslandModels
fit_model.py
fit_model_cv.py
apply_models.py
apply_model_cv.py
The following files are used by the above script to fascilitate data loading, saving/loading fitted models, and fitting models on a dask.distributed cluster.
tools/dask_tools.py
tools/load_data.py
tools/load_models.py
The fitted_models folder contains saved model parameterizations from all model fits described in the paper. The first set of models are in 2020-07-29_0a82f11673e54a9c, while the leave 1-out cross validation models are in 2020-08-23_69f5e0932e7b4970. These are described with metadata in fitted_models/model_sets.json, and are designed to be accessed and iterated thru with the functions in tools/load_models.py.
analysis/analyze_errors.R - This generates the primary error table in the manuscript, as well as a site level error table in results
analysis/analyze_cv_errors.R - This genates the CV errors
analysis/discussion_figure_site_gcc_comparison.R - This generates the discussion figure of smoothed Grassland GCC
analysis/make_error_tables.R - Make latex error tables for the manuscript
analysis/error_figures.R - Make Figures 3 and 4 showing 1:1 comparisons of the 11 model iterations
analysis/NA_desert_grassland_fig.R - Figure S1 highlighting predicted values in desert grasslands.
analysis/site_level_model_output_figure.R - Make a large figure showing predictions and actuall GCC for every site
analysis/site_map.R - create the map in the manuscript
analysis/site_table.R - create the site table in the manuscript appendix