This repository contains the code for experiments and figures for the paper "Unrecognised water limitation is a main source of uncertainty for models of terrestrial photosynthesis".

To reproduce the experiments, follow these steps:

Install environment with conda

conda create -n gpp-ml python=3.10.12
conda activate gpp-ml
pip install -r requirements.txt

Note: Install torch with cuda if applicable on your system.

Run model experiments

The following commands run experiments for the deep learning models:

python -u global_model.py --model_type LSTM --early_stopping --layernorm --modis
python -u global_model.py --model_type MLP --early_stopping --modis
python -u site_specific_model --model_type LSTM --early_stopping --layernorm --modis
python -u site_specific_model --model_type MLP --early_stopping --modis
python -u global_model.py --model_type LSTM --early_stopping --layernorm --extra_features --modis
python -u global_model.py --model_type MLP --early_stopping --extra_features --modis

The P-model experiment can be run using the script R/pmodel_global.R.