JEV sylvatic co-occurrence probability in Australia under climate change

This collection includes the input data, outputs and python code base for the publication version of JEV sylvatic co-occurrence probability. This data was originally developed as part of the Australian Climate Services National Climate Risk Assessment second pass report, to investigate how the risk of JEV across Australia may change under future climate change projections. The original outputs from that report were published previously (https://doi.org/10.25919/feye-9073).

Data consists of:
- Historic (1981 - 2010), baseline (1986 - 2005) and climate projections for 2025(2016 - 2045), 2050(2036 - 2066) and 2090(2075 - 2105) of 16 bioclimatic variables, at an 0.05 decimal degree grid across all of continental Australia. Bioclimatic variables are stored as geotiffs.
- Presence/absence data for 16 species are provided in pickled python objects (geopandas dataframes), which include observation locations and background bioclimatic data used in model building. Presence/absence data were sourced from the Australian Living Atlas and VectorMap project for the period 1950 - 2024.
- A geotiff of the scaled distance to nearest city used for bias correction is also provided.

Code:
Python codes and batch files are stored in the code folder. The primary files are analysis.py and summary.py. analysis.py handles the data and model building steps. summary.py handles the calculation of jev co-occurrence, summary statistics and model skill as well as figure generation.
Outputs:
The outputs folder contains the intermediary files and some summary tables. These can be used within summary.py, so that the models and data don't have to be regenerated. Most output files are identified by the species, model settings and data that the model has been applied to. _set.pbz2 files contain the data used to train and cross-validate the models. _best_set.pbz2 files contain the hyperparameter settings of the models.