This repository provides data, code, and supplementary information associated with the paper "Uncovering Effects of Exposure and Vulnerability on Atmospheric River Flood Damage using Interpretable Machine Learning" (doi:XX). The _data
folder contains code to collect all of the hazard, exposure, vulnerability, and damage data used to generate the random forest models. The _scripts
folder contains code to train and evaluate the random forest models, as well as code to calculate the SHAP values and ALE curves. The _results
folder contains code and formatted markdown files that reproduce all figures and numerical results from the papers. The markdown files are also available as interactive HTML documents through the links below.
- reproduce_figures.html replicates all figures and numerical results from the main body of the paper.
- figure4.html replicates Figure 4 for all four random forest models fit to different spatial and temporal scales.
- takeup.html replicates the calculations from Appendix I and includes a figure comparing observed versus expected NFIP takeup behavior by county.