Reducing the Human Impacts of Flash Floods: Development of Microdata and Causal Model to Inform Mitigation and Preparedness

This published project contains the code (in R programming language) and the data used to develop model to understand why unsafe conditions exist during flash flood events and the human impacts (fatalities and injuries) of flash floods. This study has been published here: https://doi.org/10.1007/s11069-023-05845-x. The study (1) identifies climatic, environmental, and situational factors that affect the occurrence of fatalities and injuries in flash food events and provides a predictive model to estimate the likelihood of these occurrences, and (2) develop a model to predict the likelihood of human harm resulting from a flash flood event using these factors in a census tract. Upon testing three variants of the logit model, we found the rare event logistic model (Relogit) to be the best performing one. This model can be used as a simulator tool for informing flash flood mitigation. The file ‘Flash Flood Human Harm.Rmd’ contains the code used to develop the model. The ‘flash_flood_data_all.csv’ file contains the dataset for 6065 flash flood events in the study period 2005-2019. The data were harvested from structured and unstructured data sources on the web like the NOAA Storm Events Database, American Community Survey, National Land Cover Database, TxDOT, Texas Natural Resources Information System, and National Hydrography dataset. Further details are available in the readme file. The remaining two csv files are the train and test files created using ‘flash_flood_data_all.csv’ to train a rare event logistic regression model and test its performance. The assembled data can be used to develop enhanced predictive models for estimating the impacts of future flash flooding at the community scale.