A study of road closure due to rainfall and flood zone based on logistic regression

Abstract

This study investigates the possibility of road closure due to flooding, based on varying amounts of rainfall and in different flood zones. Logistic regression analysis was employed, utilizing the road closure data from TxDOT in 2017 and the daily rainfall data provided by Harris County Flood Warning System (HCFWS). The definition of flood zones were established in three different contexts: the overall road network area, areas recognized as prone by FEMA, and areas where road closure occurred during Hurricane Harvey. Rainfall data were segmented into periods spanning 1–4 days. The findings indicate a positive correlation between rainfall intensity and the likelihood of road closures. In addition, Within a four-day window, the rainfall in the current day and 3-day prior played a more influential role in predicting road closure. Rainfall on the first day tended to wash debris into drainage pipeline, leading to blockage and substantially diminishing street drainage capabilities. This, in turn, heightened the susceptibility of roads to flooding in subsequent days. Finally, an analysis of the 2017 road closure and rainfall data revealed a discrepancy between the flood zone delineated by FEMA and those identified in this study. Nonetheless, critical infrastructures were present in both categories of areas. This study demonstrates how road closure predictions can be instrumental in identifying optimal routes to critical infrastructures, such as hospitals, in flood prone areas.