Machine learning analysis of morphometry-extreme flood links in the Lower Colorado River Basin

Abstract:

Extreme flood hazards characterize the Lower Colorado River basin (LCRB) due to the complex terrain and entrenched river channels. Evaluating basin morphometry aids understanding of the physical behavior of watersheds with respect to extreme floods events. However, extracting basin morphometric characteristics is computationally expensive and time-consuming. Conventional approaches lack effective tools for linking morphometric indices to extreme floods, thereby posing a great challenge for extreme flood prediction. In this study, we extracted 41 basin morphometric parameters for 372 watersheds in the LCRB from a 10-m DEM. We then employed the Random Forest (RF) regression to link these morphometric features to the floods-of-record. The results show that the RF with the GridSearchCV process and the OOB error rate generate similarly good results for extreme flood prediction. The most significant variables for predicting the maximum annual peak discharge (MAP) are the relative perimeter, the total basin area, and the length-area relation, followed by the main channel length, total stream length, and ruggedness number. Similar improvement for predicting the peak discharge per unit area (UP) is achieved using the maximum height of the basin, total basin relief, and the relief ratio. Based on the most significant variables, the partitional clustering K-means algorithm was used to identify three flood regions in the LCRB. This initial effort shows that data-driven machine learning, such as the RF regression and the K-means algorithm, can help link basin morphometry to measures of extreme flooding, thereby advancing our understanding of regional large flood behavior and improving flood risk analyses for the Southwestern U.S.