Evaluating levee failure susceptibility on the Mississippi River using logistic regression analysis

The goal of this investigation was to test the relative importance of geologic, geomorphic, and other physical factors that have led to levee failures through the past century along much of the Mississippi River. Logistic regression analysis assessed site characteristics at levee-failure locations identified in a database of historical levee failures along 685 km of the Middle and Lower Mississippi River (MMR and LMR). Levee site parameters were evaluated using geologic maps, soil survey maps, digital elevation models (DEMs), historic river maps, dredging reports, and field evaluations. Two statistical models were created for each river reach, one with a conservative ≥ 95% significance threshold, and another with a relaxed significance threshold (≥ 80% for the MMR, and ≥ 90% for the LMR). The conservative MMR model identified the presence/absence of channel fills as a significant and primary predictor of levee failures. The liberal MMR model identified several additional significant controls: location on a meander bend, channel width, land-cover type, and intensity of dredging. The conservative LMR model identified location of failure on a meander bend, land-cover type, channel constriction factor, and sinuosity of the channel as significant, while the liberal LMR model added: the presence of borrow pits, channel width, vegetative buffer width, and presence/absence of bank revetment. The results presented here could potentially assist engineers and decision-makers in choosing appropriate locations and designs for levees in the future.