Application of Multiple Criteria Decision Making Model for Evaluation of Levee Sustainability

Abstract

In this study we seek to improve our ability to predict the sustainability of a levee by analyzing the character of the surrounding environment. Utilizing geographic information systems (GIS), approximately 140 mi (225 km) of levees within the Lower Mississippi River Valley were divided into small segments and assigned a series of environmental factors including the configuration of Quaternary Geology with respect to the levee alignment, the hydrogeological characteristics of the alluvial aquifer beneath the levee, and soil physical properties. Next, a binary logistic regression was applied to evaluate correlation between environmental factors and development of levee distress features (seepage lines and sand boils) to generate a predictive model. Results of the logistic regression were then fed into a multiple criteria decision making (MCDM) system to categorize environments into levee sustainability groups. Logistic regression results indicated significant correlation between levee distress features and four environmental characteristics: paleo channel orientation, AASHTO soil classification, normalized difference vegetation index (NDVI), and saturated hydraulic conductivity. The predictive model generated correctly predicted the status of distress feature development with up to 62 percent accuracy. The MCDM system identified forests of sweetgum, nuttal oak, and willow oak as elevated sustainability of levees. Plots of sycamore, pecan, and American elm trees and water bodies were rated as decreasing levee lifetime. With additional development, future models may serve as tools to improve our ability to assess, maintain, and design levee systems in better coherence with their natural surroundings.