Risk assessment for dangerous sections of the levees: A case study in Guangdong Province, China

Frequently happened flood disasters often cause heavy losses to the national economy and people's livelihood. The levee is an effective engineering measure to resist flood disasters and reduce disaster losses. However, the state of levees may become dangerous with the passage of time and the change of water and sediment conditions. The analytic hierarchy process (AHP) and back propagation neural network (BP-NN) were applied in this study based on field investigation to evaluate the risk state of the collapse-type dangerous sections of levee (CDSL). The results indicated that both methods could be used to assess the states of CDSLs, and the main influencing factors were levee material and nearshore water depth. In addition, two methods could complement each other at some extent, many CDSLs that estimated inaccuracy in AHP could achieve better results in BP-NN and vice versa. This study could serve as a baseline reference for future relevant researches.