Kamal S.A.

The Use of a Distributed Hydrologic Model to Predict Dynamic Landslide Susceptibility for a Humid Basin in Puerto Rico

Abstract: 
This thesis describes the use of a distributed hydrology model in conjunction with a Factor of Safety (FS) algorithm to predict dynamic landslide susceptibility for a humid basin in Puerto Rico. The Mameyes basin, located in the Luquillo Experimental Forest in Puerto Rico, was selected for modeling based on the rich ensemble of soil, vegetation, topographical, meteorological and historic landslide data available. The basin was parameterized into the TIN-based Real-time Integrated Basin Simulator (tRIBS) with particular emphasis on vegetation parameters for broadleaf evergreen trees in tropical climates. The basin was forced with precipitation data that included a synthesized rainfall event likely to result in a landslide based on rainfall intensity-duration thresholds. The basin’s response was assessed mainly in terms of soil moisture and values of selected vegetation parameters, which served as the dynamic inputs into the FS algorithm. An off-line FS algorithm was developed and tested using typical values for parameters encountered in the Mameyes basin. Sensitivity analyses indicated that slope angle, soil cohesion and soil moisture were the most sensitive parameters in this FS algorithm. When the tRIBS / FS Algorithm combination was employed over the entire basin, landslides were indicated in 48 out of 13,169 modeled locations. The spatial distribution of landslides compared favorably to a static landslide susceptibility map developed in previous work by Lepore et al. (2008b) while the temporal distribution of landslides was correlated with rainfall events. Landslides were predicted over a range of slope angle values, including on relatively gentle slopes where the modeled soil moisture drove the instability. The results demonstrate that the tRIBS/FS algorithm combination developed in this work is able to capture the key dynamics associated with slope stability, specifically the interactions between the slope angle and the soil moisture state.

Rainfall-induced landslide susceptibility zonation of Puerto Rico.

Lepore, C. Kamal, S. A., Shanahan, P. Bras, R. L., Rainfall-induced landslide susceptibility zonation of Puerto Rico. Environmental Earth Science, 2011. DOI 10.1007/s12665-011-0976-1

Abstract: 
Landslides are a major geologic hazard with estimated tens of deaths and $1–2 billion in economic losses per year in the US alone. The island of Puerto Rico experiences one or two large events per year, often triggered in steeply sloped areas by prolonged and heavy rainfall. Identifying areas susceptible to landslides thus has great potential value for Puerto Rico and would allow better management of its territory. Landslide susceptibility zonation (LSZ) procedures identify areas prone to failure based on the characteristics of past events. LSZs are here developed based on two widely applied methodologies: bivariate frequency ratio (FR method) and logistic regression (LR method). With these methodologies, the correlations among eight possible landslide-inducing factors over the island have been investigated in detail. Both methodologies indicate aspect, slope, elevation, geological discontinuities, and geology as highly significant landslide-inducing factors, together with land-cover for the FR method and distance from road for the LR method. The LR method is grounded in rigorous statistical testing and model building but did not improve results over the simpler FR method. Accordingly, the FR method has been selected to generate a landslide susceptibility map for Puerto Rico. The landslide susceptibility predictions were tested against previous landslide analyses and other landslide inventories. This independent evaluation demonstrated that the two methods are consistent with landslide susceptibility zonation from those earlier studies and showed this analysis to have resulted in a robust and verifiable landslide susceptibility zonation map for the whole island of Puerto Rico.
Syndicate content