AbstractsGeography &GIS

Developing Tools for Earthquake-induced Landslide Hazard Maps of the Island of Hawaiʻi

by Shailesh Arun Namekar

Institution: University of Hawaii – Manoa
Year: 2016
Keywords: earthquake-induced landslide hazard maps; empirical models
Posted: 02/05/2017
Record ID: 2064136
Full text PDF: http://hdl.handle.net/10125/100731


Ph.D. University of Hawaii at Manoa 2013. The purpose of this study is to develop earthquake-induced landslide hazard maps for the Island of Hawaiʻi using various tools such as Geographical Information Systems (GIS), Artificial Intelligence (AI), and Logistic Regression (LR). The methodology for the current research consists of developing empirical models based on factors considered to be most influential on landslide susceptibility, and analytical models based on conventional slope stability analysis. Earthquake-induced landslide hazard maps were then developed using these models for the Island of Hawaiʻi. Empirical models involve systematically studying the landslide contributing factors for the entire island without using any slope stability model. This can be applied to the entire island with generally available island-wide spatial data. Empirical models involve the use of various techniques such as weight analysis, logistic regression and AI based tool such as ANFIS (Adaptive Neuro Fuzzy Inference System). In the weight analysis, different landslide contributing factors were grouped according to their relative importance. These weights were obtained and refined by an Artificial Neural Network (ANN). The map of the Island of Hawaiʻi was divided into cells of size 100m * 100m. Each landslide contributing factor was classified into a common evaluation scale of 1 to 10. Then, landslide contributing factors were multiplied by their respective weights. Finally, the results of the calculations (i.e. final rating values) were obtained. These final rating values were plotted with a color scheme in a map with ten zones of landslide vulnerability (viz., Red (High) (10) to Green (Low) (1). This map showed that weight analysis can locally yield 'false positive results. Therefore, two empirical models were developed based on ANFIS and Logistic Regression. In LR-and ANFIS-based models, probability of failure was predicted using landslide contributing factors as inputs. Hazard status values, 0 (Low hazard, i.e. low probability of failure) to 1(High Hazard, i.e. high probability of failure), were derived. Based on these hazard status values, the potential hazard maps were developed. These potential hazard maps provided an overview of potential landslide hazard zones on the Island of Hawaiʻi where more detailed study may be warranted. Though empirical models developed here do not consider seismic analysis, they give a bigger picture of landslide hazard zoning (susceptibility) for the entire island. The maps developed using empirical models can help in selecting regions where further study on slope stability is necessary. For these selected regions, slope stability analysis was carried out with new empirical models developed using explanatory variables from an analytical slope-stability model. In latter models, the general approach to the landslide zoning method is based on conventional slope stability analyses to determine the Factor of Safety (FS) and critical ground (yield) acceleration (ay) of the individual slopes…