00000ctm a22000004a 4500 UP-99796217608617116 Buklod 20221007163550.0 a grb 001 u| ta 221007s2008 xx r |||| u| | (iLib)UPD-00064996168 DENG - eng DMLUC LG 995 2008 E685 B36 Bañgate, Julius M. Developing a geoexpert system using object-based image analysis methods for the delineation of landslide hazard zones on remote sensing imagery Julius M. Bañgate. 2008 xvii, 239 leaves col ill., map "July 2008" Thesis (M.S. Remote Sensing)--University of the Philippines, Diliman. Landslides are widespread extreme natural hazards that the Philippines, being one of the most disaster prone countries in the world, experience very often. There is a need to determine landslide hazard zones and prepare the inhabitants in averting the dire consequences. This study uses remote sensing (RS) imagery and geographic information systems (GIS) to identify landslide scarps and use these together with other morphometric and topographic parameters to determine landslide prone areas. A data mining and expert system approach is utilized as the main framework integrating RS, GIS and Artificial Intelligence (AI) techniques in one application platform. An object based strategy is implemented in classifying remotely sensed images; this involves segmentation using Normalized Difference Vegetation Index (NDVI), attribution using domain knowledge generated from a statistical approach, and heuristic from the developed domain knowledge of landslide crowns. A landslide Susceptibility Index (LSI) was developed for 17 parameters using a rating scheme based on frequencies of landslide occurrence and regression analysis. Bivariate regression analysis reveals slope length factor and slope (degrees) to be the most significant parameters. A high R squared value of 0.98 was achieved in the creation of the LSI model using multivariate regression analysis. A Landslide Hazard Index (LHI) was also developed to account for the contribution of rainfall and earthquake events as triggering mechanisms. By programming in AVENUE®, an automated recognition of landslide from imagery was developed from heuristics involving the 17 parameters, LSI, LHI, ASTER image band values, landslide runout distance, Change Vector Analysis (CVA) and NDVI. Heuristics for the classification of other cover types were developed as a test for robustness of the landslide heuristics that correctly classified 71.53% of 138 landslide polygons. Outputs of this study include domain knowledge on landslides; landslide susceptibility and hazard maps; and geoExpert system. These can be useful in landslide monitoring and mitigation efforts. Also, the developed application framework is flexible and can be used in the characterization and analysis of other natural spatial phenomen Landslides Remote sensin Philippines. Landslides Remote sensing map Philippines. Landslides Geographic information system Philippines. Remote sensing Environmental aspect Philippines. Landslide hazard analysis Philippines. FI UP Thesis UPD DARCHIVES LG 995 2008 E685 B36 UPD DENG LG 995 2008 E685 B36 Thesis