Landslides predictIon based on neural network and remote sensing data

Abstract

Landslide is one of the main natural disaster that Sri Lankan intermediate zone faced. In most of the cases, property damage and vulnerability of people significantly high compared to the other natural disasters. This mainly occurs due to poor disaster forecasting methodologies, lack of early warning systems and preparedness practices. Therefore there is a vital need of implementing a model for landslides prediction. In this research, it supposed to introduce landslide forecasting model based on remote sensing methodologies and Artificial Neural Network (ANN). Landslide forecasting modelling has long history in many countries and most of the scenarios it was based on the trend line analysis, and liner and non-liner regression analytical methods and models. Remote sensing methodologies and technics are doing significant impact on landslides disaster evaluation and mitigation process in multiple sectors. Normalized Difference Vegetation Index (NDVI) has shown powerful calculation remote sensing techniques to identify vegetation, soil and build-up areas and significant variations of them through the raster calculation methods. Hence, these factors were rarely use in existing landslide forecasting models. This research is identified the significance of these calculations on landslide forecasting using ANN model with NDVI. The model was evaluated using a baseline model. The results of the model expose ability to increase the accuracy than existing landslides forecasting models.