Application of GIS to landslide susceptibility modelling, Waitaki Region, South Island, New Zealand

Abstract

Landslides worldwide and in New Zealand in particular have the potential to cause significant human and economic costs. The assessment of areas, which are prone to landsliding, is an important step in reducing those costs. The designation of areas susceptible to landsliding requires the determination of the causative factors and their interrelationships that lead to slope instability. The evaluation of the spatially varying terrain conditions as well as the spatial representation of landslides is essential for this procedure. This form of analysis has already been carried out in several former studies worldwide using a wide range of different approaches. Many of these have been undertaken with the aid of Geographic Information Systems (GIS), which provide tools for collecting, storing, manipulating and displaying spatial data. In this project a landslide susceptibility map for the Waitaki area in North Otago, New Zealand, has been developed within a GIS using a quantitative approach based on heuristic and statistical methods. The main aim is the prediction of locations that are likely to produce landslides based on the geological and geomorphological characteristics of the terrain. The spatial distribution of landslides has been correlated with topographic, geological and landscape features in order to determine which factors best explain landslide occurrences specific to this area. GIS are used to analyse and generate parameters of significance to slope stability. The first step of the project is a literature review about recent landslide susceptibility mapping techniques, followed by the identification of important input parameters to the landslide susceptibility model. The generation and assembling of the corresponding data layers within GIS forms the database of the model. For the final map construction each data layer (factor map) is assigned a weight according to the importance of the factor in the identification of landslides. Those weights are derived by the application of the statistical method 'weights of evidence'. The weighted data layers can eventually be combined to the final landslide susceptibility map, representing the posterior probability of landslide occurrences. Different combinations of those factor maps are tested by the development of a success rate. The final model includes conditions relating to rock type, dip angle, slope aspect, slope steepness, distance to streams, stream order of the nearest stream, elevation difference to the nearest stream, and distance to active faults. The model is built using the northern part of the area, whereas the validation of the best matching combination of factors maps is done in form of a prediction rate, on the basis of the existing landslides in the southern part. This prediction rate shows a satisfactory result, having 85% of the existing landslides in the first half of the higher susceptibility classes.