GIS modelling of slope stability in Phewa Tal watershed, Nepal

Abstract

Hazards are an inherent but dangerous and costly element of mountainous environments. Conventional maps of mountain hazards provide useful inventories of hazardous sites but provide little insight into the operation of the hazards. Furthermore, this approach tends to rely heavily on subjective interpretation of the landscape, which means that the results can not be replicated or transferred to other areas. Thus, alternative approaches employing the quantitative capabilities of geographic information systems (GIS) to model and predict slope stability are receiving increasing attention. This paper reports on the use of a diverse GIS database, compiled primarily from existing maps and aerial photographs, to construct a regional model of slope stability in Phewa Tal watershed, Nepal. An integral part of the research was to explore an alternative approach to the commonly used grid cell approach by employing geomorphometrically significant terrain units. The terrain units employed were created by generating line networks representing local maxima and minima extracted from elevation and curvature surfaces. One of the chief benefits of applying GIS technology in this research was the ability to georeference all of the attribute data to these terrain units. In doing so, it allowed the database to be exported into an external statistical package, where the terrain units could be statistically explored as the basic analytical unit. The application of a variety of statistical techniques resulted in logistic regression being selected as the most useful. Logistic regression successfully predicted terrain units as being either stable or unstable at a rate of approximately 90% concordance with a conventionally produced map of slope stability. The statistical probabilities of terrain unit stability were imported back into the GIS to produce a map of predicted slope stability that compared well with the conventional map of slope stability. The findings of this research suggest that the use of geomorphometrically significant terrain units extracted from a digital elevation model (DEM) are an efficient alternative to approaches using regular grid cells. In particular, the terrain units facilitated the use of logistic regression, and significantly decreased the amount of computing costs. Finally, this research also suggests that important information can be gathered from existing information sources, such as maps, aerial photographs, and written documents, thereby limiting the need for costly and time consuming field work at the reconnaissance level. Based on this latter finding, other information sources, such as satellite imagery, should be examined.