Fuzzy knowledge based GIS for zonation of landslide susceptibility

Natural phenomena do not evolve linearly. The rapidly changing dynamics of eco-processes require modeling through non-linear concepts such as non-discreet fuzzy theory. The classification of data has been done using maximum likelihood classifier and Gaussian distribution. To better capture an event such as landslide it is imperative that the inherent continuity of the causative factors of the phenomenon are well represented. And that the uncertainties involved in assigning the probabilities of occurrences are taken care of. Decision making depends upon multiple parameters and their collective evaluation. Hence multi-criteria evaluation (MCE) of fuzzy variables has been implemented. A fuzzy knowledge-based geographical information system to categorize a given region into five levels of landslide susceptibility very high, high, moderate, low, and very low, has been presented in this chapter. The approach has used membership function for causative factors of landslide. This makes the system take into account uncertainty of the contributory factors of landslide. Due weightage has been assigned to the different spatial variations of these causative factors stored in the knowledge base. The input to the system consists of satellite data and topographic maps, as layers of information for the causative factors. The system carries out decision making by using multi-criteria evaluation of each input layer for adjusting its suitability, defined by the membership values and weights towards landslide susceptibility. The test conducted shows that the highly susceptible region extends along the zone of high rock mass thrust and slope failure and occupies 20-25% of the total study area. The very high threat occupies around 9-12% of the study area also falling on the higher slopes. On comparison of the susceptibility map with ground data, it has been found that three existing slides appear in the very high and high threat susceptible zones delineated by the system.