Conclusion

Landslides occur when masses of rock, earth or debris move down a slope.  Mudslides, debris flows or mudflows, are common type of fast-moving landslides that tend to flow in channels. These are caused by disturbances in the natural stability of a slope, which are triggered by high intensity rains. Mudslides usually begin on steep slopes and develop when water rapidly collects in the ground and results in a surge of water-soaked rock, earth and debris. Causes may be of two kinds: 1. Preparatory causes & 2: Triggering causes. Preparatory causes are factors which have made the slope potentially unstable. The triggering cause is the single event that finally initiated the landslide. Thus, causes combine to make a slope vulnerable to failure, and the trigger finally initiates the movement. Thus a landslide is a complex dynamic system. An individual ‘landslide’ characteristically involves many different processes operating together, often with differing intensity during successive years.

The primary criteria that influence landslides are precipitation intensity, slope, soil type, elevation, vegetation, and land cover type. The present study demonstrated the effectiveness of two pattern recognition techniques: Genetic Algorithm for Rule-set Prediction and Support Vector Machine. The landslide hazard prediction study conducted in Uttara Kannada and Kerala have shown that small datasets can yield landslide susceptibility maps of significant predictive power. The efficiency of the model has been demonstrated by the successful validation. However, when the predicted features may have different immediate causes, one should carefully avoid including triggering factors among the predictor variables since they restrict the scope of the prediction map and convey often a poorly constrained time dimension. Beyond the sample size, the reliability of the susceptibility map fundamentally depends on a good knowledge of which environmental variables act to determine landslides, and on the availability and the quality of the data. To be reliable or simply useful, the prediction map must also be appropriately validated. The analysis showed that SVM applied on precipitation data of the wettest month with 96% accuracy was close to reality for Uttara Kannada district and GARP applied on precipitation data of the wettest quarter was more successful in identifying the landslide prone areas in Kerala.