**4. Discussions**

The rainfall thresholds defined in this study establish a minimum cut off below which chances of occurrence of rainfall is very low. Above these thresholds, the probability of occurrence increases exponentially, but still the chance of false alarms cannot be neglected. Even though rainfall is the major triggering factor, other physical factors also influence the stability of slopes. For a powerful Landslide Early Warning System to work effectively, parameters like soil moisture and soil movement/tilt etc. should be incorporated along with the rainfall thresholds. An integrated system with multiple sensors

and rain gauges can be installed in the region for this purpose. Similar researches have been carried out for the Darjeeling Himalayas [5] using Micro Electrical Mechanical System (MEMS) tilt sensors. A network of such sensors can effectively transfer the data to the authorities in real time which can be used as an effective warning system. The frequency of available rainfall data is the key factor which determines the accuracy of thresholds. In the current scenario, the temporal resolution of rainfall data available for the region is one day, and for an area of 4358 km2 only five rain gauge stations (as of 2019) are available. By establishing a network of sensors across the district, the spatial and temporal resolution of rainfall measurements can also be improved.

Several rainfall thresholds have been developed and periodically updated [43] for forecasting landslide events across the globe. Choosing the best method for establishing rainfall thresholds for a particular region requires detailed analysis and a quantitative comparison using statistical attributes [44]. Simple empirical models can also be modified conceptually by incorporating physical or hydrological parameters to improve the prediction power [45,46]. Further studies must be conducted for the area using existing models which are being practiced in different parts of the world [11,47–50] and the best suited rainfall threshold should be integrated with a sensor network and rainfall forecasting system. This research is a humble step towards achieving the goal of establishing an effective Landslide Early Warning System, which can minimize the casualties due to landslide hazards in the Idukki district.
