A methodological framework for extreme climate risk assessment integrating satellite and location based data sets in intelligent systems

We present a methodological framework to integrate the satellite as well as location based data sets to estimate extreme climate risk. The framework, in detail, has been demonstration using a study carried out to quantify extreme rainfall risks in India incorporating the influence of global (large scale oscillations) as well as local factors (population, infrastructure, economic activity) in a probabilistic model.

We use nonstationary extreme value theory along with Bayesian uncertainty analysis to model the time varying influence of oscillations such as El Nino/Southern Oscillation, Indian Ocean Dipole, and North Atlantic Oscillation in augmenting high rainfall risks in 637 districts across 29 states of India. It is found that at least 50% of the districts in 8 out of 29 states are at high risk. Extreme risk is observed in 198 (similar to 31%) and 249 (similar to 39%) districts caused by heavy downpour and extremely long wet spells, respectively.