In recent decades, the frequency and magnitude of hydroclimate (precipitation, temperature, river flow, Arctic Sea Ice, etc.) extremes are exhibiting significant year-to-year and multi-year variability along with strong secular temporal trends across the global. They inflict destructive punches on the infrastructure, ecology and, consequently on socio-economy of countries, besides taking many human lives. Thus, to effectively mitigate the negative impacts of these hydroclimate extremes, we need to understand and model their space-time variability along with their attendant uncertainties.

With this driving motivation, our group continues to develop statistical learning models for these hydroclimate variables. Bayesian hierarchical models (BHM) and tree-based methods are the two main statistical learning modeling framework that we adapted and developed. My talk will showcase the applications of these methods to a suite of hydroclimate problems – stochastic space-time weather generation, precipitation extremes in space and time, streamflow on a river network, blending satellite and ground observations, Arctic sea-ice retreat, and paleoclimate reconstructions.

These methods enable risk-based management, planning, and adaptation for a range of sectors – infrastructure, water resources, public health, agriculture, etc., and help mitigate the risks from these extremes.