Hydrological extremes are globally important natural hazards. The increase in the number of events in the last few decades has motivated the research of the spatiotemporal variability of the future extreme events like floods and drought. Data related to extreme weather events in India between January 1 and September 30 2022, as maintained by the IMD and the Disaster Management Division (DMD) under the Union Ministry of Home Affairs revealed that as many as 2,755 people lost their lives to such extreme weather events across India with Himachal Pradesh saw the highest number of human deaths, at 359. Overall, extreme weather events affected 1.8 million hectares of crop area, destroyed over 4 lakh houses and killed almost 70,000 livestock.

One of the main challenges in predicting and managing floods and droughts is the complexity and variability of the factors that contribute to these extreme events. These factors include meteorological, hydrological, and climatological processes that operate at different scales, including regional and global. In addition, these processes are often influenced by human activities such as land use and resource management, which can further increase the uncertainty of predictions.

Given the complexity of the processes involved and the potential consequences of these extreme events, it is essential to develop improved prediction and management models that can provide more accurate and timely information about the likelihood and impacts of floods and droughts. 

One common approach for predicting hydroclimatological extremes is the use of coupled atmospheric-hydrological modelling, which involves simulating simulation of atmospheric processes such as temperature, pressure, and humidity along with the movement and behaviour of water on and below the surface. This can include the prediction of river flow, groundwater recharge, and the impacts of land use and resource management on water availability.

By combining these traditional process-based modelling approaches with AI and deep learning techniques, it is possible to improve the accuracy and reliability of predictions about the occurrence and impacts of floods and droughts. This can ultimately help to better inform decision-making and risk management strategies, and reduce the negative impacts of these extreme events on communities and critical infrastructure.

 Research Objectives

1. Development of reliable atmospheric model, hydrological model and hydrodynamic model parameterization for accurate simulations of hydroclimatic fluxes.
2. To develop a robust hydrologic decision support system for predicting and managing extreme events using novel frameworks which use combinations of process based modelling and deep learning.
3. To Analyse the long term impacts of climate change on local Himalayan communities and development of better adaptation strategies that better suit local conditions.

Methodology

Our methodological spectrum comprises numerical and stochastic simulations in combination with AI based models , in particular coupled models, to quantify the process interactions. An example is the coupling of weather forecasting models with catchment models and hydrodynamic models for simulating soil storage, snow storage and the water fluxes in rivers and inundation areas. Recently, explainable AI based models and deep learning based architecture have emerged as very powerful tools to help in generating predictions and classifications. Especially in hydroclimatological sciences, a number of studies have been published to showcase the utility of these tools. However, AI based models are solely dependent on data and don't consider physics into consideration. In our proposed framework we plan to use process based simulation in combination with AI based models to develop innovative approaches for identifying patterns in large data sets.

Examples are non-stationary extreme value statistics and non-linear methods such as Bayesian Networks or Random Forests. Deep Learning capabilities aided with sound computational power will be explored for development of robust prediction models for the extreme events like floods and droughts. The joint analysis of a large number of catchments in large regions allows understanding the genesis of hydrological extremes and their changes across different space-time-scales.