ICTS

Earth System Models are extremely useful to understand the Earth’s future in the context of climate change under different possible scenarios of anthropogenic emissions and socio-economic policies. These ESMs are simplified models of various aspects of the Earth System, including climate, hydrology, ecology, biogeochemistry, and human activities at global scales, with potential applications to  local/regional dynamics. ESMs are essential  for simulating the carbon cycle and vital environmental conditions that play a crucial role in climate-ecosystem variability and change. Physics-based differential equations typically govern these models. Though the ESMs are better suited for investigating the wider consequences of climate change and understanding the interactions between climate and other Earth system processes, they are extremely computationally expensive, and often the parameterizations are inaccurate, especially for very localized processes, leading to inaccurate simulations. 

In recent years, there has been some effort to use Machine Learning (ML) to develop low-cost surrogate models for these Earth System Models and also use it to improve the parameterization of localized processes. Recent ML paradigms such as Physics-Inspired Neural Networks (PINN) and Generative Modeling can be useful to improve the ESMs. Further, Machine Learning can draw new insights about the ESM simulations using Explainable AI approaches. Weather forecasting is increasingly adopting data-driven models, powered by deep learning (DL), which are based on either Transformer based (FourCastNet, DLWP, Pangu etc) or Graph Neural Network (GraphCast). A few of these models are also being used operationally by national meteorological agencies. In addition, the rise and availability of large language models (LLMs) and their fine-tuning, has enabled remarkable outcomes in various domains and promise new directions in climate sciences too.

The aim of this workshop is to explore different aspects of the ESM  and how ML can help them to improve them and make them more accurate and efficient. The usage of state-of-the-art AI models for weather/climate forecasts too will be discussed with hands-on-workshops.

Eligibility criteria: Early career researchers (PhD within last 10 years), PhD students and industry professional are eligible to apply. Researchers/industry professionals  working on the Earth Sciences or Machine Learning will be given preference.

ICTS is committed to building an environment that is inclusive, non-discriminatory and welcoming of diverse individuals. We especially encourage the participation of women and other under-represented groups.