Machine Learning For Earth System Sciences

ABOUT THE COURSE:This course will start with a gentle recapitulation of relevant concepts of spatio-temporal statistics and data mining, following which it will take up the topics of earth system observations, earth system data analytics and earth system modeling in various domains, such as hydrology, climate and soil.INTENDED AUDIENCE:Final year undergraduate, Postgraduate and research studentsPREREQUISITES:Machine Learning (mandatory), Deep Learning (optional), a working idea of one or two domains in earth system sciences

Week 1: Recap of probability, spatio-temporal statistics (autoregression, geostatistical equation, Gaussian Processes, Extreme value statistics) Week 2:Recap of relevant Machine Learning and Deep Learning techniques (Bayesian Networks, CNN, RNN/LSTM, VaE, Interpretability, Causality) Week 3:Earth System Process Understanding: case studies (predictors of monsoon, extreme weather forecasting, climate change visualization) Week 4:Earth System Process Understanding: case studies(Extreme event analysis, networks and teleconnections, causal analysis) Week 5:Earth System Process Understanding: case studies(Extreme event analysis, networks and teleconnections, causal analysis) Week 6:Earth System Process Understanding: case studies(Extreme event analysis, networks and teleconnections, causal analysis) Week 7:Earth System Modeling: relevant concepts (Model structures, modeling challenges, model validation, data assimilation) Week 8:Earth System Modeling: applications in different domains (ML-based surrogate models, deep and shallow generators, long-term forecasting)