Who: Dr. Isha Savani
Affiliation: Isha Savani is a Research Associate at the Institute of Oceanography and Applied Geophysics (OGS) in Trieste, Italy. She is currently based in Oslo, collaborating with the environmental institute NILU and the Department of Informatics at the University of Oslo. Her research lies at the intersection of water law and atmospheric dynamics.
When: June 17th, between 13:00 and 14:00 CET (Norway time)
Where: https://uio.zoom.us/j/69096081290?pwd=bg1XKTMKjweeHqxM77VmAfbOLPD0o6.1
Title: Self Organizing Maps as Method to Identify Patterns in the Sky
Abstract
A steady increase in the magnitude of rainfall in monsoon systems has been observed across the globe. This is accompanied by an increase in extreme weather (droughts and floods), for instance in the Horn of Africa and in the Indian sub-continent. The devastating flood in 2022 in Pakistan shifted the focus of the science community at large towards the region, to better understand the relative impact of global trends (such as increase in land sea temperature and change in the direction of wind circulations) and local climate variability. The key question being asked here is “can we identify characteristic patterns (at different spatial scales) that can explain the particular extreme event, and can this help with climate forecasting”?
In this project, we aim to further contribute to the understanding by employing Self Organising Maps. Pioneered by the Finnish scientist Teuvo Kohonen in 1981, this neural network technique maps higher dimensional data to regular lower dimensional grid representing patterns. To test out the algorithm, we implement Self Organising Maps over the Indian Ocean to identify characteristic features in the distributions of Pressure and Temperature, using data from the last 100 years. The goal is to identify emerging climate patterns that could explain trends in precipitation over the study area. In this talk, I will expand on the machine learning algorithm and will share preliminary results.