SELF-SUPERVISED WEATHER DATA CLUSTERING FOR INSPECTING LOCAL CLIMATE CHANGE

As global warming worsens, we observe an increase in the frequency, duration, and intensity of weather phenomena such as heavy rainfall, heatwaves, and droughts. This project addresses the need for adaptable models by examining these impacts through local climate data, gathering evidence of shifts in climate patterns over time. To conduct the analysis, a self-supervised technique called Swapping Assignments between multiple Views (SwAV) is used alongside ERA 5 Reanalysis data, focused on an area in Brazil characterized by high population density and extensive agricultural land. The methodology has shown promising results, creating meaningful clusters based on temperature and precipitation in the selected region, which could enhance models or applications reliant on weather data.