Author(s): Srikanth Bhoopathi; Manali Pal
Linked Author(s): Manali Pal
Keywords: Machine Learning XGBoost MLR Heatwave Daily Maximum temperature
Abstract: This research aims to develop predictive models for forecasting maximum temperature for predicting heatwaves using Machine Learning (ML) techniques, specifically eXtreme Gradient Boosting (XGBoost) and Multiple Linear Regression (MLR). The study relies on five essential meteorological variables: air temperature (AT), geopotential height (HGT), relative humidity (RH), U-wind (UW), and V-wind (VW), spanning the years from 1991 to 2020). The ML models are constructed using spatially averaged atmospheric variables as predictors and IMD daily maximum temperature as the target variable. Notably, this study achieves a reasonable level of accuracy in predicting maximum temperatures for the study area with a 7-day lead time. However, as the lead time extends to 15 days, there is a noticeable decrease in model performance. It is worth mentioning that XGBoost outperforms MLR in the prediction of maximum temperatures. These results suggest the potential of utilizing spatio-temporal dynamics in meteorological variables for long term prediction of high temperatures. Importantly, both XGBoost and MLR demonstrate significant promise for consistent application within this particular framework.
Year: 2025