A MACHINE LEARNING DRIVEN MATHEMATICAL FRAMEWORK FOR PREDICTING STATION LEVEL PRESSURE AND TEMPERATURE

Abstract

Atmospheric pressure and ambient temperature are fundamental meteorological variables that influence atmospheric circulation, wind formation, precipitation processes, and extreme weather events. Accurate station-level prediction of these parameters is essential for improving short-range weather forecasting, climate diagnostics, and operational decision-making in sectors such as aviation, agriculture, and disaster management. Traditional numerical weather prediction (NWP) models, although physically comprehensive, require intensive computational resources and often struggle to resolve localized interactions over regions with complex terrain or high spatiotemporal variability. In this study, we propose a machine learning driven mathematical framework for predicting station-level pressure and temperature using the Random Forest Regressor. The model is formulated to capture nonlinear dependencies among multiple meteorological predictors without relying solely on explicit physical parameterizations. A feature importance analysis is incorporated to identify the dominant atmospheric variables and to provide interpretability of the learned relationships. The mathematical formulation includes ensemble-based regression, impurity-reduction metrics, and error-evaluation functions to quantify predictive skill. Results demonstrate that the proposed framework achieves high predictive accuracy while maintaining computational efficiency. The integration of machine learning with atmospheric data enhances the representation of localized weather behaviour and provides insights into key driving variables. This work underscores the potential of advanced ML-based mathematical models as complementary tools to traditional forecasting systems, offering scalable and reliable solutions for station-level weather prediction.