Air temperature estimation and modeling using data driven techniques based on best subset regression model in Egypt

Introduction

Accurate temperature forecasting is crucial for understanding future climate patterns¹. Climate change and global warming pose significant global challenges, intensifying water-related disasters such as floods and droughts while affecting water quality². A robust understanding of temperature variations aids decision-makers in mitigating climate change impacts and enhancing infrastructure resilience and sustainability³. While temperature is a critical variable connecting atmospheric and land surface processes in studies of hydrological, ecological, and climate change. Numerous studies have been conducted to model air temperature^{4, 5, 6–7}consistently emphasizing the importance of accurate air temperature estimation in the fields of meteorology, hydrology, and agro-hydrology. Consequently, more robust and accurate models are required to effectively capture the nonlinear dynamics of air temperature variation⁸.

Goodale et al.⁹ employed geographical information such as latitude, longitude, and altitude as predictors for interpolating temperature and precipitation in Ireland. Ninyerola et al.¹⁰ used geographic information systems (GIS) to simulate and map air temperature. Satisfactory forecasting of nonlinear air temperature behaviour in time and space is critical. Reicosky et al.¹¹ evaluated methods for estimating hourly air temperatures from daily maxima and minima, achieving reasonable accuracy on clear days but poor performance on overcast days. Their study recommended direct hourly temperature measurements for precise modeling but did not assess the impact of estimation errors. Sadler and Schroll¹² expanded this research by developing an algorithm that did not rely on predefined temperature curves, outperforming existing methods in nearly half of the cases. However, its application required constructing a site-specific normalized temperature cumulative distribution function for a full year, limiting its practicality for broader use.

The urban heat island effect^{13, 14–15}air pollution^{16, 17–18}and human mortality¹⁹ are all closely associated with air temperature measured at 2 m above ground level in urban environments^{20, 21–22}. In high density populated areas, monitoring and forecasting of maximum (T_max) and minimum (T_min) air temperatures is essential due to their association with extreme events, such as heatwaves and tropical nights^23,24. With a large population and complex infrastructure, even minor temperature variations within a city can significantly influence both human and natural environments^25,26. Consequently, understanding and monitoring the spatiotemporal...