Machine Learning-Based Forecasting of Wet-Bulb Temperature and Two-Decade Urban Climate Shifts

Rapid urbanization and climate change have exacerbated heat stress in metropolitan regions like Delhi, India. This study investigates the spatio-temporal dynamics of Wet-Bulb Temperature (WBT) and Land Surface Temperature (LST) from 2005 to 2024, and projects future WBT trends using a Long Short-Term Memory (LSTM) model. The novelty of this research lies in integrating satellite-based climate data with machine learning algorithms for early warning systems and urban resilience planning. The study reveals a significant rise in WBT over the past two decades, with projections indicating values exceeding 35 °C during extreme heat events by 2030, especially in densely built-up zones. Using LANDSAT imagery and urban expansion data, a strong positive correlation was observed between urbanization and elevated WBT levels. A Pearson correlation analysis revealed prolonged thermal stress through the strong associations between May LST values from recent years, particularly between the years 2023 and 2024 (r = 0.74) and 2015 and 2023 (r = 0.71). The thermal patterns experienced change because of increased Urban Heat Island (UHI) effects, as shown in the early-year correlations between 2005 and 2021 (r = 0.20). Spatial measurements verified that populated city centers recorded elevated WBT and LST data, which reflect the impact of urbanization and damaged vegetated areas. Real-time monitoring combined with ML-driven alerts and sustainable planning interventions needs immediate implementation, according to WBT forecasting results produced by the LSTM model. The study provides critical insights for policymakers to formulate evidence-based heat mitigation strategies, aiming to safeguard public health and labor productivity under future climate scenarios.