Long-term dynamics of urban heat island and hot spots in Wrocław: A 25-year satellite-based analysis using machine learning

Abstract

Urban areas are increasingly vulnerable to rising temperatures due to rapid land cover transformation, which intensifies the Urban Heat Island (UHI) effect, promotes the emergence of thermal hot spots, and contributes to growing thermal discomfort. Understanding long-term urban thermal dynamics is crucial for climate-responsive urban planning. While many existing studies rely on short-term analyses, single-source thermal indicators, and a single Land Use/Land Cover (LULC) classification method, this study offers a 25-year assessment (1999–2023) of UHI patterns and thermal hot spots in Wrocław, one of Poland’s largest and most urbanized cities, using an optimal LULC classification technique and an integrated composite thermal index. Landsat satellite imagery was used to derive LULC classifications and Land Surface Temperature (LST) maps. To ensure classification accuracy, four methods were evaluated: Maximum Likelihood (MaxL), Minimum Distance (MD), Support Vector Machine (SVM), and Artificial Neural Network (ANN), with ANN yielding the highest performance. A key innovation of this study is the integration of UHI and Urban Thermal Field Variance Index (UTFVI) maps to generate composite thermal stress maps, which reveal more nuanced spatial patterns of urban heat exposure than single-indicator approaches. Results indicate that 32.9% of the number of identified thermal hot spots emerged after 2010, 21.7% disappeared, and 45.4% remained stable throughout the study period. These findings underscore the critical influence of LULC changes on urban thermal environments and provide valuable insights for sustainable urban development and environmental policy-making.