Integrated assessment of urban flooding and heat island interactions: A systematic review of geospatial technologies, machine learning approaches, and microclimate dynamics

Urban flooding and the Urban Heat Island (UHI) impact are two crucial challenges driven by rapid urbanization and climate change. Understanding their linked dynamics and shared influencing factors is crucial for effective mitigation. However, despite extensive advancements, there are still certain gaps in incorporated analytical frameworks that signify the reciprocal spatio-temporal interplay among these phenomena, especially regarding microclimate variations and Land Use and Land Cover (LULC) changes and vice versa. This comprehensive review highlights the systematic evolution and the crucial significance of LULC, microclimate variables, and empirical and GIS-based techniques, incorporating machine learning approaches, while rigorously evaluating more than seventy-four reported scientific studies. Approximately 30-32 percent of studies employed hybrid learning models (e.g., CNN-LSTM) to examine UHI, while 60-65 percent used ensemble techniques (e.g., Random Forest, Gradient Boosting) for flood danger analysis. Deep learning is an effective tool for modelling complicated, nonlinear relationships, whereas ensemble models are valued for interpretability and multi-source information integration. However, the sector lacks scalable, models leveraging real-time IoT statistics and accounting for urban morphological and climatic variability. By synthesizing cutting-edge findings, this overview offers insights into the bodily and data-driven interlinkages between urban flooding and UHIs, supporting more resilient and sustainable city planning. Urban flooding and UHI phenomenon are critical urban climate challenges, and their interactions amplify hydrological and thermal stresses, threatening urban livability, infrastructure resilience, and public health. Despite advances in machine learning and geospatial technologies, most of the previous studies examine UHI and flooding separately, overlooking microclimatic variations, urban morphology, real-time IoT integration, and model transferability. This review synthesizes recent geospatial and AI-based approaches to assessing UHI-urban flood interactions and identifies key methodological and geographical gaps. A unified data-driven framework is proposed to simulate compound hazards and enhance urban climate resilience. Peer-reviewed studies reported during the time span of 2010-2025 were systematically reviewed using the PRISMA 2020 protocol, with literature sourced from Scopus, Web of Science, ScienceDirect, and Google Scholar. Studies were categorized by hazard type, analytical technique, and validation metrics (RMSE, R², MAE). Ensemble machine learning methods, notably Random Forest and Gradient Boosting, dominated flood modelling (61-65 %), while CNN-LSTM hybrid models were mostly used for UHI (30-32 %). CNN-LSTM performed best for moderate-resolution spatiotemporal data, whereas transformers and graph neural networks (GNNs) excelled for complex, large-scale scenarios. Key gaps include model interpretability and IoT integration. Transparent, geospatial AI frameworks are essential for data-driven, climate-resilient urban planning.