Systematic review analyzes machine learning applications for urban heat island modelling, highlighting methodological gaps and the need for robust models.
Key Points
This review aims to analyze the application of machine learning and deep learning in urban heat island modelling, focusing on methodological approaches and gaps.
Systematic literature review of 85 peer-reviewed studies from June 2019 to December 2024.
Analysis structured around four research questions related to model typology and evaluation.
Focus on model-centric perspective to explore ML and DL applications in UHI processes.
Static spatial snapshot models are predominant, particularly for surface-level classification and mapping tasks.
Spatio-temporal and hybrid deep learning architectures are emerging for short-term forecasting.
Only nine studies employed physics-informed approaches and the use of explainable AI techniques was limited.