AI-driven morphological optimization for resilient retrofitting of mediterranean housing stock under compounded urban heat island and heatwave effects
Alireza Karimi, David Moreno-Rangel, Antonio García Martínez
Abstract
• Compounded Urban Heat Island and heatwaves increase mean air temperature by ∼3.5°C by 2100. • Building morphology accounts for a 27% EUI gap under extreme heat stress. • 45% of CTE-compliant retrofits projected to fail regulatory cooling limits by 2100. • ML-GA optimization reduces EUI by up to 18% and carbon intensity by 22%. • A validated "form-first" framework identifies resilient Mediterranean housing design. The compounded effects of urban heatwaves (HW) and the urban heat island (UHI) phenomenon threaten the thermal resilience of Mediterranean residential buildings; yet, few studies evaluate building performance under localized future climate conditions. This study addresses this gap by combining high-resolution climate projections (EURO-CORDEX) with UHI adjustments to generate localized weather files for Madrid across historical (2006–2020), mid-term (2040–2060), and long-term (2080–2100) periods under RCP8.5. Twenty residential building archetypes were simulated in EnergyPlus to evaluate energy use intensity (EUI) and operational carbon intensity (OCI) during peak thermal stress events. Key findings show that mean air temperature during HW events may increase by ∼3.5°C by 2100, potentially doubling cooling energy demand. The results identify a significant "resilience gap": compact, well-oriented buildings achieved up to 27% lower EUI compared to less resilient typologies under identical code-compliant upgrades. Geometry-optimized designs identified via machine learning and genetic algorithms reduced EUI by up to 18% and OCI by up to 22% relative to baseline designs. Notably, up to 45% of today’s retrofitted buildings are projected to fail compliance with future CTE cooling limits, highlighting a critical need for regulatory reform. This study demonstrates the combined value of localized climate data and generative AI to inform "form-first" resilience strategies. We propose a three-tier reform for building codes, incorporating morphological resilience coefficients and AI-assisted compliance probes to ensure Mediterranean housing remains habitable and sustainable under 21st-century climatic extremes.