Environmental performance-oriented computational design for urban street renewal: a case study of Istmo de Ferreira do Amaral, Macau, China

Urban street renewal improves environmental quality in dense cities, but most existing studies rely on oversimplifiedand idealized models that overlook complex geometries and rarely integrate outdoor and indoor performance. To address these gaps, we develop anenvironmental performance-oriented computational design framework combining a realistic parametric urban model (PM-UF&BG) with environmental simulations, multi-objective optimization, and an Artificial Neural Network surrogate with SHAP analysis. Applied to 5,000 morphological variants within the case of Macau’s historic Qilou fabric, the workflow minimizes street and façade solar radiation while maximizing ground-floor interior spatial Daylight Autonomy (sDA). An equal-weighted TOPSIS analysis isolated a top-performing variant that increased sDA to 48.0 % while lowering façade and street radiation to 168 kWh/m2 and 222 kWh/m2. Post-optimization refinements based on the top 1 TOPSIS design revealed non-monotonic trade-offs between annual cooling energy and sDA of upper-floors, and window-to-wall ratios (WWR), supporting performance-driven façade design. Furthermore, SHAP analysis deconstructed the surrogate model to uncover non-linear mechanics, isolating a “medium-projection exposure paradox” at intermediate arcade depths (1–2 m) and a “downstream shading anomaly” from vertical expansions. This transferable workflow bridges data science and architectural practice to extract context-sensitive geometric guidelines during early-stage master planning.