Microclimate modeling of urban residential courtyard geometry and outdoor thermal comfort in a mediterranean climate using multiple machine-learning approaches

Courtyards, characterized as semi-enclosed and centralized outdoor spaces, serve as essential design elements in indigenous urban landscapes across the globe. The thermal conditions prevalent in these areas significantly impact their quality and functionality. This study aims to predict how the geometric parameters of urban courtyards influence outdoor thermal comfort for users. Climatic parameters were measured to validate the software utilized in this research. Subsequently, thermal comfort simulations based on Universal Thermal Climate Index (UTCI) were conducted for eight urban courtyard samples, each analyzed under ten distinct geometric configuration scenarios across winter and summer seasons, employing Ladybug and Honeybee plugins within the Grasshopper framework. Additionally, three supervised machine learning models were developed to evaluate the significance of geometric parameters. The model exhibiting the highest predictive accuracy, a decision tree, elucidated the importance and optimal ranges of geometric parameters affecting changes in UTCI temperature. Analysis indicates that wide yards with a height-to-width ratio (H/W) of less than 0.5 and vegetation coverage between 15% and 32% demonstrate the highest levels of thermal comfort. Conversely, deep yards with a H/W ratio exceeding 1.5 and vegetation cover ranging from 15% to 45% exhibit the lowest levels of thermal comfort during the summer months. Two optimal configurations were identified for both seasons: one characterized by low albedo (0.15-0.20), H/W of 0.5, and 18% vegetation cover; the other with medium albedo (0.20-0.30), H/W of 0.5, and 25% vegetation cover. These findings provide insights for urban designers and planners to improve the microclimate of urban courtyards through effective geometric strategies.