Near-surface air temperature (T_a) plays important roles in the interactions between the atmosphere and land covers, serving as crucial indicators for characterizing urban heat island and pedestrian thermal comfort. Although satellite-based observations and numerical simulations have been proposed to obtain T_a across various spatiotemporal scales, there is a lack of physically grounded, precise and easy-to-implement means to spatially resolve T_a at the centimeter level. Based on multimodal images derived from the unmanned aerial vehicle and synchronously measured meteorological parameters, this study combined the surface energy balance model and the automated machine learning to predict air temperature near the observed surfaces (T_{a_predicted}). The validations via near-ground measurements demonstrate the accuracy of this methodology, biases between predicted values and measured ones were almost maintained within 0.55 ℃. According to the spatial distribution of T_{a_predicted}, it can be inferred that, at clear and calm noon in subtropical regions, dense shrubs or lawns exhibit limited significant cooling effect on T_{a_predicted} compared to granite paving with higher reflectance. Such a finding can be attributed to limited evapotranspiration from insufficient irrigation, lower reflectance increasing shortwave absorption of vegetation surfaces, and weak winds limiting convective heat removal. The developed model can be extended to other outdoor settings with similar meteorological conditions and morphologies of local climate zone 4, but further refinement is required for robust application across more diverse spatiotemporal scenarios.

The emergence of adaptive facades offers a new approach for buildings to enhance their resilience against external weather conditions while responding to occupants’ demands, thereby improving both indoor environmental quality and energy performance. Appropriate control methods are crucial to achieving these purposes. However, most existing studies for automatic control of blinds have focused on visual comfort, leaving potential for further energy savings by reducing cooling and artificial lighting demands. Additionally, current optimization methods for slat angles are mostly simplified as a discrete process, neglecting the impact of thermal mass in building envelopes. Therefore, this paper aims to explore the energy reduction potential of window blinds by developing an iterative optimization method for devising hourly adaptive control strategies. To this end, a co-simulation platform between EnergyPlus and Python was established for the optimization and a case study in a subtropical city was conducted. The proposed strategies effectively balanced lighting and cooling demands to achieve an overall energy reduction of 7.3%–12.5% compared to reference cases while also ensuring visual comfort by mitigating glare risk and excessive daylight. These advantages were also compared with several simpler control scenarios, with analyses tailored to various glazing types and orientations. Furthermore, the optimal window configurations with blind control strategies for different orientations were determined. The findings also indicated that glass properties markedly impact the performance of control strategies, underscoring the necessity of holistically considering shading components and glazing types in the optimization to achieve optimal performance.

Urban greenery is widely recognised as a strategy to mitigate urban overheating, and its shading capacity is crucial for improving microclimate and thermal comfort. Nevertheless, research on modelling the vegetation canopy radiation transfer (VCRT) process for the microscale is lacking, and most existing VCRT models are taken out from mesoscale models. In this study, we used canopy morphology and structure as an entry point to construct a VCRT model for the microscale, from the mesoscale model. Firstly, 100 m and 1000 m were defined as the critical scales by scaling, and the effect of leaf distribution on VCRT was analysed. The VCRT model was made applicable to the microscale by introducing a scattered radiation source term and 6 leaf inclination distribution (LID) functions, and the results showed that the proposed model could improve the accuracy of canopy radiation absorptivity by about 11%. In addition, both leaf area index (LAI) and LID had significant effects on VCRT, and vegetation with LAI above 2 and spreading leaves had better shading effects. It is worth noting that urban greenery has an exciting potential for thermal comfort improvement, with the potential to regulate even extremely hot weather from near "very hot" (148 W/m²) to almost "comfortable" (69 W/m²). This study is a catalyst for improving the predictability of the VCRT process for microclimate and thermal comfort, providing theoretical support and implications for mitigating urban overheating and enhancing urban thermal resilience.