
Under the dual pressures of global climate warming and rapid urbanization, the summer heat island effect in high-density cities is becoming increasingly severe. Conventional research has predominantly concentrated on the global average effects of driving factors, overlooking the potential spatial non-stationarity of their impacts across different locations. In order to address this issue, this study takes Shenzhen, a high-density megacity in China, as a case study, and proposes the innovative integration of a geographically weighted regression framework with two advanced machine learning algorithms to construct geographically weighted random forest and geographically weighted XGBoost models. Utilizing a comprehensive multi-source open big data approach, the research establishes a five-category independent variable system encompassing two-dimensional and three-dimensional morphology, distance, geography-economics, and feature engineering, aiming to systematically reveal the spatial heterogeneity of the driving mechanisms of land surface temperature. The findings indicate the following: (1) Geographically weighted machine learning models demonstrate significantly superior predictive accuracy compared to their corresponding global counterparts, confirming the necessity of accounting for spatial non-stationarity. (2) Key drivers exhibit varying influences: in low-temperature areas, natural surface sheltering effects predominate. In moderate-temperature zones, the dynamics shift to an interplay between natural features and urban morphology. In high-temperature blocks, outcomes are determined by urban spatial structures and 3D configurations. (3) The composite index derived through feature engineering demonstrates consistent importance across temperature-tiered blocks, revealing the intricate synergistic effects of blue-green infrastructure. This study not only provides new methodological approaches and empirical evidence for understanding the spatial heterogeneity of urban thermal environments in high-density settings, but also offers direct scientific guidance for differentiated and sustainable urban heat mitigation and spatial planning.
Infants are vulnerable to heat stress within the transient thermal environments of vehicle cabins. While ventilated seat cushions offer a potential solution, their efficacy for this specific demographic under realistic conditions remains poorly quantified. This study systematically investigated the cooling performance of a ventilated seat cushion for infants by simulating a vehicle’s transient air-conditioning cooling process in a climatic chamber. An infant sweating thermal manikin was used to examine the effects of ventilation under varied metabolic rates (sleeping vs. active), clothing ensembles (sleeveless, short-sleeved, long-sleeved), and skin wetness states (dry vs. sweating). Under dry conditions, ventilation reduced final mean skin temperature across all tested clothing and metabolic-rate conditions compared with corresponding non-ventilated conditions, with reductions ranging from 1.10 °C under the sleeping metabolic rate with long-sleeved to 2.39 °C under the active metabolic rate with short-sleeved. The cooling effect was more pronounced with higher metabolic rates and smaller clothing coverage. During sweating, ventilation efficacy was significantly enhanced by latent heat transfer. The long-sleeved outfit achieved the greatest temperature reduction (2.35 °C), as the wetted fabric acted as an evaporative extension layer. Ventilation consistently provided superior cooling to seat-contact regions, especially the back and buttocks. These findings shift child seat assessment from steady-state dry evaluation to transient heat-moisture dynamics, offering quantitative guidelines for targeted ventilation and adaptive thermal management in intelligent car seats.
This paper is a reply to the paper u201CDiscussion of u2018Influence of salts and moisture on the thermal properties of porous building materialsu2019 u201D (
In May 2026, this journal published u201CInfluence of salts and moisture on the thermal properties of porous building materialsu201D, that u201Cexperimentally investigates the variation in the thermal properties of ... sintered bricks in the presence of single salts and salt mixtures across the full moisture content range from saturation to drynessu201D. This discussion voices concerns on this paperu2019s findings, starting from the implausibly small impact of moisture content on the specific heat capacity, which only increases with some 12% to 15% from dry to saturated state. Theoretical calculations, based on the rule of mixtures, infer though that the specific heat capacity should increase about 60% for clay brick and 40% for ceramic brick. Given that the paperu2019s transient measurement technique obtains the specific heat capacity from the measured thermal conductivity and diffusivity, however, flawed specific heat capacities automatically infer flawed thermal conductivities and/or thermal diffusivities. Accordingly, this discussionu2019s specific concerns on the specific heat capacity translate to general concerns on most of the paperu2019s outcomes.
Heating, ventilation, and air-conditioning (HVAC) systems account for a large share of building energy use, but advanced controllers need predictive models that are costly to develop for each building. We investigate whether IBM’s TinyTimeMixer (TTM), a compact time-series foundation model, can be adapted as a control-conditioned surrogate model for HVAC systems. We generated a simulated dataset of 32 one-year trajectories across six buildings, four climates, and two control policies, and fine-tuned TTM to jointly forecast indoor temperature and HVAC power 24 h ahead from historical operation, proposed setpoints, and weather forecasts. A model trained on all source buildings reduced RMSE by up to 36