How to rapidly map outdoor mean radiation temperatures with high-spatial-resolution from UAV-derived multimodal images: A case study in Guangzhou

Abstract

Mean radiation temperatures (MRTs) have gained widespread attention due to its close relationship with human health. Notwithstanding the availability of existing methods for evaluating MRT (such as ground-based measurements and numerical simulations), it remains a challenge to rapidly and accurately acquire MRTs with a high spatial resolution. This study thus proposed an efficient methodology for evaluating spatial MRTs of urban spaces on the microscale through multimodal images derived from unmanned aerial vehicles (UAVs). Based on UAVs and near-ground measurements, point-cloud data characterizing long- and short-wave irradiance from three-dimensional (3D) surfaces were obtained via photogrammetry and retrieval models. Spatial MRTs were then sampled based on a mapped 3D radiation field and ray-tracing. Comparison of the sampled MRTs with measured MRTs (near the ground) through globe temperatures exhibited a high consistency (R2: 0.980, RMSE: 3.080 °C) and a small difference (almost less than 2.5 °C), demonstrating the applicability and accuracy of our proposed method. In conjunction with near-ground meteorological variables, outdoor thermal comfort (OTC) indicators with a high spatial resolution were derived; further comparisons suggested the SET* and UTCI as ideal indicators to evaluate microscale OTC of hot and humid areas through our method. From the perspective of application, this method can further refine the assessment of the urban thermal environment as well as the human thermal comfort of micro-scale spaces based on simulations or remote sensing observations with lower spatial resolution, which will provide specific suggestions for redesign or redevelopment of the micro-scale areas. Additionally, the thermal mapping with high spatial resolution obtained by this method can also be used as a reference to evaluate the accuracy of related numerical models.