How to classify microclimates more validly and finely? A novel method for mapping local climate zone (LCZ) on micro-scales

Abstract

Global climate change and rapid urbanization are driving microclimate variations in urban areas, intensifying the formation of urban heat islands (UHIs) vulnerable to extreme weather. The local climate zone (LCZ) framework, using remote sensing (RS) and geographic information system (GIS), has advanced with the World Urban Database and Access Portal Tool (WUDAPT), enabling microclimate understanding for improved urban planning and climate adaptation. However, mapping LCZs at the micro-scale relies on locally available GIS data or RS imagery, data gaps, authenticity issues, and low-resolution imagery often lead to inaccurate microclimate classifications. To enhance the validity and sophistication of microclimate classification, this study introduces a novel method for mapping LCZ using unmanned aerial vehicle (UAV) photogrammetry at the micro-scale (LCZ-UAV-MS), which constructs detailed land use and land cover (LULC) and 3D real scene (3DRS) models, calculates urban surface parameters (USPs) for each basic spatial unit (BSU) through spatial statistical analysis, and employs a decision-making classifier to categorize each BSU. This study validates the proposed method using Gulangyu as the study area, employing image overlay, temperature observation, and expert knowledge, with the following results: 1) Spatial percentage showing 51 % natural environments (LCZ A, B, C, and D), 27 % neutral environments (LCZ 7, 8, 9, 10, and Y), and 22 % built environments (LCZ 1, 2, 3, 4, 5, and 6) in Gulangyu. 2) The variance in mean LST across different LCZs was 1.02 in January and 5.18 in August 2024, with temperature differences more pronounced in summer, where the built environment had a higher mean LST of 41.3 °C compared to 35.2 °C in the natural environment and 38.3 °C in the neutral environment. 3) Field research at 22 random sample sites showed that 95 % of LCZ-UAV-MS classifications matched the field, outperforming WUDAPT LCZ, which matched only 10 %. These demonstrate that the LCZ-UAV-MS accurately captures microclimate temperature variations and provides a more precise micro-scale description than the WUDAPT LCZ. This study bridges the gap of LCZ studies in specific regions and scales, enhances the applicability of the LCZ framework at the micro-scale, and provides technical support for urban blue-green infrastructure management and resilient climate design strategies.