Classification Of Urban Local Climate Zones And Research On Surface Thermal Environment Based On Airborne LiDAR Remote Sensing Data

The local climate zone(LCZ)classification system,as a standardized framework,has been widely used to study urban morphology and urban climate.Multi-source remote sensing and geographic information system(GIS)data provide strong support for large-scale LCZ mapping.Given that Li DAR data can provide height information on the ground,its potential for LCZ classification needs further attention and exploitation.The combined use of Li DAR,multispectral optical imagery,and GIS vector data for more effective LCZ classification is important to study urban morphology,surface thermal environment,and their relationship.To this end,this study combined airborne Li DAR data,Sentinel-2 optical images,large-scale building vector and road vector data for multi-scale LCZ classification and analyzed the surface thermal environment corresponding to the LCZs,taking the 23-ward area of Tokyo(Tokyo 23-Ku)as the study area.Firstly,the above data were used to calculate seven classification indicators closely related to the LCZs,including height of roughness elements,sky view factor,aspect ratio,building surface fraction,impervious surface fraction,pervious surface fraction,and surface albedo.Then,the original reference thresholds provided in the LCZ definition were moderately optimized based on previous research results and statistical values of ground samples,taking into account the actual urban conditions in Tokyo.Secondly,the fuzzy logic classification method based on linear membership function was used to implement LCZ classification with different geographic mapping units,including different sizes of grid cells(cell size from 100 m to 1 km at intervals of100 m)and blocks.The ground samples were used to validate the classification accuracy(including overall accuracy(OA),producer's accuracy(PA),and user's accuracy(UA))in the form of confusion matrices to analyze the effects of the threshold optimization and different mapping units on the spatial characteristics and classification accuracy of the LCZs.Thirdly,to reveal the characteristics and differences of land surface temperature(LST)between and within LCZ classes,multi-period(multi-seasonal and day/night)LST was obtained by Landsat-8 and ASTER.The relationship between LCZ and multiperiod LST was quantitatively and statistically analyzed by one-way ANOVA and posthoc test pairwise multiple comparison method under the mapping unit with the highest classification accuracy.Finally,correlation and regression analyses were used to explore the agreement between the airborne high-spatial-resolution thermal infrared data and the nearcontemporaneous Landsat-8 satellite thermal data under the above multiple grid cells.The main research results were as follows.(1)Using the optimization scheme of indicator's thresholds,the OAs of LCZ classification under all mapping units were significantly improved,and the PAs and UAs of most LCZ classes were also improved to different degrees.The resulting changes in spatial distribution and area percentage were more obvious for the built LCZ classes.(2)The results of LCZ classification under different mapping units differed significantly.Among the different mapping units,the LCZ classification under 100-m grid cells reached the highest OA of 80.34%.Among the multiple grid cells,the OA of LCZ classification decreased with the increase of grid-cell size,and the changes of spatial distribution and area percentage caused by the size of grid cells were more obvious in the built LCZ classes.(3)The urban form of Tokyo 23-Ku defined by the LCZs under the 100-m grid cell,was a general radial ring-like distribution pattern,in the order of LCZ 1(compact high-rise),LCZ 2(compact midrise),LCZ 3(compact low-rise),and LCZ 6(open lowrise)from the inside out.(4)Regardless of the period,statistically significant differences in mean daytime and nighttime LST were observed for most LCZ classes,reflecting the unique surface thermal characteristics of each LCZ class.Regardless of the period,the mean daytime LST of dense and open buildings(i.e.,LCZs 1-6)decreased with increasing height;and their mean nighttime LST increased with increasing height.The mean daytime and nighttime LST of dense buildings(i.e.,LCZs 1-3)were higher than those of open buildings(i.e.,LCZs 4-6)for most periods.The LSTs acquired by the airborne and satellite platforms had a strong agreement within the 400-m grid cell,and the LSTs acquired by the satellite platform were generally higher than those acquired by the airborne platform.