The Urban Thermal Environment: Spatio-temporal Dynamics And Planning Strategies

The rapid urbanization around the world has triggered global climate change to some extent through the continuous replacement of natural land cover with artificial surface.On city scale,the urbanization induced local warming has posed a great threat on the health of urban residents,biodiversity and urban sustainability.Land Surface Temperature(LST)has become a key indicator for investigating the impact of urbanization on urban thermal environment.To employ data science techniques to explore the mechanism of urban thermal environment covering cycle and process,is thus significant to guide urban planning towards urban temperature problems in a more scientific manner.However,LST is intrinsically temporal non-stationary whether on the phenomena or dynamic level.As a result,the study on urban thermal environment demands to be extended to the temporal dimension from the recent predominant spatial analysis,so as to avoid the misleading generated by using only on snapshot.The spatio-temporal pattern recognition and dynamic variation analysis will generate more comprehensive,solid and effective knowledge of urban thermal environment,and thus provide more scientific support for urban planning and management professionals.Besides,the effective collaboration between urban thermal environment research and urban planning and management domain is also a major concern.In fact,it has always been the ultimate goal for urban thermal study to provide effective guidance for the domain of urban planning and management to settle urban temperature problems,especially through the regulation of land surface composition,configuration and urban morphology.Targeted on the city scale,this paper is devoted to extend the conventional static research framework of the urban thermal environment to a more spatio-temporal dynamic manner with a special emphasis on urban inner heterogeneity.By taking Wuhan,a city with a typical subtropical monsoon climate,as example,this study investigates the spatio-temporal patterns and the dynamic variations of LST in different seasons,and studies the specific temperature problems both on the city and local scales,and then generates corresponding recommendations from the domain of urban planning and management.On one hand,multiple approaches in data science are employed for the spatio-temporal pattern and dynamic exploration given the intrinsic temporal non-stationarity of LST.On the other hand,zoning is conducted on the phenomenon level to investigate spatial heterogeneity,and further bridge with the domain of urban planning and management.Specifically,(1)time series clustering from machine learning is introduced for phenomenon zoning,and then,the Ensemble Empirical Mode Decomposition(EEMD)from digital signal processing is employed to explore the multi-timescale temporal patterns and the corresponding impact from urbanization.Results reveal that the overall trends exert considerable differences which may be potentially caused by the inconsistent levels of localized urbanization,afforestation or circular economy development.(2)Given the significant application potential of time series clustering on the spatio-temporal pattern recognition of LST,this study further investigates the optimal clustering method on short-term time series LST data.Based on the latent pattern of LST and its morphology,multiple time series clustering are applied and evaluated.In fact,such phenomenon zoning through time series clustering is able to complement the current environmental functional zoning paradigm.Specifc results reveal that the optimal clustering number for Wuhan under different spatial resolutions from 2002/2003-2017 is always 17.(3)For the mechanism study of the spatio-temporal characteristics of LST,the paper explores the dynamic variations of LST in different seasons given the possibility of spatial non-stationarity in the associations between LST and land surface indicators.The Multi-scale Geographically Weighted Regression(MGWR)model is employed for the association analysis.It breaks the conventional assumptions of global stationarity for global regression models and consistent scale for all processes for the Geographically Weighted Regression(GWR)model.The local coefficients generated from the model can thus provide unique information for site-specific mitigation from urban planning and management apartments.The evident variation of LST dynamics in different seasons also highlights the importance to consider such seasonal variation in temperature mitigation or climate modelling.Thirdly,the paper presents mitigation strategy to optimize the thermal environment pattern of the study area based on the spatio-temporal analysis results.Besides,the impact of different land surface indicators on LST in a hottest month of Wuhan is further analyzed based on regression results of MGWR.Site-specific mitigation strategies and measures from the domain of urban planning and management are then put forward for local areas with severe temperature problems.On the whole,this study facilitates our understanding of human-environment interaction in a spatio-temporal dynamic context.Besides,by targeted on urban inner heterogeneity,the urban thermal environment knowledge generated in this study can be easily absorbed by urban planners,thus can be further applied in site-specific climate-sensitive planning and design.