| Under the joint effect of climate warming and rapid urban expansion,the change of urban thermal environment has brought far-reaching influence to ecological environment and social economy,and it is of great significance to study the characteristics of spatio-temporal variation of urban thermal environment.At present,the studies of near-surface heat island mainly rely on site observation,and its coarse spatial resolution can not characterize the fine spatial structure of thermal environment.The surface temperature based on remote sensing inversion can better reflect the spatial pattern of surface heat island,but the air temperature characterization of near-surface heat island is more important to human health and comfort.Based on multi-source spatial data and in situ data,this paper carried out the study of spatial refinement of temperature,and on this basis,made an in-depth analysis of the changes and influencing factors of the thermal environment of Yangtze River Delta.Based on the site and grid temperature data,the characteristics of the thermal environment space-time changes in the Yangtze River Delta region in the past four decades were studied.The spatial representativeness of the site and grid temperature data was evaluated,and the limitations of the site and grid data were found.Therefore,based on the two ideas of spatial downscaling and remote sensing estimation,the spatial refinement of air temperature data was carried out by using four machine learning methods: multiple linear regression,BP neural network,support vector machine and random forest.By comparing the performance of different methods in the Yangtze River Delta region,the method of temperature spacial refinement suitable for highly urbanized and complex underlying surface areas was determined.Using the fine temperature data of 1km grid,the characteristics of space-time change of urban near-surface heat island in Yangtze River Delta region were analyzed,and the effect of surface heat island was compared.At the same time,the effect of land cover and its change on the thermal environment in this area was discussed.The main conclusions are as follows:(1)Based on the observation data of the site in the research area,the spatial distribution and seasonal changes of the air and soil temperature of each site were relatively consistent.In spring and summer they gradually rose from coast to inland.By autumn and winter they gradually rose from north to south.And the annual average temperature was basically low in north and high in south.Both air temperature and soil temperature showed a significant rising trend over the past 40 years.Based on site observation data,SURF monthly mean maximum temperature(R=0.991,MAE=0.671?)outperformed ERA5(R=0.996,MAE=0.957?)reanalysis product in this region.However,due to the insufficient spatial representativeness of site temperature data and the coarse resolution of the existing grid temperature data,neither of temperature data could meet the needs of fine heat island space-time change research.(2)Based on the two approaches of spatial downscaling and remote sensing estimation,spatial refinement of grid temperature data at 1km resolution was carried out using four machine learning methods: multiple linear regression,BP neural network,support vector machine and random forest,and the performances of different methods were evaluated by using site observation temperature data.The results showed that the spatial downscaling model based on random forest had the highest accuracy(R-0.992,MAE-0.905 degrees C)and the remote sensing estimation model based on random forest had the highest accuracy(R-0.993,MAE-0.786 degrees C).In comparison,the accuracy of temperature data estimated by remote sensing based on random forest was higher than the optimal model of spatial downscaling,and its spatial distribution could better reflect the temperature difference between different underlying surface types such as cities,mountainous areas,water bodies and their surrounding areas.The main reason for this phenomenon was that the study area had the characteristics of high urbanization and complex underlying surface,and the temperature grid data at coarse resolution could not reflect the high and low temperature values well.Therefore,the application of the spatial downscaling model to the high-resolution spatial variables did not reflect the spatial characteristics well.Meanwhile,remote sensing estimation retained the advantages and accuracy of site observation data in time,and made up for its discontinuous shortcomings in space,and its meshing effect was better.Therefore,based on the remote sensing estimation method of random forest,1km resolution air temperature data of The Yangtze River Delta region were generated and used to analyze the spatial and spatio-temporal changes of urban heat islands in the region.(3)Based on the 1km resolution grid temperature data,the Yangtze River Delta near-surface heat island was analyzed and compared with remote sensing surface heat island.The heat island intensity calculation showed that the near-surface and surface heat island effects of the study area had been stronger in the last 20 years.Spatially,the distribution of the two was similar,and in core cities and faster-growing areas of Yangtze River Delta the intensity was stronger.The intensity of near-surface heat islands was up to 2.8 ?,while the strength of surface heat islands was up to 4.8 ?.In terms of seasonal changes,the heat island intensity was high in summer and low in winter.Having some similarities in space and time,the intensity of surface heat islands was generally higher than that of near surface heat islands.By calculating the rate of change in heat island intensity between 2001 and 2018,the heat island intensity increase rate was highest in the rapidly expanding urban areas,while the rate of increase in more developed cities was not as high as that in rapidly expanding cities.(4)The study on the relationship between land cover and thermal environment indicated that the thermal environment change in the Yangtze River Delta was mainly caused by the continuous increase of urban land and the gradual decrease of farmland land.Based on the correlation analysis between the land cover landscape index and the thermal environment,the thermal environment of urban land was positively correlated with the Largest Patch Index and Percentage of Landscape,and negatively correlated with the degree of DIVISION.Based on the least-squares regression method to calculate the temporal coupling between air temperature and surface temperature,there was a positive coupling relationship between them in most parts of the study area,and the coupling relationship had shown a weakening trend in the last 18 years.The spatial coupling coordination between impermeable water surface and summer near-surface,and surface heat island was analyzed by using the coupling coordination model.And the coupling degree and coupling coordination degree of between impermeable water surface and summer near-surface,and surface heat island showed an increasing trend between 2001 and 2018,demonstrating the impact of urban expansion on thermal environment has been increasing.Against the background of the decreasing coupling between air and surface temperature between 2001 and 2018,the effect of impermeable water surface on the intensification of the heat island effect was gradually enhanced. |
PDF Full Text Request