Study On Three-Dimensional Urban Expansion And Spatio-Temporal Variations Of Land Surface Temperature In Zhengzhou

With the acceleration of urbanization,great changes have taken place in the three-dimensional spatial layout of the city,which in turn has a great impact on the land surface morphology,local regional climate and urban thermal environment.Studying the spatial-temporal distribution changes of Land Surface Temperature(LST)under three-dimensional urban expansion for many years not only can be conducive to disclosing the pattern of urban development and expansion in a timely and objective manner,but also be helpful for analyzing the dynamic changes of thermal environment,and provide feasible suggestions for the promotion of sustainable development and governance of thermal environmental problems.Based on extracting building height with stereo image pairs,Random Forest classification and the Flexible Spatiotemporal Data Fusion(FSDAF)method,combined with multi-source remote sensing datasets and meteorological site data from2013 to 2017,the central area of Zhengzhou city was selected to conduct explorations on three-dimensional(3D)urban expansion qualitative and quantitative,spatial distribution characteristics of LST in different time series and multi-factor driving forces of LST changes by the technology of RS and GIS.The results indicate that:(1)The 3D urban expansion of study area from 2013 to 2017 is pretty significant.In the horizontal direction,the area of construction land increases by 104.26km~2 and shows the trend of developing towards the east,southeast and north,while the change of building density is mainly characterized by the transition from no-building area and low-density area to a greater density area.Among them,the area of the low-density regions reduces by 18.96km~2 with the most prominent changes.As for the vertical direction,it is mainly manifested by the increase of building height.The average height of buildings in the four-ring range increases from 26.33m to 37.37m.On the whole,the overall structure is characterized by a sharp decrease in single-storey,low-rise and multi-storey buildings,and a rapid increase in high-rise and super-high-rise buildings.(2)Based on FSDAF spatial-temporal fusion algorithm,with using the LST data of MODIS and Landsat,the monthly high spatial-resolution LST datasets from 2013 to2017 were generated.Based on the quantitative evaluation and analysis,it shows determination coefficient R~2?0.670 and the overall rootmean square error RMSE ranges from 1.30?to 2.36?,which indicates that the high accuracy of predicted Landsat-like LST meets the needs of this study.The LST distribution of different time series show much more intense effect of central high temperature agglomeration and the increasing trend of mean LST in various municipal districts year by year.The monthly series LST show the alternation of"hot island"and"cold island",and the interannual center high temperature agglomeration effects spread from June to September to April to October.The characteristics of seasonal LST present"hot island"in summer,and"cold island"in winter,and high temperature agglomeration in the southeast of the study area in spring and autumn.However,interannual outstanding characteristics are showed with"heat island"spreading from summer to spring and autumn.The annual LST series characteristics maintain the high temperature and sub-high temperature coverage of the surface in central,southern and southeastern parts of study area.(3)With the predicted monthly Landsat-like LST data,the multi-factor driving forces of LST changes were analyzed quantitatively.As shown in the results,the extremely strong nonlinear cubic polynomial correlation between the building density and LST(0.896<R~2<0.951)as well as the strongly complex nonlinear correlation model between building height and LST(0.716<R~2<0.868)both show significant seasonal differences and staged changes.What's more,the mean LST of various land types LST also presents obvious seasonal differences.In summer,the sequence of average LST is construction land,unused land,agricultural land,woodland/grassland,and water area.However,the sequence of average LST is unused land,agricultural land,woodland/grassland,construction land,and water area in winter.NDBI-LST basically maintains a pretty strongly positive driving mode(except from November to the next February,R~2 is always greater than 0.3),and it expresess much more stable and robust than the negative driving model of NDVI-LST(0.41<R~2<0.57),which only shows strong correlation in summer.The influence of various climate factors on LST is characterized significantly by the strongly positive driving force of temperature to LST(R~2 is 0.878 and 0.911,respectively)and the negative correlation between aerosol(PM2.5)and LST(R~2 is 0.696 and 0.765).