Research On The Impact And Mechanism Of Forest Change On Land Surface Temperature In China Based On Remote Sensing And Land Surface Mode

Forests are the main body of terrestrial ecosystems,playing an important role in protecting biodiversity and mitigating climate change.Since the reform and opening up,China has implemented a series of large-scale forest ecological projects,and forest change has become one of the most important vegetation change processes in China.Land surface temperature（LST）is an important biophysical parameter for studying the interaction between terrestrial ecosystems and climate change.Quantifying the impact of forest change on LST is of great significance for studying the feedback of forest ecosystem to climate change in china.Many studies have explored the impact of forest change on LST from the perspective of carbon cycle,ignoring the impact of forest change on LST through biophysical processes.However,due to the spatiotemporal discontinuity of remote sensing vegetation parameters data and the uncertainty of numerical models in vegetation simulation,the biophysical effects and related mechanisms of forest changes,especially changes in forest internal characteristics（such as greenness and connectivity）on LST are still unclear.Therefore,this paper reconstructed the MODIS leaf area index（LAI）data based on an advanced unsupervised machine learning method,and improved improved the Noah-MP land surface model for LST simulation.Using the reconstructed LAI data and other multi-source remote sensing data combined with the improved land surface model,this paper adopted offline simulation,decomposed temperature method（DTM）and trading space for time analysis method to discuss the impacts of forest greening and connectivity change in China in the past two decades on LST and related mechanisms from the perspective of biophysics.The main conclusions of the study are as follows:（1）The comprehensive reconstruction method（GANSG）was proposed in this study,which can better realize the time series reconstruction of MODIS LAI products by using Generative Adversarial Networks and improving S-G filter,which was better than the other five reconstruction algorithms.The GANSG method improved the ability to interpolate low-quality pixels by adopting the Generative Adversarial Network（GAN）interpolation method.The GANSG method achieved high-quality reconstruction of LAI time series by improving the traditional S-G filter whith better convergence.Compared with the three classical reconstruction algorithms（adaptive S-G filter,asymmetric Gaussian method and double logic method）and two cutting-edge spatio-temporal reconstruction methods（Spatio-temporal S-G filter and modified temporal spatial filter）,GANSG method can better deal with the continuous loss of high-quality pixels in the LAI time series,and can accurately identify the phenological characteristics of vegetation.Station observation verification showed that the reconstruction effects of the six methods can be ranked as follows:GANSG>Spatio-temporal S-G filter>Modified temporal spatial filter>adaptive S-G filter>asymmetric Gaussian method>double logic method.（2）The new vegetation parameterization scheme based on the reconstructed LAI data was proposed to improve the LST simulation of the Noah-MP land surface model by providing more accurate vegetation information,and the new scheme affected the LST simulation by changing the simulation of each component in the surface energy balance.The LAI affects the LST simulation by affecting the simulation of each component in the surface energy balance.Compared with the new satellite LAI scheme（OBS）,the LAI obtained by the original parameter table scheme（TAB）lacked spatial heterogeneity,while the original dynamic vegetation scheme（DVEG）overestimated the LAI value and failed to simulate the phenology of vegetation such as forests.The validation based on 2304 automatic weather station observations in vegetation covered areas showed that the OBS scheme can improve the underestimation of the original scheme on the LST simulation and reduced the unbiased root mean square error.The order of improvement on the LST simulation on the four vegetation types was:grassland>forest>farmland>other vegetation.In the Noah-MP LSM,the LAI change of dense vegetation mainly affects the LST simulation by affecting the competition of shortwave net radiation（SW）,latent heat flux（LE）and sensible heat flux（H）.However,the LAI change of sparse vegetation mainly affects the LST simulation by changing the SW and surface heat flux（G）.（3）In the past two decades（2001-2018）,Forests in China have shown a greening trend.Forest greening had a cooling effect on the surface through biophysical processes.Latent heat and sensible heat dominated the biophysical processes of forest greening affecting LST.Forests in China generally showed a greening trend,and the LAI change trend was 0.352m²m^-2decade^-1.97.34%of the forest showed an upward trend in LAI,which increased with the decrease of latitude.The average annual LAI growth trend of evergreen forests was greater than that of deciduous forests,and that of broadleaf forests was greater than that of needleleaf forest.Forest greening in China had a cooling effect on the land surface.The average impact of forest LAI changes on the annual mean LST was-0.146°C decade^-1,increasing with latitude.The DTM method was used to further explore the biophysical mechanism of forest greening on LST in China.The results showed that the annual mean contributions of SW,LE and H changes to LST caused by LAI changes were6.31%,47.87%and 43.42%,respectively.The LAI changes dominated the annual mean LST change of needle forests through H,and the LAI changes dominated the annual mean LST change of deciduous broadleaf forests through LE.The dominant process of LAI changes in evergreen broadleaf forests affecting annual average LST depended on the difference in effect intensity of H in autumn and winter and LE in spring and summer on LST.（4）In the past two decades（2003-2018）,the core forest area in China has increased,improving the spatial connectivity of forests.The increased connectivity enhanced the cooling effect on the surface through biophysical processes.The competition of shortwave net radiation and turbulent heat flux dominated the biophysical processes by which connectivity changes affect LST.Multi-source satellite observations showed that during 2003-2018,the net increase area of core forests with high connectivity was 45300 km²,while the net increase area of marginal forests with low connectivity was 124200 km².The proportion of forests with high connectivity（>0.8）increased from 45.7%to 49.1%.Core forests with high connectivity had a cooling effect on the land surface and maintained a lower diurnal and seasonal temperature range.Compared with the nearby marginal forests and croplands,the daily cooling effects of the core forest were 0.42°C and 1.12°C,respectively.Core forests in low latitude had a stronger cooling effect than that in high latitude,and core forests in high latitude were more capable of maintaining seasonal temperature ranges than that in low latitude.The biophysical mechanism of core forests affecting local LST was the competing effect between SW and turbulent heat fluxes（LE and H）.Compared with croplands,core forests had larger leaf and canopy structures,thereby enhancing evapotranspiration and cooling the surface.Therefore,core forests dominated the cooling process of the land surface through LE compared with croplands.While compared with the marginal forests,the core forests had higher LAI values,with a greater roughness,thus dominating the cooling process of the land surface through H.