A Methodology To Fuse Multi-source Land Cover Data For Noah-MP And Its Impact On Simulation Effects Of Surface Temperature And Soil Moisture

Land cover data is one of the important basic data for land surface process simulation.In the land surface models(LSMs),some parameters that reflect land surface features,such as leaf area index,land surface albedo and canopy roughness,are determined by land cover type Currently,there are large differences between the commonly used global remotely sensed land cover products,including data accuracy,classification system,space-time scale and so on,which restrict the application and development of the LSMs.In this study,the global remotely sensed land cover data including FROM-GLC and MODIS and China Land Use Data(CNLULC)were used as fusion objects,and China vegetation map was used as auxiliary data Finally,the China fused land cover data(CFLC)was applied to the simulation of Noah-MP The simulation study of land surface temperature and soil moisture was carried out to analyze the impact of different land cover data on the simulation of Noah-MP.The mainly results are as follows:(1)The classical D-S evidence theory has limitations in dealing with the conflicts of evidences.In this study,the fusion failure occurred in 54.2%of the regions when using the classical D-S evidence theory.The D-S evidence theory improved by mathematical model method had a strong ability to fuse conflicting evidence.The fusion result CFLC maintained the classification accuracy of CNLULC.The overall spatial consistency of CFLC and CNLULC on the six classes reached 84.5%.The Kappa coefficient was 0.796,and the overall accuracy reached 80.3%,which realized the conversion between multiple land cover classification systems and the classification system adopted by Noah-MP.Compared with global land cover data,the accuracy of CFLC was higher than that of MODIS and FROM-GLC global land cover data.The fusion results had a good degree of confidence in Northwest China and North China,but a relatively low degree of confidence in the southern hilly areas,southwest mountainous areas,and parts of the Qinghai-Tibet Plateau.The uncertainty of the fusion result mainly came from the uncertainty of the input data and the uncertainty of the affinity score(2)In order to study the impact of different land cover data on the simulation of ground temperature,the study designed three sets of experiments,including USGS/TG,MODIS/TG,and CFLC/TG.These three sets of experiments simulated the ground temperature of China in 2014 respectively.The results showed that the three groups of experimental simulation results were generally underestimated.The simulation result of the newly produced CFLC data was generally superior to the other two data in terms of bias,root mean square error(RMSE)and correlation coefficient(R).By comparing the spatial distribution of the RMSE of the simulated value and the measured data,the newly produced CFLC data improved the simulation effect in North China and the Jiang huai region,and the RMSE was below 2?.The change of land cover type had a significant impact on the spatial distribution characteristics of ground temperature,especially in the updated construction land area,where the temperature change was above 3?.Overall,the integrated land cover data improved the simulation of ground temperature.(3)In order to study the impact of different land cover data on the simulation of soil moisture,the study designed three sets of experiments,including USGS/SM,MODIS/SM and CFLC/SM.These three sets of experiments simulated the soil moisture at the depth of 10cm of China in 2014 respectively.The results showed that the USGS/SM and MODIS/SM experimental simulation results on the daily scale were underestimated for a long time,accounting for 74.8%and 65.4%of the total days respectively.The improvement effect of CFLC/SM was obvious,and the underestimated days of the simulation results accounted for 51.3%of the total days.the simulation result of the newly produced CFLC data was generally superior to the other two data in terms of bias,RMSE and R.Compared with the default land cover data of the model,the new land cover data improved the stations with large root mean square errors in the southeast and southwest regions,and reduced the number of stations with root mean square errors greater than 0.15 m3/m3.Compared with the simulation results of the default land cover data,the simulation results of the new land cover data was higher in the Qinghai-Tibet Plateau region and the southern region and lower in most parts of the Northeast.Overall,the integrated land cover data improved the model simulation of soil moisture at the depth between 0 and 10 cm.