Inversion Of 1km Surface Soil Moisture In Permafrost Areas Along The Qinghai-Tibet Engineering Corridor During The Thawing Seaso

Under global warming,the Qinghai-Tibet Plateau(QTP)has experienced phenomena such as the retreat of glaciers,the expansion of some lakes,and the degradation of permafrost,which have had a profound impact on the QTP water cycle process.The high-resolution soil moisture products over the QTP permafrost regions are limited because of the harsh natural environment.This study takes the permafrost regions along the Qinghai-Tibet engineering corridor as the study area.Firstly,take the relevant surface variables of soil moisture,such as land surface temperature,normalized difference vegetation index,enhanced vegetation index,slope,aspect,elevation,soil texture,longitude and latitude as the input variables,and take the in-situ soil moisture as the target variable,to establish two kinds of soil moisture retrieval models at the daytime and nighttime based on the random forest regression method.At the same time,two soil moisture products at 1 km spatial resolution based on daytime retrieval model and nighttime retrieval model in the study area from May to September 2015 to 2018were obtained.Secondly,the ascending and descending soil moisture retrieve models were constructed by taking land surface temperature,normalized vegetation index,enhanced vegetation index,slope,aspect,elevation,soil texture,latitude and longitude,and backscatter as input variables,and in-situ soil moisture as the target variables.Similarly,one soil moisture product at 1 km spatial resolution based on ascending retrieval model May to September2015 to 2018 was obtained.Then I have explored whether the backscatter can improve the accuracy of the soil moisture retrieve model by comparing the retrieve results of the daytime retrieve model with the retrieve results of the ascending model retrieve.The results showed that:(1)Daytime and nighttime soil moisture retrieval models both have good retrieval accuracy.The correlation coefficients between the estimated and in-situ soil moisture in the validation period are higher than 0.97,the root mean square error are the same with 0.03m~3/m~3,and the bias are both lower than-0.0002 m~3/m~3.The simulation results of each station were better.(2)The ascending soil moisture retrieve model and the descending soil moisture retrieve model constructed by introducing the backscatter have high accuracy.Two models are manifested with R between in-situ soil moisture and retrieved soil moisture greater than 0.95,RMSE of 0.03 m~3/m~3,and the bias of-0.002 m~3/m~3.In addition,the simulation results of the model in most sites have a strong correlation with the measured value,and the retrieved error is low.(3)Three sets of 1 km surface soil moisture products retrieved using the daytime retrieve model,nighttime retrieve model and ascending retrieve model have been verified to have high accuracy at three independent stations.In addition,the spatial distribution characteristics of the 3 predicted surface soil moisture products were consistent with that of the vegetation types,showing a decreasing trend from southeast to northwest.(4)The retrieved soil moisture product based on the daytime model without backscatter and ascending model with backscatter are compared.The result shows ascending model is better at 3 validation sites,and its spatial distribution trend is more consistent with vegetation types than that of the daytime model.This means adding backscatter as a retrieve predictor can improve the accuracy of the existing model.