Inversion Of Permafrost Thickness In The Da And Xiao Hinggan Mountains Based On Integrated Physical Sounding Techniques And Its Influencing Factors

The permafrost development,distribution and degradation of the permafrost zone in the Da and Xiao Hinggan Mountains,which is an important part of the northeastern ecological environment,has attracted much attention.In this paper,three permafrost study areas were selected based on the environmental conditions of the ground source-Wuling area in Yichun,Gongbeila River Reserve in Heihe,and Huzhong area in the Da Hinggan region-and classified into four types of features-woodland,scrub meadow,mud roadside,and streamside.Field surveys were conducted in August 2021.The survey experiment is based on two types of physical sounding equipment,ground-penetrating radar and transient electrometer,to survey the upper and lower limits of the permafrost layer(active layer and permafrost layer)and obtain data on the active layer and permafrost layer thickness.The correlation analysis of permafrost development,distribution and natural elements is carried out based on the measured permafrost data and natural elements data using multiple linear regression.The inversion of the active layer and permafrost thickness distribution in the study area was carried out based on three machine learning models using remote sensing image information,feature types and measured data.The main conclusions were obtained as follows:(1)In this paper,ground-penetrating radar and transient electromagnetic equipment are used to make field measurements of the permafrost zone in the Da and Xiao Hinggan Mountains.The depth range of the active layer is determined by the difference in dielectric constant data from the ground-penetrating radar high-frequency electromagnetic waves,and the depth range of the permafrost layer is determined by the difference in resistivity data from the pulsed electromagnetic field of the transient electromagnetic instrument.The data collected is visualised using the relevant software matched to the equipment to obtain the final depth data for the upper and lower permafrost limits(active layer and permafrost layer thickness).In order to ensure the effectiveness and accuracy of the two types of physical probing equipment to measure the thickness of the permafrost layer,validation experiments were carried out in two ways,namely pit probing and metal brazing methods,and borehole temperature comparison,and the validation data were basically consistent with the measured data,proving the feasibility of the two types of physical probing equipment to measure the permafrost.(2)In this paper,four types of features,namely woodland,scrub and meadow,streamside and dirt road,were selected as the comparative analysis areas for analyzing the factors influencing the thickness of permafrost layer and data were collected to obtain the preliminary analysis results of the differences in the thickness of permafrost: the active layer thickness in the Da and Xiao Hinggan Ridge is mostly concentrated in the range of0.8-2m,and the thickness of permafrost layer is mostly concentrated in the range of 15-60m;the development of permafrost beside water systems and dirt roads is significantly worse than that in places with high vegetation cover;the thicker the active layer,the thinner the permafrost layer.The permafrost layer is thinner where the active layer is thicker,and there is delamination and faulting between the active layer and the permafrost layer and between the permafrost layer and the permafrost layer at the same measurement site in the island-shaped seasonal permafrost area.Five factors,namely surface cover type,soil temperature,soil moisture,latitude and elevation,were selected as the correlates of active layer and permafrost thickness variation in the Da and Xiao Hinggan Mountains using multiple linear regression models for correlation analysis.By analysing the significance,Durbin-Watson and correlation coefficients,we obtained that the type of feature has a greater influence on the change of permafrost thickness,with streamside water systems having the greatest influence on the thickness of the active layer and small roadsides having the greatest influence on the thickness of the permafrost layer;soil temperature and moisture have a greater influence on the change of permafrost thickness,with a positive trend of correlation with the thickness of the active layer and a negative trend of correlation with the thickness of the permafrost layer;on a large spatial scale The influence of latitude on permafrost thickness is greater among the factors latitude and altitude,with latitude having the most significant influence on permafrost thickness.(3)The training set was created by matching the waveband information in the remote sensing images of Gaofen 6 with the field measurement data(feature type,active layer thickness and permafrost layer thickness)in a one-to-one correspondence,and by using three machine learning models MLR,RF and SVM to build the relevant inversion models according to the training set and test set,and finally the inversion of the spatial distribution of permafrost in a certain area of the survey sample area was carried out,and the corresponding results were obtained.The corresponding results were obtained.The three models are combined by scaled extraction to form an optimised model,which in turn improves the accuracy of the inversion.