Surface temperature data: The surface temperature (GST) of 14 HOBO monitoring stations in the study area, with a time series from January 1, 2024 to December 31, 2024.

    DEM data: 30m data from the Space Shuttle Radar Topography Mission (SRTM).

    Vegetation type data: Global 30 meter land cover fine classification product of the Earth Big Data Science Engineering Data Sharing Service System.

    Perform outlier removal and quality control on the surface temperature (GST) of 14 HOBO monitoring stations in the study area, and calculate the daily average surface temperature (GSTdaily); Project all environmental factor grids uniformly as WGS1984-Albers and resample to a resolution of 30 meters. Using ArcGIS' multi value extraction tool to obtain environmental factor values at 14 HOBO sites; Based on DEM extraction of slope orientation, it is reclassified into four categories: sunny slope, shady slope, semi shady and semi sunny slope, and flat slope.

    Construction of regression model for "surface temperature environmental factors": Accumulated temperature with daily average temperature greater than 0 ℃ is counted to obtain accurate measured melting index (TDDobservated, unit: ℃ · day) for each station. Select key factors that affect surface thermal conditions. Extract elevation, terrain moisture index (TWI), and normalized vegetation index (NDVI) based on DEM. Construct a multiple linear regression model with TDDobservated as the dependent variable and altitude, TWI, and NDVI as independent variables. Apply the regression model to the entire watershed, perform calculations using environmental factor raster data, and generate preliminary TDD prediction data for the study area. Calculate the residual between the measured and predicted values at each site. Using the inverse distance weighting method (IDW) for spatial interpolation of residuals to generate a continuous TDD residual surface. Overlay the preliminary prediction map with the residual surface to obtain the corrected final melting index distribution data.

    Construction of ecological terrain classification units and E-factor inversion: Grid overlay analysis was performed in ArcGIS to combine surface cover and slope orientation, generating a total of 36 ecological terrain units. Based on the active layer thickness (ALTobserved) obtained from actual drilling and pit exploration, as well as the corresponding melting index (TDD) at each point, the E factor (Eoobserved) of each measured point is calculated using Stefan's deformation formula. Construct a correspondence table for "classification units - E factors", assign E factor values to spatial classification layers through attribute joins, and generate spatially continuous E factor distribution data.

    Calculation of Active Layer Thickness (ALT): Calculate the ALT of the study area pixel by pixel based on Stefan's equation. To further reduce model bias, perform secondary correction on the ALT simulation results and calculate the deviation between the measured point ALTobserved and the simulated value ALTcalculated. Similarly, the IDW method is used to generate the ALT residual surface, which is then overlaid with the preliminary calculation results to obtain the final corrected active layer thickness distribution data.

    The regression model of "surface temperature environmental factors" has a coefficient of determination (R2) of 0.67. Compared with the measured data, the final simulation result of the thickness of the active layer has a determination coefficient (R2) of 0.64 and a root mean square error (RMSE) of 0.41 m, which meets the requirements of mapping accuracy.