Raw data: actual drilling and pit exploration data: 347 permafrost sample points and 310 seasonally frozen soil sample points; Environmental variable data: Thirteen environmental variables, including terrain, vegetation, climate, and soil, were selected as predictive factors.

    Terrain factor: Extracting altitude, slope, aspect, and terrain undulation based on digital elevation model (DEM).

    Vegetation factor: Using MOD13A3 remote sensing products to extract normalized vegetation index (NDVI); Obtain aboveground and underground biomass data based on the Chinese forest vegetation carbon storage dataset.

    Meteorological factors: Extracting spring precipitation from the ERA5 Land dataset; The surface temperature (GST) data is based on actual drilling and surface temperature measurement data, which are obtained in advance through a random forest model simulation and used as key intermediate variable inputs.

    Soil and thermal condition factors: integrated calculation of melting index, freezing index, humus thickness, and soil moisture content.

    Data preprocessing: Perform spatial registration and standardization on all multi-source raw raster data mentioned above. The unified projection coordinate system is WGS1984_ Albers, and the spatial range is cropped to the boundary of the study area. The spatial resolution of all variables is uniformly downscaled to 30 m using resampling techniques, and the format is unified as GeoTIFF to ensure strict spatial matching of multi-source data.

    Using ArcGIS' Extract Multi Values to Points feature, extract 13 environmental variable values corresponding to each sample point and construct a high-dimensional dataset of "sample environment features". The constructed sample dataset includes the target variables (classification labels: 1 represents permafrost, 0 represents seasonal permafrost) and their corresponding feature vectors. Perform integrity checks on the extracted results, eliminate samples containing missing values (NoData) or outliers, and ensure the quality of the input data for the model.

    Random Forest Model Construction: Stratified Random Sampling is used to divide the dataset into a training set (70%) and a testing set (30%). Build a random forest classification model based on the scikit learn machine learning library in Python environment. To address the issue of sample imbalance, set the class_ceight parameter to 'balanced'. Optimize key hyperparameters through grid search, and ultimately determine the number of decision trees (n_estimators) to be 1000, the maximum depth (x_depth), and the minimum number of samples for node splitting (min_stamples_split), and fix the random seed (random_state) to ensure the reproducibility of the results. Train the model with 13 environmental variables as feature inputs and frozen soil types as labels.

    This data is modeled using machine learning methods (random forest) to calculate confusion matrix, overall accuracy, precision, recall, F1 Score, and Kappa coefficient. The results show that the model has high consistency (Kappa>0.6), and the simulation accuracy of frozen soil distribution reaches 0.86.