This dataset is a spatial grid data of permafrost temperature in the Kamalan River Basin of the Dongpo Tahe area in the Greater Khingan Range of Northeast China. The grid unit values represent the permafrost temperature. The dataset is based on the measured ground temperature data obtained from on-site mechanical/manual drilling and exploration pits in the research area, and integrates multi-source environmental spatial datasets such as climate, soil, and terrain obtained from the Google Earth Engine (GEE) platform and authoritative data sources as model prediction variables. By using Python and ArcGIS to complete the preprocessing process of environmental factor format conversion, WGS84 coordinate system spatial registration, target resolution resampling, and normalization, based on three core factors: terrain (elevation, slope, aspect, terrain undulation, terrain humidity index, terrain position index, etc.), soil (soil texture, land cover, bedrock burial depth), and meteorology (temperature, precipitation, freeze-thaw index), the geothermal gradient model, environmental similarity model, and random forest algorithm are comprehensively used to invert and generate the spatial distribution products of permafrost temperature at the scale of the study area. To ensure data reliability, a five fold cross validation was used to evaluate the prediction accuracy of the random forest model, and ArcGIS was used for grid data visualization verification and spatial logic analysis to ensure that the spatial distribution of permafrost temperature conforms to the regional permafrost distribution pattern and has no significant outliers. This dataset can provide key basic data support for simulating hydrological processes, ecological environment assessment, and engineering construction planning in cold regions.