Abstract: To improve the accuracy of identifying geothermal anomaly areas by thermal infrared remote sensing, a plateau railway corridor is selected as the research area. The daytime surface temperature is retrieved using the singlewindow algorithm based on Landsat8 data, and the nighttime surface temperature is obtained by ECO2LSTE.  Next, the altitude, slope, aspect, cumulative solar radiation, NDVI, NDSI, NDWI, and albedo are selected as input factors using the random forest method, and the surface temperature without the solar radiation effect is obtained. Finally, the surface temperature after removing the solar radiation effect, fault density, buffer distance to river, and magnetic anomaly are selected as index factors. The geothermal anomaly area is quantitatively identified using the certainty factor model, and the results are evaluated using the known hot spring points. The results reveal the following: (1) The random forest method can effectively remove the temperature changes caused by nongeothermal factors, reduce the effect of solar radiation, and improve the accuracy of identifying geothermal anomalies. (2) The evaluation results of geothermal anomaly areas are consistent with the distribution of known hot spring points, and the identification results of high and mediumanomaly areas are reliable. (3) The plateau railway corridor mainly passes through seven geothermal anomaly areas, and the lengths of the line passing through the high and mediumanomaly areas are 72 km and 125 km, respectively. LudingKangding, Litang basin and its surrounding areas, Batang, Gongjue, ChangduChaya, Basu, and BomiLulang are the sections that need to focus on preventing high temperature and heat damage. The research results have preliminarily realized the quantitative and refined identification of geothermal anomalies in the plateau railway corridor.

Key words:  , plateau railway, geothermal anomaly, quantitative identification, surface temperature, solar radiation effect, random forest