Spontaneous coal combustion (SCC) is a hidden danger in gob areas of underground working faces and is characterized by strong concealment and difficulty in detection.Gas is a key indicator used to characterize the degree of SCC.It is feasible to classify SCC risk areas using the gas volume fraction and its variation patterns.In this study, taking the gas in the gob area of the 401,103 working face in the Mengcun Coal Mine as the research object, the variation features of CO and O2 vs. the length of the gob area were analyzed; on the basis of the nonparametric kernel d. estimation method, the kernel d. distribution patterns of the volume fractions of O2 and CO were analyzed, the early warning threshold of SCC was determined, and a classification method for SCC risk areas was proposed.The results in this study revealed the following: The statistical features of the gas showed that the O2 volume fraction of the gob area changed linearly, and the CO volume fraction met second-order polynomial change features. The kernel densities of the CO and O2 volume fractions increased with increasing gob area length.The highest kernel d. of CO was distributed within 10 m behind the shelf, and the highest kernel d. of O2 was distributed in the range of 20 ∼ 25 m. The length of the critical gob area for SCC at the 401,103 working face was approx. 30 m, and the critical volume fractions of O2 and CO were 15% and 0.009%, resp. The early warning critical value divided the gob area into four quadrants and three risk levels of safe, potential risk and danger.The dynamic effect of the change in the gas volume fraction on the risk level was analyzed.Compared with the traditional use of single data points to classify SCC risk areas, the determination of SCC early warning thresholds and the classification of danger areas on the basis of gas statistical features significantly improved the accuracy of early warnings.This study provides novel ideas for the early warning of SCC.