Abstract: Five interpolation methods—ordinary kriging (OK), co-kriging with the auxiliary variable of dissolved oxygen (CK-DO), co-kriging with the auxiliary variable of UV254 (CK-UV254), regression kriging (RK), and ordinary kriging combined with BP neural network (BP-OK)-were applied to predict the spatial distribution patterns of COD in the surface water of Xinhe Lake in Qiaokou Town, Hunan Province. The results show that the COD concentration in Xinhe Lake has a strong space autocorrelation and the semivariance models fit well, which provides a good theoretical basis for the kriging interpolation. Moreover, results of the five interpolation methods are in agreement with the local investigation of pollution sources and can reflect the spatial distribution of COD in Xinhe Lake. CK-UV254 is the most accurate method, and the accuracy of each interpolation method shows a declining trend with the decrease in interpolation points.