Taking the precipitation forecasting errors for the reach from Cuntan to Wanxian and inflow flood forecasting errors for Three Gorges Reservoir as an example, the Frank copula, Gumbel copula and Clayton copula were used to analyze the relevance formation of two forecasting errors based on discussing the statistical correlations for the two forecasting errors, and the selection of copula function was performed on the rule of minimum of sum square variation, then the theoretical bivariate joint distribution frequencies were compared to the results of dependent condition. The results demonstrate that considering the correlations between the two forecast errors has some impact on frequency of joint distribution, and on the negative correlation condition, the calculation of the joint distribution between two forecast errors can be simplified. These results may give guidance to the forecasting operation risk management of reservoirs.