Genetic programming and correlation coefficient methods were used to identify the key hydrologic indicators of fish resources. The results show that the fish biomass downstream of the dam is positively correlated with the annual maximum flow, high pulse duration and the average monthly flow in October and December. The biomass of the four major Chinese carps is positively correlated with the rise rate of water level, 7-day maximum flow and the average monthly flow in May and June. The biomass of the Chinese sturgeon is positively correlated with the base flow index and the rise rate of water level, but it is negatively correlated with the annual minimum flow, and the number of daily hydrologic reversals. These pertinent hydrologic indicators reflect the influences of the magnitude, change rate and duration of water discharge on fish. Nonlinear functions are established between pertinent hydrologic indicators and fish resources. By comparing the simulated values with the observed ones, it can be concluded that the genetic programming method is more suitable than the correlation coefficient method for identifying the key hydrologic indicators affecting fish resources.