Based on the daily accumulated precipitation from the European Centre for Medium-Range Weather Forecasts (ECMWF) as the forecast data, and the daily accumulated precipitation from ground meteorological stations as observations, experiments of precipitation forecast on the township scale are carried out by means of the frequency matching method. The results show that, compared with the bilinear interpolation, the frequency matching method is more effective in increasing the anomaly correlation coefficient and the equitable threat score, as well as decreasing the root-mean-square error, the false alarm rate of light rain and the missing rate of heavy rain. The frequency matching method can greatly calibrate and improve the refined precipitation forecast on the township scale.