Abstract:Using frequency or area matching method, an experiment is performed to correct precipitation forecast bias from AREM model over China during June-August, 2012 (rainy season). The method is first introduced including its meteorological principle and mathematical means, and its performance is evaluated. Then in depth analysis is done to understand its advantages and limitations, and future research directions are suggested. Three conclusions are drawn from this study. (1) The method can effectively eliminate biases in precipitation amount and areal coverage. The corrected precipitation amount and area coverage are much closer to the observations. (2) The method works better when bias is higher. (3) This method, in principle, has no ability to correct precipitation position errors, but through precipitation area alternation, large area of erroneous or spurious light rain from model forecasts can be effectively removed, resulting in significantly improved TS and ETS scores for light precipitation. Therefore, this calibration method can greatly increase the accuracy of a “rain or no rain” dichotomous forecast in numerical weather prediction. In addition, five possible approaches are proposed to remedy the inability of this method in correcting precipitation position error. This method can also be used at single station. Hydrological application of this method is apparent since an accurate input of precipitation amount and areal coverage is critical to a hydrological forecast model over a watershed. Besides for precipitation, this method can also be applied to correct forecast bias for other variables like fog.