概率调整法在气候模式模拟降水量订正中的应用

应用概率调整法订正区域气候模式系统PRECIS在SRES A1B情景下模拟的各季节全国日降水量。以第95百分位降水量为阈值，利用Γ分布分段拟合1962年12月—1972年11月的模拟值，构建传递函数，得到1991年12月—2001年11月的订正值。结果表明:全国平均日降水量空间分布的模拟改善明显，偏差百分率高于100%的格点比例从23.5%降低到1.0%；对各地区平均降水月循环的模拟结果改善，冷季降水较暖季更接近观测，提高拟合优度是改进订正方法的关键；多数地区连续干日数、连续5 d最大降水量及极端降水贡献率的空间强度、概率分布与空间相关性的订正效果显著。总体来说，该方法对模拟中国区域降水的平均态与极端降水均有明显改善，有助于气候评估工作的展开。

Correction Based on Distribution Scaling for Precipitation Simulated by Climate Model

A statistical bias correction based on piecewise Γ distribution fitting to construct seasonal transfer function is applied to the precipitation simulated by a regional climate model PRECIS under the SRES-A1B emission scenario over China. The transfer function (TF) is derived from the control period of December 1962—November 1972, fitting the cumulative probability density function of both simulated and observed precipitation with Γ distribution. The 95th percentile precipitation is chosen to be the threshold and precipitation below and upon the threshold are fitted, respectively. When compared with wholesale fitting, this method can better fit the distributions of both small/medium precipitation and extreme precipitation. Then the TF is applied for the validation period of December 1991-November 2001. The correction strategy is based on the assumption that discrepancies between model and observation stay constant with time.Results show that PRECIS can reproduce the spatial distribution of mean and extreme precipitation, while the biases exist. The biases are larger if the topography is more complex. If the region is high or low in altitudes, the bias tends to be positive or negative, while Sichuan Basin is the exception, where large positive biases occur.The correction based on the piecewise Γ distribution fitting can well correct the spatial distribution of the mean precipitation over China, especially over the original large-bias regions, and the grids in which the bias percentages used to be larger than 100% are reduced from 23.5% down to 1.0%. Simulation of region-averaged monthly precipitation is significantly improved, especially over Southwest China and the Tibet Plateau regions. Precipitation in cold seasons is better corrected, while it has relatively larger biases in warm seasons especially in June due to a wide range of precipitation, which may bring difficulties during fitting. So, it's crucial to improve the fitting probability in warm seasons.The piecewise Γ distribution fitting correction also does a quite good job in correcting the extreme precipitation. The spatial distribution, probability density distribution and spatial correlation coefficient of consecutive dry days, the maximum 5-day precipitation amount and the contribution of extreme precipitation are corrected significantly, except for maximum 5-day preciptiation amount in East China, contribution of extreme precipitation in Northwest China and the Tibet Plateau are overcorrected. These show that the technique has the ability to correct the extreme precipitation.In general, the correction results are satisfying, which implies that the piecewise Γ distribution fitting correction is capable of improving the reproduction of both mean and extreme precipitation simulated by regional climate model PRECIS over China, which is useful for assessment research.