GRAPES_RAFS系统2m温度偏差订正方法研究

Study on Bias Correction for the 2 m Temperature Forecast of GRAPES_RAFS

本文通过对2013年6月20日至7月20日GRAPES(Global and Regional Assimilation and Prediction System)_RAFS(Rapid Analysis and Forecast System)系统每天8个时次每3 h的2 m温度预报进行分析,发现各时次的预报均能较好地表征2 m温度日变化特征,但预报与实况存在一定的偏差,其中西藏东部川西高原、云贵高原、江南武夷山脉偏低于实况可达3℃,而华北地区偏高于实况3℃以上。为了减小GRAPES_RAFS系统偏差对2 m温度预报的影响,本文采用平均法、双权重平均法、滑动平均法和滑动双权重平均法分别对GRAPES_RAFS系统2 m温度预报产品进行偏差订正,并对订正前后的结果进行检验分析和对比。结果表明:2 m温度订正后的平均误差大部地区减小到(-1~1℃),而均方根误差大部地区降低到2.5℃内。对于偏差较大地区,订正效果更为明显,如西藏东部川西高原,经过订正,平均误差绝对值由订正前3℃以上降低到1℃内,而RMSE由订正前4℃以上控制到3℃内。对比四种订正方法,双权重订正方法与平均法订正整体效果接近,但对个别站点,双权重订正法要优于平均法,经过滑动的订正方法比无滑动的订正方法订正效果更好,订正效果最好的是滑动双权重平均法,全国平均误差大部分在(-0.5~0.5℃)内,不超过(-1~1℃)的范围。

The 8-times daily 3 h forecast of the 2 m temperature forecast of GRAPES_RAFS(Global and Regional Assimilation and Prediction System_Rapid Analysis and Forecast System) during the period from20 June to 20 July in 2013 is analyzed in this paper.It is found that the forecast can show the diurnal variation of the 2 m temperature fairly well,but,some deviations exist between the forecast and the observation.For instance,the forecasted values are 3℃ lower on average than the observation in Western Sichuan Plateau in eastern Tibetan,Yungui Plateau and Wuyi Mountains,but 3℃ higher than the observations in North China.In order to diminish the influence from those deviations,bias correction is conducted by mean method,biweight method,moving mean method and moving-biweight method respectively.The values before and after the bias correction are analyzed and compared.The results show that the average error is reduced to(-1-1℃) in most regions while the RMSE(root mean square error) is less than 2.5℃.The bias correction is more effective in areas with larger deviations like Western Sichuan Plateau in eastern Tibetan where the absolute of the average error is reduced from above 3℃ to within 1℃ and the RMSE is reduced from above 4℃ to within 3℃.Comparing the four methods of bias correction,we see that the biweight method is generally as effective as the mean method,but it works better than the mean method at some particular sites.The moving methods are more effective than the non-moving methods.The movingbiweight method is the most effective method,which can reduce the average error down to be within(-0.5-0.5℃) in most area of China and no larger than(-1-1℃).