北疆春季气温和霜冻网格精细化预报订正及检验

Correction and Verification for Grid Refined Forecast of Temperature and Frost in Spring in Northern Xinjiang

选用中国气象局下发的0.05°×0.05°的气温格点预报指导产品和陆面数据同化系统(CLDAS)逐时气温实况数据资料,设计了三种基于平均滤波的温度智能网格预报订正算法,对2019年4—5月北疆平原地区逐日20时起报的未来240 h的逐3 h气温预报产品进行订正,并对比分析检验三种订正后产品和下发的指导预报共四种产品的预报效果。结果表明:经过三种滤波订正后,气温和霜冻预报准确率及稳定性明显提高。在分时效检验结果中,三种订正产品较原指导预报产品的气温均方根误差分别减小了0.79、0.85、0.88℃,气温预报准确率分别提高了6.11%、6.38%、6.46%,霜冻预报准确率分别提高了3.00%、5.81%、7.31%,霜冻逐24 h持续时间预报均方根误差降低了4.21、4.41、4.35 h。在分区域检验结果中,三种订正产品较原指导预报产品的气温均方根误差分别降低了0.66、0.71、0.90℃,气温预报准确率分别提高了5.7%、6.1%、6.1%,霜冻预报准确率在准噶尔盆地东南部海拔高度600~1200 m区域提高明显,分别提高了2.5%、4.8%、5.4%,其他霜冻区域提高不明显。霜冻逐24 h持续时间预报均方根误差平均降低了0.81、0.63、0.56 h。相较而言,最优集成算法的订正预报效果最好。

Using the 0.05°×0.05°temperature grid forecast guidance products issued by China Meteorological Administration and the hourly temperature data of land surface data assimilation system(CLDAS),this paper designs three intelligent grid temperature forecast correction algorithms based on the average filtering and corrects the 3 h temperature forecasts with 240 h lead time starting from 20:00 BT every day in the northern Xinjiang Plain from April to May 2019.Then the forecast effect of three kinds of revised products and guidance forecast products are compared and tested.The results show that the accuracy and stability of air temperature and frost forecast were obviously improved after the three filtering corrections.In the results of time-division test,compared with the original guide forecast product,the root mean square error(RMSE)of the three corrected products decreased by 0.79,0.85 and 0.88℃on average,the accuracy of temperature forecast increased by 6.11%,6.38%and 6.46%on average,the accuracy of frost forecast increased by 3.00%,5.81%and 7.31%on average,respectively.Moreover,the RMSE of 24 h frost forecast decreased by 4.21,4.41 and 4.35 h,respectively.In the regional test results,the RMSE of the temperature of the three revised products decreased by 0.66,0.71 and 0.90℃,and the accuracy of the temperature forecast increased by 5.7%,6.1%and 6.1%,respectively.The accuracy of frost forecast increased significantly in the area of 600-1200 m above sea level in the southeast of Junggar Basin,increasing by 2.5%,4.8%and 5.4%respectively,but not obvious in other frost areas.The RMSE of 24 h frost duration forecast was reduced by 0.81,0.63 and 0.56 h.In comparison,the effect of the revised forecasts by the optimal ensemble algorithm is the best.