3种常用温度MOS预报方案的对比试验  

Comparison Test of Three Common Temperature MOS Forecasting Schemes

作　　者：高万泉[1] 李玉娥[1] GAO Wanquan;LI Yu'e

机构地区：[1]保定市气象局,河北保定071000

出　　处：《山西科技》2020年第6期59-63,共5页Shanxi Science and Technology

摘　　要：为了寻求最好的温度预报方法,进一步提高温度预报质量,利用2016—2018年的ECMWF细网格产品资料,对目前常用的3种MOS预报方案——经典MOS预报、动态回归预报、相似回归预报进行了对比应用试验,并针对动态回归的最优训练期日数进行了探寻。结果显示:这3种MOS预报方案预报质量均较高,达到了目前实际业务需要,对预报员有很大的参考价值。3种方案中动态回归预报最优,动态回归的最佳训练期日数最低、最高气温略有不同,最低气温的训练期日数为35~40天,最高气温的训练期日数为20~35天。In order to find the best temperature forecasting method,further improve the temperature forecasting quality,and use the ECMWF fine grid product data to compare the three commonly used MOS forecasting schemes:classical MOS forecasting,dynamic regression forecasting and dynamic similarity regression forecasting.The optimal training period days for dynamic regression were explored.The results show that the forecast quality of these three MOS forecasting schemes is high,which meets the current actual business needs and has great reference value for forecasters.The dynamic regression forecasting is the best among the three schemes,the best training days of dynamic regression for minimum and the maximum temperature are slightly different.The best training days of minimum temperature are 35~40 days and the best training days of maximum temperature are 20~35 days.

关 键 词：经典MOS预报 动态回归 相似回归 温度预报 

分 类 号：P456.7[天文地球—大气科学及气象学]