| 摘要: |
| 本文利用C#程序语言逐日读取欧洲模式(Ec)、中国模式(T639)、德国天气在线、美国天气、中国天气、中央气象台指导预报六个模式预报中的日最高气温和日最低气温预报值,建立六家模式日最高、最低气温预报数据库,使用LINEST函数对最高气温和最低气温做多元回归分析,得出六家模式的集成预报结果,结果表明：集成预报比六家模式的预报准确率都高；同时,建立本地化订正方法,将原六家模式最高最低气温预报值进行订正后再集成,结果表明：订正后的集成预报比直接集成预报准确率又提高了0～4%,说明订正方法的使用对提高最高、最低气温预报准确率有一定的效果；另外,通过对各家模式预报结果和集成预报结果的检验分析,不仅为预报员择优使用数值预报产品提供参考依据,也为研究数值预报产品释用提供一定的参考方法。 |
| 关键词: 气温预报;LINEST函数;数值预报产品;多元线性回归 |
| DOI： |
| 投稿时间：2018-04-24修订日期：2018-07-06 |
| 基金项目: |
|
| Application of analysis of Multi - nodel consensus forecast for temperature |
| zhangyurong |
| (Bayannaoer Meteorological Bureau) |
| Abstract: |
| According to the European model (Ec), China model (T639), Germany, America, China online weather weather weather, the Central Meteorological Station forecast of six models in the prediction of daily maximum temperature and minimum temperature forecast value, use the LINEST function to do multiple regression analysis, integrated forecasting results, obtained six model results show that: integrated forecast than the six model forecast accuracy is higher; at the same time, setting up a local correction method, the original six model of maximum and minimum temperature forecast value was corrected after integration, the results show that the integrated forecast than the direct integration and improve the prediction accuracy rate of 0 ~ 4%, note the use of the correction methods to improve the highest and lowest temperature forecast accuracy has certain effect; in addition, through the test of the model prediction and forecast results analysis of the results, not only for the forecasters use preferred The numerical forecast product provides the reference basis, also provides the certain reference method for the research numerical forecast product release application. |
| Key words: temperature forecast; LINEST function; numerical forecast product; multiple linear regression |
