A seasonal forecast scheme for spring dust storm predictions in Northern China

Meteorological applications

article

Tao Gao¹  Zhang Xuebin²  Wulan¹ 

[1] Inner Mongolia Meteorological Institute;Climate Research Division, Science and Technology Branch

关键词: dust storms;    Northern China;    spring seasonal forecast;    forecast signals;    predictors;    hindcasts;   

DOI  :  10.1002/met.175

学科分类：社会科学、人文和艺术（综合）

来源: Wiley

【 摘 要 】

The theme discussed in the present study is that of spring dust storm seasonal forecasts in Northern China. A comprehensive investigation of observations collected from 65 stations in Northern China, which studied strong winds for 35 years (1971–2005) and dust storms for 48 years (1961–2008), concluded that strong winds, which are recognized as a crucially dynamic factor, have unsurprisingly proven to be strongly related to dust storm activity. Therefore, determining effective predictors for strong winds should be helpful in spring dust storm forecasts. By employing this idea, comprehensive correlation analyses among the strong winds, dust storms and other influential elements from the oceans and the atmospheric circulations can be seen. From the spatial correlation fields between prior sea surface temperatures and the strong winds, four regions with higher oceanic coefficiencies are confirmed. The method of EOF (Empirical Orthogonal Function) decomposition is adopted to extract forecast signals from prior precipitation in Northern China and sea surface temperatures of those regions. The multivariable step-regression model is employed to select efficient predictors and the multivariable regression model is used to create forecast equations. With the cross validation approach, six series of 48 year hindcasts with six different predictor sets are conducted. Furthermore, the three-classification forecast method is used to judge successful or failed dust storm forecasts. Together, forecast skills of probability of detection and skill score suggest that series forecasts are better than random forecasts. The best forecast skill is gained from using the predictor set selected by the multivariable step-regression model.

【 授权许可】

CC BY|CC BY-NC|CC BY-NC-ND