时间序列法在临近空间大气风场预报中的应用

受多种因素影响，临近空间大气环境要素复杂多变，预报难度很大.本文采用时间序列法中的自回归滑动平均（ARMA）模型对临近空间大气风场开展统计预报方法研究，基于廊坊（39.4°N，116.7°W）中频雷达在88km高度的大气纬向风数据开展预报试验.本次预报试验的样本数据为2015年9月24日至10月24日风场数据，利用过去7天数据对未来第8天风场数据进行预报.试验结果显示，ARMA模型对临近空间大气风场预报有一定的适用性.当风场变化规律性较强，即样本数据风场呈现出比较显著的24h周期性变化时，ARMA模型预报效果较好；当风场发生突变时，预报效果变差.与实测数据的对比结果表明，ARMA模型预报结果的误差在9~27m·s-1，预报效果优于同阶自回归（AR）模型，略优于高阶AR模型. 

Application of Time Series Method in Forecasting Near-space Atmospheric Windormalsize

Due to many factors, near-space environment is complex and variable. Atmospheric environmental elements are hard to be forecasted. In this paper time series method is applied to the near-space wind forecasting. Autoregressive Moving Average (ARMA) model is adopted. The zonal wind data at 88km altitude of Langfang (39.4°N, 116.7°W) MF radar form September 24 to October 24, 2015 is used for the forecasting test. In this test the data of past 7 days was used to forecast the data of the 8th day. Results suggest that ARMA model has certain applicability in forecasting the near-space atmospheric wind. The forecast effect is better when the winds have stronger regularity of change, i.e., when the sample data show a significant 24-hour cycle, the forecast effect is better, and is worse when the winds have a mutation. Compared with the observed data, results show that the forecast error of ARMA model is 9~27m·s-1, and the forecast result of ARMA model is better than that of AR model with the same order, and is slightly better than that of higher-order AR model.