旋转经验正交函数分解回归方法在东北夏季气温季节预测和成因诊断中的应用

An application of the REOF and stepwise regression to seasonal forecast and cause diagnosis for summer temperature anomaly in northeast China

  • 摘要: 利用中国东北4省/区73个地面气象观测站1971—2011年6—8月月平均气温,以及NCEP/NCAR再分析1971—2011年月平均高度和NOAA月平均海表温度(SST)资料,基于主成分回归(PCR)预测方法的思想,用同年1—5月北半球大气环流和全球SST场建立了东北地区夏季气温的统计预测模型,该建模方法是主成分回归方法的变形,计算方法较为简易,对气温等级的季节预测有较好的预报效果;并用其计算过程做了前期气候成因诊断。考虑到旋转经验正交函数分解样本误差较小、空间模态结构清晰,但特征向量的时间系数在不同时段有所变化的特点,故使用旋转经验正交函数分解整个时段的东北夏季气温场,然后基于旋转经验正交函数分解结果,进行前期影响因子甄别,最后建立多元线性逐步回归预测模型,建模期为前30年,独立样本预报期为后11年。30 a逐年交叉回报检验和11 a独立样本预测效果都显示该模型具有较高的预报技巧,尤其对气温等级的预测具有参考价值。用预测模型的中间过程诊断了东北夏季各月某一类典型气温异常的前期成因:5月北极涛动和北大西洋涛动(AO/NAO)、北太平洋涛动(NPO)以及副热带纬向一致异常型(SZ)等大气大尺度低频波动对6月吉林和辽宁省气温异常有显著影响;3月北极涛动和北大西洋涛动、东太平洋涛动(EP)及欧亚遥相关型波列对7月内蒙古东北部气温异常有显著影响;5月在北半球中高纬度大尺度低频波列不显著的情况下,SZ型低频波列对8月内蒙古东北部部分地区和黑龙江中、西部等地气温异常有显著影响;前期海温呈厄尔尼诺(拉尼娜)型、同时北大西洋海温三极子为正(负)位相,一般与导致东北6月和7月偏冷(暖)的大气环流型相匹配;春末热带印度洋全区海表温度一致异常模态(IOBM)正(负)位相与导致东北8月偏暖(冷)的大气环流型相匹配。

     

    Abstract: Based on observed monthly surface air temperatures from the 73 stations in northeastern China from June to August during 1971-2011, the NCEP/NCAR reanalysis Ⅰ monthly height at 500 hPa and the NOAA extended reconstructed monthly Sea Surface Temperature (SST) V3b data during 1971-2012, and employing the statistical prediction idea of principal component regression (PCR), the paper constructed a seasonal forecast model of the modified PCR for classified summer temperatures in northeastern China using the atmospheric circulation and SST of January-May this year as predictors, which made a clear explanation on the early climatic causes of the Northeast summer temperature anomalies. Because of the qualities of smaller sampling errors, significant spatiality structure, but flexible time series of eigenvectors for the Rotated Empirical Orthogonal Function (REOF), the whole time-space series data of surface air temperature from 1971-2011 is analyzed by REOF for gaining full predictand information, the predictors are then diagnosed based on the relationships between REOF eigenvectors and preceding climatic conditions, and at finally the forecast model is built by the multiple stepwise regressions. The training period of the forecast model is the former 30 years of the time series, and the last 11 years in the time series are the period of the independenttest samples, in which the cross hindcast validation and independent test all showed skillful, especially effective for the classified temperature forecast.The preceding causes of summer temperature anomalies in northeastern China are able to be found by the processes of the built model, and the results show that the certain configured situation of AO/NAO, NPO, and subtropical zonal mode (SZ) of like-signed anomalous pressure in May is significantly responsible for the surface air temperature anomaly of southeastern Northeast in the following June; and the cooperation of AO/NAO, East Pacific Oscillation, and Eurasian pattern type 1 teleconnection wave in March is responsible for the following surface air temperature anomaly of northeastern Inner Magnolia and western Northeast in July; and subtropical zonal mode pattern associated with no apparent planetary wave trains in the mid-high latitudes in May is responsible for the surface air temperature anomaly in the portion of northeastern Inner Magnolia and the central and western part of Heilongjiang Province in coming August, with both of which located in northern Northeast; and, on the other hand, the causes of SST anomaly show that the anomalous SST distribution like El Niño (La Nia) pattern associated with the North Atlantic Tripole pattern in positive (negative) phase is significantly matched with atmospheric circulations leading to the lower (higher) summertime surface air tempertures in Northeast, and Indian Ocean Basin Mode(IOBM) in positive (negative) phase in late spring is related to the circulations leading to the higher (lower) surface air temperature in Northeast in August.

     

/

返回文章
返回