利用RHtest提供的两种检验方法:惩罚最大T检验(The Penized Maximal T Test,PMTT)和惩罚最大F检验(The Penized Maximal F Test,PMFT),对新疆大风区域13个国家级气象站点建站至2014年逐年平均风速资料进行了均一性检验和订正;通过在新疆地区特定环境下对该两种误报率分布相对均匀且微小偏移检测能力较高的方法进行评估,找出每种方法在新疆地区实际应用当中的适用性,最终通过订正改善新疆风要素年值历史序列的均一性,结果表明:(1)站址迁移、仪器换型、测风手段变化及台站环境变化等非气候变化因素均对13个待检站点年平均风速序列的均一性产生显著影响,其中仪器换型造成的影响尤为突出。(2)当参考站均一性较好时,PMTT方法检测结果与台站元数据一致性较好,反之则存在虚假跳变点或漏检。(3)PMFT方法因为不依靠参考站,能很好的弥补新疆大风区域站点稀疏、参考站点筛选困难和元数据不完整等弊端,该方法对迁站和仪器换型的检测能力较强。(4)数据订正后,总体变化趋势没有大的改变,呈现下降趋势,符合风要素在全国范围的总体变化趋势,数据的均一性得到很大改善。
英文摘要:
Two tests of RHtest method,the Penalized Maximal T Test(PMTT)and the Penalized Maximal F Test(PMFT),were used in this paper to conduct homogeneity test and data correction for the annual average wind speed data of 13 national meteorological stations in gale areas of Xinjiang,China from their establishment to 2014. The introduction of empirical penalty function has greatly improved the non-uniform distribution of false alarm rate and the algorithms test capability. PMTT is a test method that depends on reference sequence and uses the average offset degrees of the time buckets before and after different nodes in the sequence to find the discontinuity point. Due to the consideration of the relative position among different nodes,this method can eliminate the impact of different sample lengths on test results. While PMFT does not require a reference sequence in the test, but was proposed based on a series improvement of the two-phase regression statistical test,which can avoid effectively the test error due to non-uniform reference sequence. In this paper,comparative analysis was conducted between the results of two test methods,and finally created the corrected data sequence. Results show as follows: (1)the factors of non-climate change such as station relocation,instrument model change,anemometry means change and station environment change had big impacts on the homogeneity of annual average wind speed sequences of the 13 stations in the wind areas of Xinjiang,of which the instrument model change had the biggest impact.(2)According to PMFT,all the 13 stations had significant trip points at the significance level of 95%;while for PMTT method,only 10 stations met the testing requirements under the constraints of reference station distribution and factor dependency,and even appeared false trip points.(3)When the reference station was of good homogeneity,the test results of PMTT method was in good consistency with the station metadata,and showed good test effect for subtle change of the instrument height and environment;otherwise false trip points or omission appeared. PMFT method,as its independence from the reference station,can make up well the defects caused by sparsely-located stations,difficulties in selecting reference station,and incomplete metadata. In addition to relocated station and instrument model change factors,this method also had good test effect for the impact caused by instrument height change and observation site environment change.(4)After corrected,the annual average wind speed sequences of the 12 stations except for that of Shisanjianfang Station overall showed no big trend change,which all showed a downward trend in line with that of the whole country. The corrected sequences collection is of great reference value for environmental assessment and wind energy application of Xinjiang in the future.