基于EOF迭代的自动气象站气温观测资料修复方法

沈王彬; 李昕; 秦正坤; 张冰

doi:10.3878/j.issn.1006-9895.2103.21021

基于EOF迭代的自动气象站气温观测资料修复方法

doi: 10.3878/j.issn.1006-9895.2103.21021

沈王彬^{1, 2,},
李昕³,
秦正坤^{1, 2},
张冰³

1.
南京信息工程大学大气科学学院，南京 210044
2.
南京信息工程大学资料同化研究与应用联合中心，南京 210044
3.
南京气象科技创新研究院/江苏省气象科学研究所，南京 210009

基金项目: 国家重点研发计划资助项目2018YFC1507302，国家自然科学基金青年项目41805076，中国气象科学研究院基本科研业务费专项2019Z006

详细信息

作者简介:
沈王彬，男，1995年出生，博士研究生，主要从事气象资料的质量控制与同化研究。E-mail: wangbinshen@nuist.edu.cn

中图分类号: P413
计量
- 文章访问数: 470
- HTML全文浏览量: 95
- PDF下载量: 113
- 被引次数: 0
出版历程
- 收稿日期: 2021-01-21
- 录用日期: 2021-05-24
- 网络出版日期: 2021-06-03
- 刊出日期: 2022-03-16

Restoration Method for Automatic Station Temperature Observation Data Based on EOF Iteration

Wangbin SHEN^{1, 2
,},
Xin LI³,
Zhengkun QIN^{1, 2},
Bing ZHANG³

1.
School of Atmospheric Science, Nanjing University of Information Science & Technology, Nanjing 210044
2.
Joint Center of Data Assimilation Research and Applications, Nanjing University of Information Science & Technology, Nanjing 210044
3.
Nanjing Joint Institute for Atmospheric Sciences/Jiangsu Institute of Meteorological Sciences, Nanjing 210009

Funds: National Key Research and Development Program of China (Grant 2018YFC1507302), National Natural Science Foundation of China (Grant 41805076), Fundamental Research Funds of the Chinese Academy of Meteorological Sciences (Grant 2019Z006)

摘要

摘要: 全国目前已经建成了近7万个自动气象观测站点，然而自动气象站观测资料一直存在资料质量较低的问题，大量错误资料的存在极大影响了其在气象研究中的应用，因此对错误观测的数据进行准确的修复是一项重要工作。本文利用2019年12月1日00:00至7日23:00（北京时），共168个时次的地面自动站温度观测资料，在利用EOF（Empirical Orthogonal Function）质量控制方法识别异常观测资料的基础上，提出了一种基于迭代EOF分析方法的错误资料修复方法。通过理想修复试验的精度分析表明，新修复方法能够很好地修复错误的地面自动站观测气温，修复方法的误差约为0.48°C。而基于Cressman插值等这一类依赖单点观测信息进行修复的方法更容易受到小尺度信号干扰而引入非自然观测信息，对地面温度的修复误差可以达到1.55°C。实际的修复结果分析也证明新修复方法充分利用了EOF分析方法的时空分离作用和模态正交性特点，通过迭代方法逐步消除错误资料的影响，从而获得了与周边观测资料有更好时空连续性的修复结果。
- 自动气象站 /
- 地面温度 /
- 经验正交函数分解
Abstract: With the construction of about 70,000 automatic weather stations across China, a comprehensively automatic meteorological observation has been realized. However, the real application of this kind of observation always suffers from their low quality. A large number of error data seriously affects the practical application of observation. Therefore, it is a particularly important task to repair these abnormal observations. Using a total of 168 times of hourly surface temperature observations of automatic weather station during December 1–7, 2019, which is provided by the Jiangsu meteorological bureau, a restoration method based on the empirical orthogonal function method is proposed. The accuracy analysis of ideal restoration experiments shows that the new restoration method can well repair wrong observations with an error of about 0.48 degrees centigrade. The methods based on the Cressman interpolation, which rely on a single point observation information, are more vulnerable to small-scale signal interferences and introduce unnatural observation information, with the surface temperature repair error reaching up to 1.55 degrees centigrade. The analysis of the actual repair results also proves that the new repair method makes full use of the time-space separation and modal orthogonality of the EOF analysis method and gradually eliminates the influence of wrong data through an iterative method to obtain better space-time continuity repair results with the surrounding observation data.
- Ground observation automatic station /
- Surface temperature /
- Empirical Orthogonal Function

HTML全文

图 1 地面自动观测站点（黑色点）水平空间分布

Figure 1. Spatial distribution of ground automatic observation stations (black points)

下载: 全尺寸图片幻灯片

图 2 2019年12月1～7日质量控制方法剔除资料量的时间变化曲线

Figure 2. Hourly data count of abnormal data detected by the quality control method during December 1–7, 2019

下载: 全尺寸图片幻灯片

图 3 2019年12月1日11:00（第36时次）局部温度数值（单位：°C）。图中阴影表示地形高度（单位：m），红色为剔除站点

Figure 3. Spatial distribution of the observed temperature (units: °C) at 1100 BJT (Beijing time) December 1, 2019 (36th hour). The shading indicates the terrain (units: m) and the red dots represent the abnormal stations

下载: 全尺寸图片幻灯片

图 4 2019年12月4日00:00～23:00（a）用来重构的气温分布（单位：°C）以及（b）$\left| {T_{x,s}^{{t_1} - 1} - T_{x,s}^{{t_1}}} \right|$随迭代次数的变化曲线

Figure 4. (a) Spatial distribution of temperature (units: °C) for reconstruction and (b) variation of $\left| {T_{x,s}^{{t_1} - 1} - T_{x,s}^{{t_1}}} \right|$ with the number of iterations for station$s$ from 0000 BJT to 2300 BJT on December 4, 2019

下载: 全尺寸图片幻灯片

图 5 重构过程中观测值与重构值的差随迭代次数的变化曲线（单位：°C）。不同颜色分别代表不同的模态

Figure 5. Temperature differences between the observed and the reconstructed temperature varying with the number of iterations (units: °C). Different colors represent different modes

下载: 全尺寸图片幻灯片

图 6 2019年12月4日14:00（第87个时次）EOF质量控制方法识别出的错误站点（红色圆点）分布

Figure 6. Spatial distribution of abnormal data (red points) at 1400 BJT December 4, 2019 (87th hour)

下载: 全尺寸图片幻灯片

图 7 2019年12月4日14:00（第87个时次）观测气温、重构气温以及误差（观测减去重构）的分布（单位：°C）：（a）观测气温；（b）EOF迭代法重构气温；（c）EOF迭代法重构气温的误差；（d）Cressman插值法重构得到的气温；（e） Cressman插值法重构气温的误差

Figure 7. Spatial distributions (units: °C) of observed and restored temperature and their difference distribution at 1400 BJT December 4, 2019 (87th hour): (a) Observed temperature; (b) restored temperature of EOF method; (c) restoration error of EOF method; (d) reconstructed temperature of Cressman interpolation method; (e) reconstruction error of Cressman interpolation method

下载: 全尺寸图片幻灯片

图 8 观测气温与EOF迭代法（黑色点）以及Cressman插值法重构气温（灰色点）散点图（单位：°C）

Figure 8. Scatter plots between the observed and reconstructed temperature of the EOF iteration method (black spots) and the Cressman interpolation method (gray spots), units: °C

下载: 全尺寸图片幻灯片

图 9 EOF迭代法（蓝色实线）以及Cressman插值（红色虚线）的重构误差概率密度曲线（单位：°C）

Figure 9. The probability density function of the reconstruction error for the EOF iteration method (blue solid line) and the Cressman interpolation method (red dotted line) during 1–7 December 2019 (units: °C)

下载: 全尺寸图片幻灯片

图 10 2019年12月4日14:00（第87个时次）（a）局部观测气温、（b）EOF迭代法重构气温以及（c）Cressman插值气温分布，单位：°C。图中黑色阴影表示地形高度（单位：m）

Figure 10. Spatial distribution of (a) observed temperature, (b) restored temperature by the EOF iteration method and (c) restored temperature by the Cressman interpolation method at 1400 BJT on December 4, 2019 (87th hour), units: °C. The shading indicates the terrain (units: m)

下载: 全尺寸图片幻灯片

图 11 2019年12月4日10:00～20:00观测气温（红色）、重构后气温（蓝色）、Cressman插值后（黑色）气温以及周围站点的气温（灰色）序列（单位：°C）

Figure 11. Time series of (a) observed temperature and (b) restored temperature by the EOF iteration method and (c) restored temperature by the Cressman interpolation method during 1000 BJT–2000 BJT on December 4, 2019 (units: °C)

下载: 全尺寸图片幻灯片

参考文献(31)

[1]	Alvera-Azcárate A, Barth A, Sirjacobs D, et al. 2009. Enhancing temporal correlations in EOF expansions for the reconstruction of missing data using DINEOF [J]. Ocean Sci., 5(4): 475−485. doi: 10.5194/os-5-475-2009
[2]	Annan J D, Hargreaves J C. 2013. A new global reconstruction of temperature changes at the Last Glacial Maximum [J]. Climate Past, 9(1): 367−376. doi: 10.5194/cp-9-367-2013
[3]	蔡彦枫, 张灿亨, 黄勇. 2015. 广东沿海地区极值风速空间插值方法的对比研究 [J]. 南方能源建设, 2(S1): 187−192. doi: 10.16516/j.gedi.issn2095-8676.2015.S1.042 Cai Yanfeng, Zhang Canheng, Huang Yong. 2015. Study on the comparison of spatial interpolation methods used to estimate extreme wind speed spatial distribution using irregularly distributed station data in Guangdong coastal area [J]. Southern Energy Construction (in Chinese), 2(S1): 187−192. doi: 10.16516/j.gedi.issn2095-8676.2015.S1.042
[4]	曹丽娟, 严中伟. 2011. 地面气候资料均一性研究进展 [J]. 气候变化研究进展, 7(2): 129−135. doi: 10.3969/j.issn.1673-1719.2011.02.009 Cao Lijuan, Yan Zhongwei. 2011. Progresses in research of homogenization of climate data [J]. Adv. Climate Change Res. (in Chinese), 7(2): 129−135. doi: 10.3969/j.issn.1673-1719.2011.02.009
[5]	陈兴旺. 2011. 自动气象站数据质量控制方法及误差分析 [J]. 科技资讯(23): 129−131. doi: 10.16661/j.cnki.1672-3791.2011.23.206 Chen Xingwang. 2011. Data quality control method and error analysis of automatic meteorological station [J]. Sci. Technol. Inf. (in Chinese)(23): 129−131. doi: 10.16661/j.cnki.1672-3791.2011.23.206
[6]	Cressman G P. 1959. An operational objective analysis system [J]. Mon. Wea. Rev., 87(10): 367−374. doi:10.1175/1520-0493(1959)087<0367:AOOAS>2.0.CO;2
[7]	Duan S B, Li Z L, Wang N, et al. 2012. Evaluation of six land-surface diurnal temperature cycle models using clear-sky in situ and satellite data [J]. Remote Sensing of Environment, 124: 15−25. doi: 10.1016/j.rse.2012.04.016
[8]	Everson R, Cornillon P, Sirovich L, et al. 1997. An empirical eigenfunction analysis of sea surface temperatures in the western North Atlantic [J]. J. Phys. Oceanogr., 27(3): 468−479. doi:10.1175/1520-0485(1997)027<0468:AEEAOS>2.0.CO;2
[9]	Göttsche F M, Olesen F S. 2001. Modelling of diurnal cycles of brightness temperature extracted from METEOSAT data [J]. Remote Sens. Environ., 76(3): 337−348. doi: 10.1016/S0034-4257(00)00214-5
[10]	江志红, 黄群, 李庆祥. 2008. 近50年中国降水序列均一性检验与订正研究 [J]. 气候与环境研究, 13(1): 67−74. doi: 10.3878/j.issn.1006-9585.2008.01.09 Jiang Zhihong, Huang Qun, Li Qingxiang. 2008. Study of precipitation series homogeneous adjustment and their correction over China in the last 50 years [J]. Climatic Environ. Res. (in Chinese), 13(1): 67−74. doi: 10.3878/j.issn.1006-9585.2008.01.09
[11]	李良富, 王汉杰, 刘金玉, 等. 2006. 基于黑板模型的地面气象数据质量控制 [J]. 气象科技, 34(2): 199−204. doi: 10.3969/j.issn.1671-6345.2006.02.018 Li Liangfu, Wang Hanjie, Liu Jinyu, et al. 2006. Surface meteorological data quality control based on blackboard model [J]. Meteor. Sci. Technol. (in Chinese), 34(2): 199−204. doi: 10.3969/j.issn.1671-6345.2006.02.018
[12]	李天祺, 朱秀芳, 潘耀忠, 等. 2015. MODIS陆地表面温度数据重构方法研究[J]. 北京师范大学学报(自然科学版), 51(S1): 70–76. Li Tianqi, Zhu Xiufang, Pan Yaozhong, et al. 2015. Study on reconstruction methods of MODIS LST products[J]. J. Beijing Norm. Univ Nat. Sci. ) (in Chinese), 51(S1): 70–76. doi: 10.16360/j.cnki.jbnuns.2015.s1.011
[13]	Lorenc A C, Hammon O. 1988. Objective quality control of observations using Bayesian methods, theory, and a practical implementation [J]. Quar. J. Roy. Meteor. Soc., 114(480): 515−543. doi: 10.1002/qj.49711448012
[14]	Qin Z K, Zou X, Li G, et al. 2010. Quality control of surface station temperature data with non-Gaussian observation-minus-background distributions [J]. J. Geophys. Res., 115(D16): D16312. doi: 10.1029/2009JD013695
[15]	邵宇行, 秦正坤, 李昕. 2021. 基于EOF的高时空分辨率自动站温度观测资料质量控制 [J]. 大气科学学报, 已录用. Shao Yuhang, Qin Zhengkun, Li Xin. 2021. High spatial and temporal resolution automatic station temperature observation data quality control based on EOF [J]. Trans. Atmos. Sci. (in Chinese), in press.
[16]	Shen W B, Qin Z K, Lin Z H. 2019. A new restoration method for Radio frequency interference effects on AMSR-2 over North America [J]. Remote Sens., 11(24): 2917. doi: 10.3390/rs11242917
[17]	盛峥, 石汉青, 丁又专. 2009. 利用DINEOF方法重构缺测的卫星遥感海温数据 [J]. 海洋科学进展, 27(2): 243−249. doi: 10.3969/j.issn.1671-6647.2009.02.015 Sheng Zheng, Shi Hanqing, Ding Youzhuan. 2009. Missing satellite-based sea surface temperature data reconstructed by DINEOF Method [J]. Adv. Mar. Sci. (in Chinese), 27(2): 243−249. doi: 10.3969/j.issn.1671-6647.2009.02.015
[18]	Sorjamaa A, Lendasse A. 2007. Time series prediction as a problem of missing values: Application to ESTSP2007 and NN3 competition benchmarks[C]//Proceedings of 2007 International Joint Conference on Neural Networks. Orlando, FL, USA: IEEE, 2948–2953. doi: 10.1109/IJCNN.2007.4371429
[19]	陶士伟, 仲跻芹, 徐枝芳, 等. 2009. 地面自动站资料质量控制方案及应用 [J]. 高原气象, 28(5): 1202−1209. Tao Shiwei, Zhong Qiqin, Xu Zhifang, et al. 2009. Quality control schemes and its application to automatic surface weather observation system [J]. Plateau Meteor. (in Chinese), 28(5): 1202−1209.
[20]	王海军, 杨志彪, 杨代才, 等. 2007. 自动气象站实时资料自动质量控制方法及其应用 [J]. 气象, 33(10): 102−109. doi: 10.3969/j.issn.1000-0526.2007.10.015 Wang Haijun, Yang Zhibiao, Yang Daicai, et al. 2007. The method and application of automatic quality control for real time data from automatic weather stations [J]. Meteor. Mon. (in Chinese), 33(10): 102−109. doi: 10.3969/j.issn.1000-0526.2007.10.015
[21]	王晓春, 蔡雅萍, 梁幼林. 1994. 用Bivariate-Test方法分析温度资料的不均一性 [J]. 大气科学, 18(S1): 856−867. doi: 10.3878/j.issn.1006-9895.1994.z1.10 Wang Xiaochun, Cai Yaping, Liang Youlin. 1994. To detection of the inhomogeneity of yearly-mean surface temperature data using the Bivariate-Test Method [J]. Chinese Journal of Atmospheric Sciences (in Chinese), 18(S1): 856−867. doi: 10.3878/j.issn.1006-9895.1994.z1.10
[22]	Wu P H, Yin Z X, Yang H, et al. 2019. Reconstructing geostationary satellite land surface temperature imagery based on a multiscale feature connected convolutional neural network [J]. Remote Sens., 11(3): 300. doi: 10.3390/rs11030300
[23]	徐浩然. 2019. 地面气象观测自动化存在的弊端及其对策 [J]. 河南科技(22): 151−152. doi: 10.3969/j.issn.1003-5168.2019.22.053 Xu Haoran. 2019. The disadvantage and countermeasure of surface meteorological observation automation [J]. Henan Sci. Technol. (in Chinese)(22): 151−152. doi: 10.3969/j.issn.1003-5168.2019.22.053
[24]	薛纪善. 2009. 气象卫星资料同化的科学问题与前景 [J]. 气象学报, 67(6): 903−911. doi: 10.3321/j.issn:0577-6619.2009.06.001 Xue Jishan. 2009. Scientific issues and perspective of assimilation of meteorological satellite data [J]. Acta Meteor. Sinica (in Chinese), 67(6): 903−911. doi: 10.3321/j.issn:0577-6619.2009.06.001
[25]	叶小岭, 施林红, 熊雄, 等. 2016. AI方法在地面气温观测资料质量控制中的应用 [J]. 气候与环境研究, 21(1): 1−7. doi: 10.3878/j.issn.1006-9585.2015.15073 Ye Xiaoling, Shi Linhong, Xiong Xiong, et al. 2016. Application of AI method to quality control in surface temperature observation data [J]. Climatic Environ. Res. (in Chinese), 21(1): 1−7. doi: 10.3878/j.issn.1006-9585.2015.15073
[26]	张慧. 2012. 分析自动站资料对气象预报的重要性 [J]. 农业与技术, 32(12): 124. doi: 10.3969/j.issn.1671-962X.2012.12.106 Zhang Hui. 2012. The importance of analyzing data from automatic stations for weather Forecasting [J]. Agric. Technol. (in Chinese), 32(12): 124. doi: 10.3969/j.issn.1671-962X.2012.12.106
[27]	张颖超, 姚润进, 熊雄, 等. 2017. PSO-PSR-ELM集成学习算法在地面气温观测资料质量控制中的应用 [J]. 气候与环境研究, 22(1): 59−70. doi: 10.3878/j.issn.1006-9585.2016.16013 Zhang Yingchao, Yao Runjin, Xiong Xiong, et al. 2017. Application of PSO-PSR-ELM-Based ensemble learning algorithm in quality control of surface temperature observations [J]. Climatic Environ. Res. (in Chinese), 22(1): 59−70. doi: 10.3878/j.issn.1006-9585.2016.16013
[28]	Zhao J B, Lange H, Meissner H. 2020. Gap-filling continuously-measured soil respiration data: A highlight of time- series-based methods [J]. Agric. Forest Meteor., 285–286: 107912. doi: 10.1016/j.agrformet.2020.107912
[29]	赵煜飞, 任芝花, 张强. 2011. 适用于全国气象自动站正点相对湿度资料的质量控制方法 [J]. 气象科学, 31(6): 687−693. doi: 10.3969/j.issn.1009-0827.2011.06.003 Zhao Yufei, Ren Zhihua, Zhang Qiang. 2011. Quality control procedures for hourly relative humidity data from national automatic weather stations [J]. Scientia Meteor. Sinica (in Chinese), 31(6): 687−693. doi: 10.3969/j.issn.1009-0827.2011.06.003
[30]	周芳成, 唐世浩, 韩秀珍, 等. 2021. 云下遥感地表温度重构方法研究 [J]. 国土资源遥感, 33(1): 78−85. Zhou Fangcheng, Tang Shihao, Han Xiuzhen, et al. 2021. Research of reconstructing missing remotely sensed land surface temperature data in cloudy sky [J]. Remote Sens. Land Resour. (in Chinese), 33(1): 78−85.
[31]	朱江, 徐启春, 王赐震, 等. 1995. 海温数值预报资料同化试验Ⅰ. 客观分析的最优插值法试验 [J]. 海洋学报, 17(6): 9−20. Zhu Jiang, Xu Qichun, Wang Cizhen, et al. 1995. Data assimilation experiment of SST numerical forecast. Part I. Experiment of optimal interpolation method for objective analysis [J]. Acta Oceanol. Sinica (in Chinese), 17(6): 9−20.