主办单位:中国气象局沈阳大气环境研究所
国际刊号:ISSN 1673-503X
国内刊号:CN 21-1531/P

气象与环境学报 ›› 2017, Vol. 33 ›› Issue (1): 34-43.doi: 10.3969/j.issn.1673-503X.2017.01.005

• 论文 • 上一篇    下一篇

基于逐时降水站点资料空间插值方法对比研究

许娈1,2, 董美莹1, 陈锋1   

  1. 1. 浙江省气象科学研究所, 浙江 杭州 310008;
    2. 南京信息工程大学气象灾害教育部重点实验室, 江苏 南京 210044
  • 收稿日期:2015-09-25 修回日期:2016-05-30 出版日期:2017-02-28 发布日期:2017-02-28
  • 通讯作者: 董美莹,E-mail:dongmy_zj@163.com。 E-mail:dongmy_zj@163.com
  • 作者简介:许娈,女,1986年生,工程师,主要从事灾害天气研究,E-mail:xuluan_new@126.com。
  • 基金资助:
    国家自然科学基金(41375056)、浙江省气象科技计划项目(2013ZD01)、浙江省科技计划项目(2014C33056)和气象灾害教育部重点实验室(南京信息工程大学)开放课题(KLME1407)共同资助。

Comparison study of spatial interpolation methods based on hourly precipitation data from automatic weather stations

XU Luan1,2, DONG Mei-ying1, CHEN Feng1   

  1. 1. Zhejiang Meteorology Science Institute, Hangzhou 310008, China;
    2. Key Laboratory of Meteorological Disaster, Ministry of Education, Nanjing University of Information Science and Technology, Nanjing 210044, China
  • Received:2015-09-25 Revised:2016-05-30 Online:2017-02-28 Published:2017-02-28

摘要: 自动站降水资料的格点化是目前气象业务及研究亟需解决的问题,空间插值方法是数据格点化最直接有效的方法。本文选取3个降水个例分别作为大范围强降水、大范围弱降水和局地强对流3种类型降水的典型代表,采用8种常用的插值方法,设计3组试验,通过交叉检验对比8种插值方法降水的插值效果及站点密度对插值效果的影响。结果表明:高密度站点试验表明(站点平均距离约为9.0 km),8种插值方法降水的插值效果依次为CR、IW、NN、LP、KR、TL、MQ和SP,8种插值方法之间的差异小于样本间的差异,即插值效果主要取决于站点的分布而不是插值的方法。低密度站点试验表明(站点平均距离约为30.0 km),大范围降水个例(包括大范围强降水和大范围弱降水)插值的均方根误差(Root Mean Square Error,RMSE)显著增大,不同插值方法之间的差异也增大;而局地强对流个例中,插值后均方根误差增大幅度较小,不同插值方法之间的差异较小。利用CR、IW和NN等3种插值方法开展站点密度敏感性试验,试验表明站点密度提高有利于减小降水插值误差,但站点平均距离达13.0 km甚至更密时,降水误差减小的趋势变缓。

关键词: 逐时降水, 空间插值, 站点密度, 敏感性

Abstract: The gridding of precipitation data from automatic weather stations is of the urgent need to be dealt with in current meteorological service and research,and spatial interpolation is the most direct and efficient method to fulfill it.In this study,three cases were chosen to be the typical representations of wide range intense rainfall,wide range weak rainfall and local convective weather,respectively.Based on the three sets of experiments,eight common methods for interpolation are adopted to compare the interpolation results of various methods and the influence due to station densities.The first set of experiments which is based on high-density stations (the average distance between two stations is approximately 9 km) shows that the interpolation result of CR is the best,followed by IW,NN,LP,KR,TL,MQ,SP in that order.The differences among various interpolation methods are less significant than those among samples,suggesting that it is station distributions,instead of interpolation methods,that play a dominant role in the interpolation results.The second set of experiments which is based on low-density stations (the average distance between two stations is nearly 30 km) shows that the values of RMSE (root mean square error) increase significantly,so do the differences among eight methods.The values of RMSE have a slight increase after interpolation,and the results among various methods are similar in the case of local strong convection.Sensitivity experiments are carried out using the three interpolation methods such as CR,IW,and NN,indicating that RMSE will diminish with the increase of station densities,but the trend will slow down in case that the average station distance reduces to no more than 13 km.

Key words: Hourly precipitation, Spatial interpolation, Station density, Sensitivity

中图分类号: