中国农业气象 ›› 2017, Vol. 38 ›› Issue (07): 417-425.doi: 10.3969/j.issn.1000-6362.2017.07.003

• 论文 • 上一篇    下一篇

基于改进卡尔曼滤波算法的雷达定量降雨估算

曲小康,芮小平,于雪涛,雷秋良   

  1. 1.中国科学院大学资源与环境学院,北京 100049;2.石家庄铁道大学交通运输学院,石家庄 050043;3.中国农业科学院农业资源与农业区划研究所/农业部面源污染控制重点实验室,北京 100081
  • 收稿日期:2016-11-24 出版日期:2017-07-20 发布日期:2017-07-14
  • 作者简介:曲小康(1989-),硕士,研究方向为基于地理信息系统的气象预警方法。E-mail:qxkang@126.com
  • 基金资助:
    河北省自然科学基金“京津冀地区强对流天气下多场因子驱动的风暴体外推方法研究”(D2016210008);河北省社会科学基金“京津冀地区强对流天气预警及应对策略研究”(HB15SH015)

Quantitative Rainfall Estimation Using Weather Radar Based on Improved Kalman Filter Method

QU Xiao-kang, RUI Xiao-ping, YU Xue-tao, LEI Qiu-liang   

  1. 1.College of Resource and Environment, University of Chinese Academy of Sciences, Beijing 100049, China; 2. Transportation Institute, Shijiazhuang Tiedao University, Shijiazhuang 050043; 3.Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences/Key Laboratory of Non-point Source Pollution Control, Ministry of Agriculture, Beijing 100081
  • Received:2016-11-24 Online:2017-07-20 Published:2017-07-14

摘要: 针对雷达定量降雨估算误差较大的问题,本文提出一种使用改进卡尔曼滤波对雷达估算值进行校准的方法。先确立G/R(自动气象站测量值/天气雷达估算值)校准因子模型,并应用普通卡尔曼滤波方法对G/R校准因子建立预测系统和测量系统,同时引入系统参数的校准过程和系统误差的自适应估计过程,动态调整卡尔曼滤波中各项参数值;最后将滤波后的G/R因子用于校正雷达定量降雨估算,得到较准确的降雨估算值。利用长春市天气雷达2015年8月19-20日和2016年月8月6-7日两次降雨过程的雷达产品和加密自动站逐小时的降雨资料,对卡尔曼滤波方法进行检验分析。结果表明:改进卡尔曼滤波和普通卡尔曼滤波校准后雷达降雨估算结果优于未校准的降雨估算结果,普通卡尔曼滤波方法和改进卡尔曼滤波方法的平均相对误差分别从0.6047减至0.3557和0.2645,从0.8052 减至0.3096和0.1715,且改进算法效果优于普通卡尔曼滤波算法,校准后雷达降雨估算准确度明显提高。

关键词: G/R校准因子, 改进卡尔曼滤波, 定量降雨估算, 自适应估算

Abstract: To minimum the error of radar rainfall evaluation, an improved Kalman filter method was presented to calibrate the radar quantitative rainfall estimation (QRE). Firstly, the G/R (rain gauge rain rate/radar rain rate) calibration factor model was established. Secondly, the prediction and measurement system of G/R was set up based on the Kalman filter (KF). The calibration process of system parameters and adaptive estimation process of system error was introduced to adjust the parameters of KF dynamically. Thirdly, the G/R calibration ratio was used to correct radar quantitative rainfall estimation. The radar and rain gauge hourly rain data of two rain cases on 2015-08-19-20 and 2016-08-06-07 from Changchun were used to test the efficiency of the proposed method. The results showed that the QRE result with KF calibration was better than that without calibration. And the average relative errors of two rain cases were reduced from 0.6047 to 0.3557 and 0.2645, from 0.8052 to 0.3096 and 0.1715 by ordinary KF and improved KF respectively. Moreover, the improved KF was even better than the ordinary KF.

Key words: G/R ratio, Improved Kalman filter, Quantitative rainfall estimation, Adaptive estimation