地球信息科学学报 ›› 2015, Vol. 17 ›› Issue (9): 1054-1062.doi: 10.3724/SP.J.1047.2015.01055

• • 上一篇    下一篇

东北地区TRMM数据降尺度的GWR模型分析

刘小婵(), 张洪岩*(), 赵建军, 郭笑怡, 张正祥, 朴梅花   

  1. 东北师范大学地理科学学院,长春 130024
  • 收稿日期:2014-09-04 修回日期:2015-06-15 出版日期:2015-09-10 发布日期:2015-09-07
  • 通讯作者: 张洪岩 E-mail:liuxc225@nenu.edu.cn;zhy@nenu.edu.cn
  • 作者简介:

    作者简介:刘小婵(1990-),女,福建厦门人,硕士生,研究方向为遥感与GIS应用。E-mail: liuxc225@nenu.edu.cn

  • 基金资助:
    中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室开放基金项目“中国东北地表物候遥感反演及交叉验证”;中国博士后科学基金项目(2014M561272);中央高校基本科研业务费专项资金项目(14QIVJJ025);吉林省博士后科研项目启动经费项目(RB201353);吉林省科技发展计划项目(20150520069JH)

Spatial Downscaling of TRMM Precipitation Data Based on GWR Model in Northeast China

LIU Xiaochan(), ZHANG Hongyan*(), ZHAO Jianjun, GUO Xiaoyi, ZHANG Zhengxiang, PIAO Meihua   

  1. School of Geographical Science, Northeast Normal University, Changchun 130024, China
  • Received:2014-09-04 Revised:2015-06-15 Online:2015-09-10 Published:2015-09-07
  • Contact: ZHANG Hongyan E-mail:liuxc225@nenu.edu.cn;zhy@nenu.edu.cn
  • About author:

    *The author: SHEN Jingwei, E-mail:jingweigis@163.com

摘要:

利用东北地区2000-2010年93个气象站点观测数据作为“真实值”,对TRMM降水数据进行精度验证,发现研究区TRMM降水数据与观测数据之间具有明显的线性相关性,且TRMM降水数据数值偏大于观测值,表明TRMM降水数据在东北地区具有一定的可信度。对东北地区多年平均、2001、2010年的TRMM数据,进行GWR模型降尺度研究,得到1 km的新降水数据,并与全局OLS回归模型进行对比。结果表明:(1)相比全局OLS回归模型,GWR模型的降尺度结果可获得更好的RRMSE,说明GWR模型更适用于东北地区TRMM数据的降尺度研究;(2)东北地区GWR模型的降尺度分析结果与观测数据之间的相关系数在0.44-0.97之间,且分布较分散;(3)经过降尺度的TRMM降水数据,在空间分辨率上有较大提高,能更真实地反映研究区的降水特征,为该数据小尺度的应用研究奠定基础。

关键词: TRMM, 降尺度, GWR, 东北地区

Abstract:

The availability of precipitation data with high spatial resolution is critical for several applications, such as hydrology, meteorology and ecology. The Tropical Rainfall Measuring Mission (TRMM) data sets can provide effective precipitation information, but at a coarse resolution (0.25°). Therefore, it is very necessary to improve its resolution. The existing TRMM-downscaling methods tend to use ordinary linear regression (OLS), which is known as a global model. However, it ignores the local characteristics. In this paper, the relationship between TRMM and Normalized Difference Vegetation Index (NDVI) was explored by using a local regression analysis approach that is known as geographically weighted regression (GWR). The relationship was used to construct the precipitation downscaling model, which then produces 1 km downscaled precipitation data. The OLS model and GWR model were tested for the data of Northeast China from 2000 to 2010. The accuracy of the downscaled data was validated by the observed precipitation data from 93 meteorological stations located in the study area. Some conclusions can be drawn from our study: (1) there is a strong correlation between TRMM data and the observation data obtained from meteorological stations (R = 0.9172). Overall, the TRMM precipitation is higher than the observed data at all stations. (2) Two downscaling methods were applied in this study, and the results show that the downscaled precipitation based on GWR model produces better results. It produces better R values and the reduced RMSE. Thus, the GWR model is more suitable for the spatial downscaling of TRMM. (3) The correlation coefficient between the downscaled precipitation based on GWR model and the observed data is ranging between 0.44 and 0.97, and its spatial distribution is disperse. (4) The downscaled precipitation data improves the spatial resolution (from 0.25° to 1 km), which can better reflect the characteristics of the precipitation in the study area. It could provide more accurate and realistic precipitation data for the studies at small scales.

Key words: TRMM precipitation, spatial downscaling, GWR, Northeast China