地理研究 ›› 2008, Vol. 27 ›› Issue (5): 1161-1168.doi: 10.11821/yj2008050019

• 气候与全球变化 • 上一篇    下一篇

一种改进的生成区域日降水场的方法及精度分析

林忠辉, 莫兴国   

  1. 中国科学院地理科学与资源研究所 生态网络观测与模拟重点实验室,北京 100101
  • 收稿日期:2007-08-22 修回日期:2008-07-11 出版日期:2008-09-25 发布日期:2008-09-25
  • 作者简介:林忠辉(1971-),男,四川德阳人,副研究员。主要从事生态系统模拟模型的区域应用研究。Email:linzh@igsnrr.ac.cn.
  • 基金资助:

    国家自然科学基金项目 (40671033 );国家"863"计划项目(2006AA10Z228)和中国科学院地理科学与资源所领域前沿项目(066U0607SZ)资助。

Daily precipitation interpolation overChina with DAYMET model

LIN Zhong-hui, MO Xing-guo   

  1. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • Received:2007-08-22 Revised:2008-07-11 Online:2008-09-25 Published:2008-09-25
  • Supported by:

    国家自然科学基金项目 (40671033 );国家"863"计划项目(2006AA10Z228)和中国科学院地理科学与资源所领域前沿项目(066U0607SZ)资助。

摘要:

利用全国687个气象站点11年的日降水数据,对基于地理特征和统计回归的函数拟合类模型DAYMET生成中国区域日降水场的能力进行了验证。交叉验证表明,DAYMET模型估计日降水累计得到的年降水量的绝对偏差11年平均为29.8%,年降水总量估计偏差低于20%的站点占48.3%。鉴于中国陆地区域降水深受季风的影响,不同方位气象站点对插值点的影响也有所不同,引入了站点不同方位对插值的影响权重,对DAYMET模型进行了改进,改进后年降水量的绝对偏差降为27%。与梯度距离平方反比法相比,该方法具有较高的区域降水插值精度。还以无定河流域降水插值为例,说明降水插值精度的高低与区域内雨量站点的多寡紧密相联。

关键词: 降水, 空间插值, DAYMET模型

Abstract:

Based on daily records of precipitation from 687 weather stations over China from 1990 to 2000, we interpolate and validate daily precipitation interpolation using DAYMET. The technique interpolates on a systematic grid daily precipitation from surrounding stations based on the principles of a weighted Gaussian filter. In this study we calibrated DAYMET using the China climate data base and modified DAYMET considering the different effects of the dimensional orientation of surrounding stations on the local point interpolation, which allows us to interpolate daily precipitation for any location within the country. We validated the technique using crossing validation, and compared the MAE to the result of the methods named gradient plus inverse distance squared (GIDS). Our results can be summarized as follows: without modification, the mean absolute error(MAE) of daily precipitation interpolation using DAYMET is 1.8mm, the value for GIDS is 2.2mm,and the MAE for annual total precipitation is 29.8% for DATMET. After modification, the MAE of annual precipitation for DAYMET is 27%. The reason for considering high MAE of the two methods is that the weather stations in China are sparsely distributed in space, especially in Western China. To decrease the MAE of interpolation, the addition of the precipitation observations from hydrological station network should be considered. We chose Wudinghe Basin located in Western China with an area of 30261 km2, only 3 weather stations, but 93 precipitaition stations of hydrological station network, to test it. With the surrounding 30 climate stations, the interpolation MAE of annual precipitation is 77mm(28%) for DAYMET,but aided by precipitation stations of hydrological station network, the MAE is 48mm(12.1%)。

Key words: precipitation, spatial interpolation, DAYMET model