作 者：
Cao Hong-xin;Hanan, Jim Scott;Liu Yan;Liu Yong-xia;Yue Yan-bin;Zhu Da-wei;Lu Jian-fei;Sun Jin-ying;Shi Chun-lin;Ge Dao-kuo;Wei Xiu-fang;Yao An-qing;Tian Ping-ping;Bao Tai-lin
单 位：
Yangzhou Univ, Coll Agr, Yangzhou 225009, Peoples R China;Univ Queensland, Ctr Biol Informat Technol, Indooroopilly, Qld 4068, Australia;Jiangsu Acad Agr Sci, Inst Agr Econ & Informat, Engn Res Ctr Digital Agr, Nanjing 210014, Peoples R China;Yangtze Univ, Agron Coll, Jinzhou 434025, Peoples R China;Nanjing Agr Univ, Agron Coll, Nanjing 210095, Jiangsu, Peoples R China;Agr Technol Extens Stn Luntai Cty, Luntai 841600, Peoples R China
关键词：
crop models;validation methods;comparison
摘 要：
In this paper, the many indices used in validation of crop models, such as RMSE (root mean square errors), S-d (standard error of absolute difference), d(a) (mean absolute difference), d(ap) (ratio of d(a) to the mean observation), r (correlation), and R-2 (determination coefficient), are compared for the same rice architectural parameter model, and their advantages and disadvantages are analyzed. A new index for validation of crop models, d(ap) between the observed and the simulated values, is proposed, with d(ap)<5% as the suggested standard for precision of crop models. The different kinds of validation methods in crop models should be combined in the following aspects: (1) calculating d(a) and d(ap); (2) calculating the RMSE or S-d; (3) calculating r and R-2, at the same time, plotting 1:1 diagram.
