依据最小二乘法原理推求作物水分响应模型参数时可能存在参数不合理、模型模拟精度尚可提升、试验处理 少而难于估计模型参数等问题。针对这些问题, 以 Jensen 模型参数推求为例, 引入自由搜索( Free Search, FS) 算法 率定作物水分响应模型参数, 将 FS 单个动物每步探查行走的位置向量作为参数的一组潜在解, 利用 FS 算法的动 物群体迁移行为推求一组最优参数, 从而改善作物水分响应模型的适用性。三个实例研究结果表明: FS 算法概念 清晰, 操作便捷, 可以直接率定作物水分响应模型参数, 不需转换作物水分响应模型的数学形式, 能够有效避免参数 出现负值或大于 1 等不合理现象, 并提升作物水分响应模型的模拟精度; 当难于采用线性回归方法时, 可以考虑利 用 FS 算法推求作物水分响应模型参数。

Using the least squar e metho d to est imate t he parameters of the mo del o f cr op respo nse to w ater ( MCRW) may pres2 ent some pro blems such as unreasonable parameters, unsatisfying simulatio n accuracy, and inest imable par ameter due to insuffi2 cient irrigation ex perimental tr eat ments. T o so lv e these pr oblems, Free Sear ch ( FS ) w as used to estimat e the parameters of Jensen mo del in this paper. The po sitio n v ecto r of each ex plor ator y step of the individual animal of FS was taken as a set of po2 tential solutions o f t he parameter, and a set o f o ptimal parameters w as estimated based on t he animal g ro up mig ration o f the FS alg orithm. Thr ee case studies showed that the FS alg orithm can calibrate dir ectly the parameters o f Jensen mo del w ithout con2 v erting the mo del. FS can also effectively avo id unreasonable phenomenon such as a negative parameter or a parameter g reater than 1, and impr ove the simulation accuracy of the Jensen mo del. When it is difficult to adopt the linear r egr essio n metho d, FS alg orithm can be used to estimate the pa rameters of M CRW

国家重点研发计划( 2016YFC0400101 ) ; 国家自然科学基金( 51009026 ) ; 农业部农业水资源高效利用重点实验室开放课题 (2015002)