基于全局敏感性分析和贝叶斯方法的WOFOST作物模型参数优化  

Parameters optimization of WOFOST model by integration of global sensitivity analysis and Bayesian calibration method

作　　者：何亮 侯英雨[1] 赵刚[2] 邬定荣[3] 于强[4,5] 

机构地区：[1]国家气象中心,北京100081 [2]德国波恩大学作物科学研究组,波恩D-53115 [3]中国气象科学研究院,北京100081 [4]中国科学院地理科学与资源研究所陆地水循环与地表过程重点实验室,北京100101 [5]澳大利亚悉尼科技大学生命科学学院,悉尼2007

出　　处：《农业工程学报》2016年第2期169-179,共11页Transactions of the Chinese Society of Agricultural Engineering

基　　金：公益性行业(气象)科研专项(GYHY201306052,GYHY201506001)

摘　　要：作物模型参数的敏感性分析、标定和验证可以提高模型的效率和精准度,进而为模型应用做好准备工作。该研究结合参数全局敏感性分析方法以及贝叶斯后验估计理论的马尔科夫蒙特卡洛(Markov Chain Monte Carlo,MCMC)方法,以华北栾城站三年的冬小麦观测数据(叶面积和地上生物量)为参照,对WOFOST模型的55个品种参数进行了敏感性分析、筛选和优化。发现:1)对叶面积影响较大的参数为:生育期为0、0.5、0.6和0.75时的比叶面积、生育期为1.5时的最大光合速率、叶面积指数最大增长率;对地上干物质影响较大的参数为:生育期为1.5时的最大光合速率、生育期为0时的比叶面积、35℃时叶面积的生命周期、生育期为0时的散射消光系数、生育期为1.8时的最大光合速率、储存器官的同化物转换效率。2)潜在和雨养产量水平下,最大叶面积和地上生物量对参数的敏感性差异不大。3)马尔科夫蒙特卡洛方法(MCMC)可以对WOFOST模型品种参数较好地优化;设计的3种校正-验证方案中,第1种方案(用1998-1999年作为校正年份,1999-2000年,2000-2001年作为验证年份)模拟效果最好。4)优化后的参数,模型对潜在产量水平模拟较好,一致性指数均大于0.9,相对均方根误差小于20%;而对有水分胁迫的雨养情况下比潜在产量水平的模拟结果差,表明模型对水分胁迫的模拟不足。该研究为WOFOST模型区域应用和模型调整优化提供科学理论依据。Crop model calibration and validation are essential for model evaluation and application. It is important for model application to accurately estimate the values of crop model parameters and further improve the capacity of model prediction. In the previous researches, trial-and-error method was widely used in model calibration and validation. The deficiency of this method was subjective selection of parameter values and time-consuming processes. To overcome these issues, the optimization methods such as general likelihood uncertainty estimation（GLUE）, genetic algorithm（GA） and shuffled complex evolution（SCE-UA） algorithm were alternative method for model calibration and validation. However, it is a problem to decide which parameters for optimization. It is essential to select the most sensitive parameters among hundreds of parameters in the crop model for optimization. To avoid subjective selection of parameters for calibration and validation, we used the global sensitivity analysis method of model parameters and the Markov Chain Monte Carlo（MCMC） method based on Bayesian theory to optimize the crop genetic parameters in the WOFOST（world food studies）, and the data of three-year winter wheat field experiment in Luancheng in the North China Plain were adopted. The main objectives were： 1） to analyze the sensitivity and uncertainty of WOFOST brought by 55 crop genetic parameters using the extended Fourier amplitude sensitivity test; 2） to calibrate and validate the WOFOST using the MCMC method after sensitivity analysis. We found that： 1） The most sensitive parameters for maximum leaf area index（MAXLAI） in the crop growth period were successively： specific leaf area at development stage of 0, 0.5, 0.6, and 0.75, maximum CO2 assimilation rate at development stage of 1.5, and maximum relative increase in LAI（RGTLAI）; 2） The most sensitive parameters for total above ground production（TAGP） in the crop growth period were successively： maximum CO2 assimilation rate at deve

关 键 词：模型 作物 优化 WOFOST 全局敏感性分析 MCMC 模型参数优化 

分 类 号：S512.1[农业科学—作物学]