基于留一交叉验证法的APSIM-Maize产量模拟  

Yield simulation from APSIM-Maize by using the leave-one-out cross validation approach

作　　者：杨雪宁 张永强[1] 张选泽 马宁 张俊梅 YANG Xue-Ning;ZHANG Yong-Qiang;ZHANG Xuan-Ze;MA Ning;ZHANG Jun-Mei(Key Laboratory of Water Cycle and Related Land Surface Processes,Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences,Beijing 100101,China;Sino-Danish College,University of Chinese Academy of Sciences,Beijing 100101,China;Dalat Banner Agricultural and Animal Husbandry Technology Extension Center,Dalat 014300,Inner Mongolia,China)

机构地区：[1]中国科学院地理科学与资源研究所陆地水循环及地表过程重点实验室,北京100101 [2]中国科学院大学中丹学院,北京100101 [3]达拉特旗农业技术推广中心,内蒙古达拉特014300

出　　处：《作物学报》2023年第10期2854-2860,共7页Acta Agronomica Sinica

基　　金：“科技兴蒙”行动重点专项(十大孔兑综合治理与水资源集约高效利用集成示范)(KJXM-EEDS-2020005);鄂尔多斯科技重大专项(鄂尔多斯陆地生态系统碳储量、碳汇核算及潜力评价)(2022EEDSKJZDZX016)资助。

摘　　要：作物生长模型APSIM广泛应用于作物估产和农业生产管理中,在观测数据有限的情况下,开展留一交叉验证是提高模型模拟能力的关键途径。本研究以内蒙古十大孔兑地区春玉米为研究对象,量化分析了APSIM-Maize模型模拟2012—2019年间玉米产量对关键参数的敏感性,并根据参数敏感性强弱对APSIM-Maize模型进行交叉验证与参数率定,提高了模型模拟能力。主要结果为:(1)影响春玉米产量的敏感性参数由强到弱依次是:蒸腾效率系数、辐射利用效率、开花到成熟的积温、出苗到拔节的积温、开花到灌浆的积温、潜在灌浆速率、光周期和最大穗粒数;(2)交叉验证时,APSIM-Maize模型各参数变异系数在1.06%~23.32%之间波动,总体上模型参数变异性小,可靠性高;(3)APSIM-Maize模型经过参数率定后的模拟产量与实测产量具有较好的一致性(R^(2)=0.72,RMSE=401.5kg hm^(-2)),模型在十大孔兑地区春玉米产量的评估中表现出较好的适应性。本研究为在农田试验数据有限情况下提高模型率定参数的可靠性提供了新的研究思路和科学依据。The Agricultural Production Systems Simulator(APSIM)has been extensively used in crop yield estimation and agri-cultural management.Leave-one-out cross validation is a key way to improve the model simulation capability with the limited filed data.In this study,we evaluate the parameters sensitivity to maize yield and evaluate the reliability of parameters with the leave-one-out cross validation based on the sensitivity rank of parameters.These results showed that:(1)The sensitive parameters for maize yield were ranked in descending order as:transpiration efficiency coefficient,radiation use efficiency,thermal time from flower to maturity,thermal time from emergence to end of juvenile phase,thermal time from flowering to start grain-filling,po-tential grain growth rate,photoperiod slope,and potential grains per head.(2)The coefficients of variation of parameters fluctu-ated from 1.06%to 23.32%,which indicates that the variation of parameter was small and the reliability of parameters was high.(3)The adjusted APSIM-Maize model performed well in simulating spring maize yield(R^(2)=0.72;RMSE=401.5 kg hm^(–2)),which indicates that the model had great adaptability in estimating spring maize yield in the Ten Kongduis.Our study provides an insight to improve the reliability of parameters with limited field data.

关 键 词：APSIM-Maize模型 交叉验证 春玉米 

分 类 号：S513[农业科学—作物学]