温度升高下APSIM模型春小麦籽粒生长参数敏感性分析及优化

Sensitivity analysis and optimization of spring wheat grain growth parameters under APSIM model with the increase of temperature

为有效识别基于APSIM模型籽粒生长参数中春小麦产量敏感性参数,快速并准确的估算当地模型参数。使用甘肃省定西市安定区凤翔镇安家沟村1971—2018年的气象数据和2000—2018年旱地春小麦大田试验数据,并利用EFAST方法对进行了5个增温梯度(0℃、0.5℃、1.0℃、1.5℃和2.0℃)下32个模型参数进行敏感性分析。粒子群算法对各个增温条件下均敏感的参数进行优化验证。结果表明:不同温度变化梯度下,对旱地春小麦产量影响最大的籽粒生长模型参数有9个,分别为消光系数、每克茎籽粒数量、穗粒数、单株最大籽粒质量、灌浆到成熟积温、出苗到拔节积温、株高、最大比叶面积和光合叶片老化的水分胁迫斜率。并且对产量敏感性强度有着显著的差异,其中消光系数和每克茎籽粒数量是对春小麦产量影响最大的参数,其他参数在不同温度下对春小麦产量的敏感性顺序存在差异。利用粒子群算法针对这9个参数进行优化,相较于优化前,优化后的春小麦产量、开花期和灌浆期籽粒干物质的均方根误差、归一化均方根误差和模型有效性指数均得到了显著改善,参数优化后开花期、灌浆期、成熟期产量的均方根误差平均值分别由13.50 kg hm-2减小到5.99 kg hm-2、183.17 kg hm-2减小到69.44 kg hm-2、141.69 kg hm-2减小到48.51 kg hm-2,归一化均方根误差平均值分别由4.94%减小到2.19%、10.92%减小到4.65%、8.39%减小到2.87%,模型有效性指数平均值分别由0.894提高到0.979、0.893提高到0.981、0.898提高到0.988。优化后的参数有效地提高了模型的预测精度。此研究为APSIM模型在当地应用和模型参数校准提供了科学依据。

In order to effectively identify spring wheat yield sensitivity parameters in grain growth parameters based on APSIM model,the local model parameters were quickly and accurately estimated.Using the meteorological data of Anjiagou Village,Fengxiang Town,Anding District,Dingxi City,Gansu Province from 1971 to 2018 and the field test data of dryland spring wheat from 2000 to 2018,the sensitivity analysis of 32 model parameters under five temperature gradients(0℃,0.5℃,1.0℃,1.5℃,and 2℃)was conducted by EFAST method.The particle swarm optimization algorithm is used to optimize and verify the parameters sensitive at all temperatures.The results showed that,under different temperature gradients,there were nine grain growth model parameters that had the greatest influence on the yield of spring wheat in dry land,which were extinction coefficient,the number of grains per gram of stem,the number of grains per ear,the maximum grain mass per plant,the accumulated temperature from filling to maturity,the accumulated temperature from emergence to jointing,plant height,the maximum specific leaf area,and water stress slope of photosynthetic leaf aging.The sensitivity intensity of spring wheat yield was significantly different,among which extinction coefficient and the number of seeds per gram of stem were the most influential parameters in spring wheat yield,and the sensitivity sequence of other parameters was different under different temperatures.Particle swarm optimization algorithm was used to optimize the nine parameters.Compared with before optimization,the optimized spring wheat yield,root mean square error of grain dry matter,normalized root mean square error and model validity index were significantly improved.After parameter optimization,the optimized spring wheat yield,root mean square error of grain dry matter and model validity index were significantly improved.The mean square error of yield at maturity stage decreased by 13.50–5.99 kg hm–2,183.17–69.44 kg hm–2,and 141.69–48.51 kg hm–2,respectively.The normalized root-mean-square error decreases by 4.94%–2.19%,10.92%–4.65%,8.39%–2.87%,and the average model validity index increases by 0.894–0.979,0.893–0.981,and 0.898–0.988,respectively.The optimized parameters can effectively improve the prediction accuracy of the model.This study provides a scientific basis for the local application of APSIM model and the calibration of model parameters.