基于鲸鱼算法的旱地春小麦农田土壤N_(2)O排放模型参数优化  

Optimization of N_(2)O Emission Parameters in Dryland Spring Wheat Farmland Soil Based on Whale Optimization Algorithm

作　　者：牟树佳 董莉霞[1] 李广[2] 燕振刚[1] 逯玉兰 MU ShuJia;DONG LiXia;LI Guang;YAN ZhenGang;LU YuLan(School of Information Science and Technology,Gansu Agricultural University,Lanzhou 730070;School of Forestry,Gansu Agricultural University,Lanzhou 730070)

机构地区：[1]甘肃农业大学信息科学技术学院,兰州730070 [2]甘肃农业大学林学院,兰州730070

出　　处：《中国农业科学》2025年第3期537-547,共11页Scientia Agricultura Sinica

基　　金：国家自然基金项目(32360438);甘肃省高校教师创新基金项目(2025A-094);甘肃省拔尖领军人才项目(GSBJLJ-2023-09);甘肃省财政专项(GSCZZ 20160909)。

摘　　要：【目的】采用鲸鱼优化算法(Whale Optimization Algorithm,WOA)对APSIM模型中与土壤N_(2)O排放相关的默认参数进行优化,以提高模型在模拟我国西北半干旱农业区土壤N_(2)O排放的准确性和适用性,为精确评估和管理农业活动中的温室气体排放提供支持。【方法】采用2020—2021年甘肃省定西市安定区安家坡旱农综合长期定位试验站测定的田间试验数据,结合气象局提供的1970—2021年气象数据,对APSIM模型中N_(2)O形成阶段的4个关键参数(土壤硝化潜力nitrification_pot、铵态氮在半最大利用效率时的浓度nh4_at_half_pot、反硝化系数dnit_rate_coeff、反硝化水系数计算的幂项dnit_wf_power),采用鲸鱼优化算法(WOA)进行单目标多参数优化。通过比较APSIM模型默认参数模拟值、优化参数模拟值和实测值的误差,评估优化后的APSIM土壤N_(2)O排放模型的准确性。【结果】通过多次执行优化程序,最终获得了4个参数的最优组合。其中,土壤硝化潜力为7.62 mg·kg^(-1)·d^(-1)、铵态氮在半最大利用效率时的浓度为49.3 mg·kg^(-1)、反硝化系数为0.00063、反硝化水系数计算的幂项为0.64。与APSIM模型的默认参数相比,决定系数R 2从0.432提升至0.719,均方根误差RMSE从39.42μg·m^(-2)·h^(-1)减少至25.37μg·m^(-2)·h^(-1),归一化均方根误差NRMSE从18.51%下降至11.92%。鲸鱼算法在优化过程中展现出显著的全局搜索能力和快速收敛性。优化后的APSIM模型在模拟土壤N_(2)O排放方面的精度显著提高,表明该方法可以实现对模型参数的快速准确率定。【结论】通过应用鲸鱼算法,4个关键参数得到了精确调整,使得模型的预测误差显著减小,明显地改善了APSIM模型在模拟土壤N_(2)O排放的性能。优化后的模型在我国西北半干旱农业区表现出更高的准确性和适用性,同时也证明了该优化策略的有效性。【Objective】In order to improve the simulation accuracy of N_(2)O emissions by using APSIM model,this study used Whale Optimization Algorithm(WOA)to optimize the default parameters related to soil N_(2)O emissions in the APSIM model to improve the accuracy and applicability of the model in simulating soil N_(2)O emissions in the semi-arid agricultural region of northwest China,for providing support for precise assessment and management of greenhouse gas emissions in agricultural activities.【Method】This study used field experimental data measured by the Anjiapo integrated long-term positioning test station in Anding District,Dingxi City,Gansu Province from 2020 to 2021,combined with meteorological data provided by the Meteorological Bureau from 1970 to 2021,to optimize the four key parameters of N_(2)O formation stage in the APSIM model(soil) nitrification potential(nitration_pot),concentration of ammonium nitrogen at semi maximum utilization efficiency(nh4_at-half_pot),denitrification coefficient(dnit_rate-coeff),and power term of denitrification water coefficient(dnit_wf_power)using the WOA for single objective and multi parameter optimization.The accuracy of the optimized APSIM soil N_(2)O emission model was evaluated by comparing the errors between the default parameter simulation values,optimized parameter simulation values,and measured values of the APSIM model.【Result】Through multiple executions of the optimization program,the optimal combination of four parameters was ultimately determined.Among them,the soil nitrification potential was 7.62 mg·kg^(-1)·d^(-1),the concentration of ammonia nitrogen at semi maximum utilization efficiency was 49.3 mg·kg^(-1),the denitrification coefficient was 0.00063,and the power term of the denitrification water coefficient calculation was 0.64.Compared with the default parameters of the APSIM model,the coefficient of determination R2 increased from 0.432 to 0.719,the root mean square error(RMSE)decreased from 39.42 to 25.37μg·m^(-2)·h^(-1),and the normaliz

关 键 词：N_(2)O排放 APSIM模型 参数优化 鲸鱼优化算法 春小麦 旱地土壤 

分 类 号：S512.12[农业科学—作物学] TP18[自动化与计算机技术—控制理论与控制工程]