基于自适应模糊推理系统的汾河流域参考作物蒸散量模拟研究  

Simulation research of reference crop evapotranspiration based on ANFIS in Fenhe basin of China

作　　者：葛杰 刘园[1,2] 雒舒琪 曹绮欣 GE Jie;LIU Yuan;LUO Shuqi;CAO Qixin(Power China Northwest Engineering Corporation Limited,Xi’an 710065,China;Shaanxi Provincial Water Ecological Environment Engineering Technology Research Center,Xi’an 710065,China;Northwest A&F University,Yangling 712100,China)

机构地区：[1]中国电建集团西北勘测设计研究院有限公司,西安710065 [2]陕西省水生态环境工程技术研究中心,西安710065 [3]西北农林科技大学,陕西杨凌712100

出　　处：《灌溉排水学报》2024年第1期69-77,共9页Journal of Irrigation and Drainage

基　　金：国家自然科学基金面上项目(51979222);陕西省科技统筹创新计划项目(2016KTZDNY-01-01)。

摘　　要：【目的】有效提高缺少气象资料条件下汾河流域参考作物蒸散量(ET0)计算精度。【方法】选取汾河流域及附近7个气象站点1960—2017年逐日气象资料,根据不同气象要素组合,构建16种基于自适应模糊推理系统(ANFIS)的ET0模拟模型,并与Hargreaves-Samani、Irmak-Allen、Makkink模型进行比较,评价ANFIS模型在汾河流域的适用性及可移植性。【结果】(1)ANFIS模型能很好地展现ET0与各输入因子之间的非线性关系;仅输入Tmax、Tmin、Ra建立的ANFIS2模型模拟精度(平均R^(2)为0.882,平均NSE为0.876,平均RMSE为0.341 mm/d)能满足使用要求,随着输入气象要素数量的增加,模型模拟精度不断提高;(2)在输入因子相同时,ANFIS模型精度高于Hargreaves-Samani、Irmak-Allen、Makkink模型;(3)ANFIS模型在汾河流域具有很强的泛化能力和可移植性,不同分区建立的ANFIS模型相互移植时具有较高精度(平均R^(2)为0.983,平均NSE为0.978,平均RMSE为0.134 mm/d)。【结论】因此,在缺少气象资料时,基于自适应模糊推理系统建立的ET0模拟计算模型可作为汾河流域ET0计算的推荐模型。【Objective】In order to effectively improve the calculation accuracy of reference crop evapotranspiration(ET0)in Fenhe basin,【Method】Daily meteorological data in the past 58 years(1960—2017)were collected from seven meteorological stations in Fenhe basin and it’s vicinity.According to the combinations of different meteorological elements,16 ET0 calculation models were built based on adaptive network-based fuzzy inference system(ANFIS),and were compared with Hargreaves-Samani,Irmak-Allen and Makkink,which applicability were evaluated.【Result】①The non-linear relation between ET0 and each input factor can be well displayed by ANFIS model.The simulation accuracy of ANFIS2 model established by only inputting Tmax,Tmin and Ra can meet the usage requirements(R^(2)=0.882,NSE=0.876,RMSE=0.341 mm/d).With the increase of meteorological elements,the simulation accuracy of ANFIS model was improved.②The simulation accuracy of ANFIS model was higher than that of Hargreaves-Samani,Irmak Allen,and Makkink models when in the same input condition.③The ANFIS model had strong generalization ability and portability in the Fenhe basin.High accuracy was shown in mutual portability of ANFIS models established in different zone(R^(2)=0.983,NSE=0.978,RMSE=0.134 mm/d).【Conclusion】Therefore,the model based on adaptive network-based fuzzy inference system(ANFIS)can be used as a recommended model for the calculation of ET0 in the Fenhe basin in the absence of meteorological data,which can provide a technical support for the efficient utilization of agricultural water resources in the Fenhe basin.

关 键 词：汾河流域 参考作物蒸散量 自适应模糊推理系统 可移植性 

分 类 号：P426.2[天文地球—大气科学及气象学]