基于ANFIS的陕西省参考作物蒸散量计算

为有效提高陕西省参考作物蒸散量（ET₀）的计算精度,选取陕西省6个气象站点57 a（1960—2016年）每日气象资料,构建8种基于自适应神经模糊推理系统（ANFIS）的ET₀计算模型,将其与Hargreaves-Samani、Makkink和Iramk等三种在陕西省ET₀计算精度较高的模型进行比较。结果表明：ANFIS模型能较好地反映气象因子与ET₀之间的复杂非线性关系,在仅有气温数据时,ANFIS模型具有足够的精度（平均R²=0.894,平均R_MSE=0.558 mm/d,平均MRE=18.258%）,ANFIS模型的精度随着气象因子数量的增加而增加;在相同气象条件下,ANFIS模型模拟效果最好;ANFIS模型具有较强的泛化能力,基于不同站点也有较高的精度（平均R²=0.974,平均R_MSE=0.276 mm/d,平均MRE=8.608%）,具有很好的可移植性。因此,在缺少气象数据时,ANFIS可以作为陕西省ET₀的计算模型,为农业用水管理及水资源优化配置提供帮助。

Calculation of reference crop evapotranspiration in Shaanxi province based on ANFIS

To effectively improve the prediction accuracy of reference crop evapotranspiration (ET₀) in Shaanxi province, eight ET₀ calculation models based on the ANFIS were constructed. The data used are obtained from 6 meteorological stations in Shaanxi province. The ET₀ calculated by FAO 56 P-M was taken as a standard value. ANFIS was compared with three models, namely, Hargreaves-Samani, Iramk and Makkink. The results show that the ANFIS model can better reflect the complex linear relationship between meteorological factors and ET₀. When only the temperature data can be used, the ANFIS model has sufficient accuracy (R²=0.894, R_MSE=0.558 mm/d, MRE=18.258%), with the increase of meteorological factors, the accuracy of ANFIS model is improved. The simulation accuracy of ANFIS model is higher than that of Hargreaves-Samani, Iramk and Makkink models under the same input condition. The analysis shows that the ANFIS model has strong generalization ability in the province, and it has higher accuracy based on different sites (R²=0.974, R_MSE=0.276 mm/d, MRE=8.608%). Therefore, the ANFIS model can be used as a recommended model for the calculation of ET₀ in Shaanxi in the absence of meteorological data, and has good portability, which can provide a scientific basis for agricultural water management and optimal allocation of water resources.