基于思维进化算法优化的东北地区参考作物蒸散量估算

Estimation of Reference Crop Evapotranspiration in Northeast China Based on Mind Evolutionary Algorithm Optimization

准确估算参考作物蒸散发(Reference crop evapotranspiration,ET0)对于农业水资源管理至关重要。东北地区是我国最重要的粮食产区,但该区域纬度相对较高、气温相对较低,ET0影响因素多、估算的不确定性高。研究选取东北地区20个代表性气象站点1961-2019年气象数据,采用Mann-Kendall非参数趋势检验及反距离加权插值法模拟东北地区ET0时空变化特征,并利用思维进化算法(Mind Evolutionary Algorithm,MEA)优化模型参数,以FAO-56 Penman-Monteith公式计算结果为标准值,比较9种不同输入因子的模型精度,结果表明:(1)1961-2019年东北地区ET0的年平均值在567.81~1 080.66 mm之间,东北地区北部的年均ET0值呈上升趋势,中部平原及南部沿海呈下降趋势;(2)通过对东北地区20个站点使用不同类型模型计算ET0的评估,优化前精度表现:辐射型模型>湿度型模型>温度型模型。其中Mak模型在东北地区的计算精度最高,相应的R^(2)、NSE、RMSE、和MAE中位数值分别为0.801、0.786、0.570 mm/d和0.331 mm/d;(3)MEA算法优化后,对9种经验模型的R^(2)、NSE、RMSE和MAE提升幅度分别为14.43~47.15%、14.84~50.47%、5.42~46.79%、7.47~39.86%。优化后的Mak模型相应的R^(2)、NSE、RMSE、和MAE中位数值分别为0.910、0.907、0.510 mm/d、0.291 mm/d。因此,在气象资料缺乏情景下,Mak模型可作为东北地区ET0计算的最优模型,并且MEA算法优化能够高效提高模型计算精度,实现了准确性和效率之间更优化的平衡。

Accurate estimation of reference crop evapotranspiration(ET0)is crucial for effective agricultural water resource management.The Northeast region of China,a vital grain-producing area,presents unique challenges due to its relatively high latitude and lower temperatures,which contribute to numerous factors affecting ET0 and high estimation uncertainty.This study uses the Mind Evolutionary Algorithm(MEA)to optimize the ET0 model and compares the accuracy of nine different input factors to identify the best ET0 calculation model for the region.Meteorological data from 20 weather stations spanning from 1961 to 2019 were applied.The Mann-Kendall trend test and inverse distance weighting interpolation were used to analyze the spatiotemporal variations of ET0.The MEA was then applied to optimize the model parameters,and the results were compared with the FAO-56 Penman-Monteith formula.Between 1961 and 2019,the results showed that the annual average ET0 in the Northeast region ranged from 567.81 to 1080.66 mm.The northern part of the region showed an increasing trend in ET0,while the central plains and southern coastal areas exhibited a decreasing trend.Before optimization,the accuracy ranking of different models for ET0 calculation at the 20 stations was as follows:radiation-based models>humidity-based models>temperature-based models.The Mak model demonstrated the highest level of accuracy,with median values of R^(2),NSE,RMSE,and MAE at 0.801,0.786,0.570 mm/d,and 0.331 mm/d,respectively.In addition,after optimization using the MEA,the improvements in R^(2),NSE,RMSE,and MAE for the nine empirical models ranged from 14.43%to 47.15%,14.84%to 50.47%,5.42%to 46.79%,and 7.47%to 39.86%,respectively.The optimized Mak model showed median values of R^(2),NSE,RMSE,and MAE at 0.910,0.907,0.510 mm/d,and 0.291 mm/d,respectively.Therefore,in scenarios with limited meteorological data,the Mak model can be considered the optimal choice for ET0 calculation in the Northeast region.The MEA optimization improves the model accuracy and achieves a better balance between accuracy and efficiency.