我国不同参考作物蒸散经验模型适用性分析及参数修正

为提高不同参考作物蒸散经验模型在全国范围内的精度和适用性，并简化计算方法，依据1961—2020年全国585个气象站点的逐日地面观测资料(统计数据未包含港、澳、台地区，下同)，选用平均相对百分误差、标准差等量化指标，采用最大似然和最短距离法，从月尺度和年尺度上对以Penman-Monteith模型为标准的Hargreaves-Samani(HS)模型、Makkink(MK)法、Priestley-Taylor(PT)模型、Hargreaves(Har)模型、Mc Cloud(MC)模型进行比较分析和系数修正。结果表明:1)修正前除MC模型外月尺度平均相对百分误差平均值均低于20%，PT模型和MC模型年尺度平均相对百分误差平均值均高于25%，精度最低；2)修正后5种模型的月尺度和年尺度平均相对百分误差平均值均低于8%，计算精度明显提升，其中PT模型的修正效果最好，月尺度和年尺度的平均相对百分误差平均值均低于3.7%，与修正前相比均降低了80%以上；3)利用最短距离法进行聚类分析后发现，区域修正后Har模型和MK模型的修正效果最好，年蒸散量相对百分误差平均值分别为8%和5.1%。修正后的不同参考作物蒸散经验模型在全国范围内具有更高的适用性。

Applicability analysis of the empirical models for evapotranspiration of different reference crops in China and their parameters corrections

In order to improve the accuracy and applicability of different reference crops evapotranspiration empirical models in China and simplify the calculation method, the mean relative percentage error, standard deviation and other quantitative indicators were selected, and the maximum likelihood and the shortest distance method were used based on the daily surface observation data of 585 meteorological stations in China from 1961 to 2020(Data do not include those of HongKong, Macao and Taiwan regions. The same below). The Hargreaves-Samani(HS)model, Makkink(MK)method, Priestley-Taylor(PT)model, Hargreaves(Har)model and Mc Cloud(MC)model with Penman-Monteith model as the standard are compared and corrected at the monthly and annual scales. The results show that: 1)The average values of the mean relative percentage error at the monthly scale of all other models are less than 20% except the MC model before correction. The average values of the mean relative percentage error at the annual scale of both PT model and MC model is higher than 25%, indicating these two model display the lowest accuracy; 2)The average values of the mean relative percentage errors at the monthly and annual scales of the five models after correction are less than 8%, proving that the calculation accuracy is significantly improved. Among them, the PT model has the best correction effect, and average values of the mean relative percentage errors of the monthly and annual scales are both less than 3. 7%, which at reduced by more than 80% compared with that before correction; 3)After the cluster analysis by the shortest distance method, it is found that the Har model and MK model have the best correction effect after regional correction, and average values of the mean relative percentage errors of annual erapotras piration value are 8% and 5. 1%, respectively. Therefore, the revised different reference crop evapotranspiration empirical models have higher nationwide applicability prospects.