秦淮河流域五种参考作物蒸散量估算方法的比较及改进

基于秦淮河流域内部及周边共7个气象站2000-2013年的逐日气象资料，使用FAO-56 Penman- Monteith、Irmak-Allen、Makkink、Turc、Jensen-Haise和Hargreaves共6种方法估算各站点逐日参考作物蒸散量（ET0）。以FAO-56 Penman-Monteith结果为标准，修正其余5种方法估算公式的原始经验系数，并通过平均绝对误差、平均相对误差、相关系数等精度评价指标和Wilcoxon非参数检验法，分别从年、月尺度对比分析5种方法修正前后的估算结果，旨在获得一种适于秦淮河流域的数据要求低，估算过程简单，精度较高的ET0估算方法。分别以5种方法的ET0日值为自变量，P-M法ET0日值为因变量，建立逐月线性回归方程，寻找经验系数的修正倍数，对5种方法经验系数进行逐月修正。结果表明，使用原始经验系数时，年尺度上，Irmak-Allen、Makkink、Turc法存在较大误差，Hargreaves法相关性较差，均不适于秦淮河流域；月尺度上，Irmak-Allen法在5-8月，Turc在9-11月，Hargreaves法在4月及9-11月适用性较好，其余月份误差较大，Makkink和J-H法分别在1-12月和3-11月存在显著差异，故5种方法均不能代替P-M法在年内12个月使用。使用修正后经验系数，年尺度上Makkink法适用性最好，平均绝对误差和平均相对误差分别为14.9mm·a-1和1.4%，相关系数为0.89，无显著差异，其次为Turc法，I-A法估算结果仍存在显著差异，Hargreaves法相关性仍较差；月尺度上，从估算精度考虑，Turc和Makkink法搭配使用，4-10月推荐使用Turc法，其平均绝对误差为2.1～6.1mm·mon-1，平均相对误差为2.9%～4.3%，无显著差异，月平均相对误差波动较小，稳定性好，1-3月和11-12月推荐使用Makkink法，其平均绝对误差为1.2～2.4mm·mon-1，平均相对误差为3.2%～5.7%，无显著差异，月平均相对误差波动较小，稳定性好，从时间连续性考虑，推荐使用Hargreaves法，其平均绝对误差为1.9～10.4mm·mon-1，平均相对误差为3.9%～9.2%，无显著差异，月平均相对误差波动较小，稳定性好。

Comparison and Modification of Five Crop Reference Evapotranspiration Models for Qinhuai River Basin

The daily reference crop evapotranspiration (ET0) was estimated using the FAO-56 Penman-Monteith and five other methods (Irmak-Allen, Makkink, Turc, Jensen-Haise and Hargreaves) and meteorological data from seven sites located in inside and surrounding areas of the Qinhuai River Basin for the period of 2000-2013.Taken the FAO-56 Penman-Montieth method as a reference, the original empirical coefficients of five other methods were calibrated. The results were analyzed with mean absolute error (MAE), mean relative error (MRE), correlation coefficient (r) and non-parametric Wilcoxon test at an annual and a monthly scale respectively. We aimed to obtain one method which requires less data with high accuracy for the Qinhuai River Basin. With daily results of five methods as independent variable and daily results of FAO-56 Penman-Montieth method as dependent variable, monthly linear regression equations were established. Monthly correction parameters could be found based on these equations. This research indicated that, Irmak-Allen, Makkink, Turc and Hargreaves methods were not applicable at an annual scale when the original empirical coefficients were used. At a monthly scale, when original empirical coefficients were used, large biases and significant differences were found for most of the methods in most months except May to August for Irmak-Allen method, September to November for Turc method, April and September to November for Hargreaves method. After model calibration, the Makkink method performed best followed by Turc method at annual scale. The MAE,MRE and r was 14.9mm·y-1,1.42% and 0.89 respectively. No significant difference existed between the results of Makkink and the FAO-56 Penman-Montieth method. In contrast, there were significant differences between the results of Irmak-Allen and FAO-56 Penman-Montieth method. The Hargreaves method was still not applicable due to a poor correlation. At a monthly scale, considering the accuracy of estimation, Makkink and Turc methods were used by collocation. Turc method was recommended to use from April to October, MAE and MRE was 2.1-6.1mm·mon-1 and 2.9%-4.3%, Makkink method was recommended to use in the period from January to May and November to December, MAE and MRE was 1.2-2.4mm·mon-1 and 3.2%-5.7%. No significant difference existed and coefficient of variation of monthly MRE was small in the two periods for each method. Considering continuity of time, Hargreaves method was recommended to use from January to December, MAE and MRE was 1.9-10.4mm·mon-1 and 4.0%-9.2%.No significant difference existed and coefficient of variation of monthly MRE was small from January to December for Hargreaves method.