基于Hargreaves-Samani回归修正的作物参考蒸散发计算适用性研究

为提高Hargreaves-Samani（H-S）模型计算参考蒸散发的精度，利用西北黄河流域与长江中下游平原共128个气象站点1961—2010年的逐日气象资料对H-S模型进行回归修正，以Penman-Monteith（P-M）模型为标准，评价了H-S改进模型H-S_CORR模型的计算精度，并且以第六次国际耦合模式比较计划（CMIP6）气候模式来对H-S_CORR模型进行了未来适应性评价。结果表明：修正后，在验证期内，长江中下游平原4个分区的平均绝对误差（MAE）和均方根误差（RMSE）的平均值分别下降了6.21 mm·月^-1和6.38 mm·月^-1；西北黄河流域4个分区的MAE和RMSE的平均值分别下降了9.26 mm·月^-1和9.23 mm·月^-1，2个研究区域修正后的决定系数（R²）比修正前最少提高1%。在CMIP6气候模式的未来气候情景下R²均达到0.98以上，具有良好的适应性。该研究修正的模型方法可为仅有气温数据的地区提供较高精度的参考蒸散发估算方法，为高频灌溉提供较为准确的数据基础。

Applicability of reference crop evapotranspiration calculation based on Hargreaves-Samani regression correction

The Hargreaves-Samani (H-S) model for calculating reference crop evapotranspiration (ET₀) was regressed and corrected using daily meteorological data from 128 meteorological stations in the Northwest Yellow River Basin and the middle and lower reaches of the Yangtze River from 1961, 1981, to 2010. The calculation accuracy of the H-S upgraded model H-S_CORR model was evaluated using site data from 2011 to 2020 and the Penman-Monteith model was used as the reference. The ACCESS-CM2 model and the future test scenario SSP2-4.5 are also used to assess the H-S_CORR model’s future adaptability as part of the Sixth International Coupling Model Comparison Program (CMIP6) climate simulation experiment. The results show that the mean absolute error (MAE) of the four subdomains in the middle and lower reaches of the Yangtze River decreases from 2.58-24.28 mm to 1.53-7.99 mm per month after correction, and the root mean square error (RMSE) per month decreases from 3.22-24.56 mm to 1.96-9.27 mm during the validation period. Monthly, both MAE and RMSE decrease by 6.21 mm and 6.38 mm, respectively. The MAE of the four subregions of the Northwest Yellow River Basin decreases from 2.51-34.1 mm per month to 1.11-8.94 mm per month, and the RMSE decreases from 3.02-34.58 mm to 1.43-10.46 mm per month. MAE and RMSE monthly averages decrease by 9.26 mm and 9.23 mm, respectively. Most months in the two study areas have a coefficient of determination (R²) greater than 0.9, whereas a few months have an R² greater than 0.8, and the corrected R² value is at least 1% higher than before the correction. The R² values in the CMIP6 climate model’s future climate scenarios are greater than 0.98, showing strong adaptability. Therefore, the H-S_CORR model’s performance in the Northwest Yellow River Basin, as well as the middle and lower reaches of the Yangtze River, is improved; it can now better simulate the seasonal cycle and long-term trend of ET₀, as well as improve the accuracy of reference crop evapotranspiration calculations.