考虑“蒸发悖论”的洱海灌区逐日参考作物蒸散发预测

【目的】为估算参考作物蒸散发(ET₀)和灌溉实时预报调度、区域农业干旱评估提供依据。【方法】以滇中高原上洱海湖滨灌区的大理气象站为例,探究“蒸发悖论”现象出现的时期,采用气象因子线性回归模型、蒸发皿折算系数K_p模型、气象因子+蒸发皿蒸发(E_pan)多元回归模型、Normal Copula模型等4种方法计算逐日ET₀进行预测对比,并与Penman-Monteith公式计算所得的ET₀进行对比。【结果】①1954—2018年大理站20 cm蒸发皿蒸发量呈下降趋势,ET₀和气温呈上升趋势,但ET₀的上升趋势更平缓;虽然在长时间序列上ET₀和蒸发皿蒸发量有相反的变化趋势,但在年代际存在显著的差异性,1960年和2000全年以及四季均出现“蒸发悖论”,1970年则是全年以及夏、秋、冬三季出现“蒸发悖论”,1990年仅夏季出现“蒸发悖论”,2010年秋季出现“蒸发悖论”。②在未出现“蒸发悖论”时期,加入E_pan后的气象因子多元回归模型法(ET_0,Epan+Metr)所得逐日ET₀预测结果与标准值的误差最小,其次为单纯的气象因子多元线性回归模型法(ET₀,Metr),最差为K_p模型法(ET₀,K_p);加入E_pan后的气象因子多元回归模型(ET₀,ET_0,Epan+Metr)逐日ET₀预测的相对误差(ERR)小于15%、20%、25%的样本数达到了79.18%~90.16%、89.32%~97.23%、94.79%~98.36%。③出现“蒸发悖论”时,蒸发皿蒸发与ET₀的变化趋势相反,只能采用Copula联合分布函数模型预测,构建T-T_max二维Normal Copula模型的精度更高,ERR小于15%、20%、25%的样本数为73.70%~86.56%,82.51%~92.95%,89.89%~98.52%。【结论】通过M-K检验判别是否处于“蒸发悖论”期,以决策选用加入E_pan后的气象因子多元回归模型,还是T-T_max二维Normal Copula模型,二者均可显著提高逐日ET₀预测模拟的精度。

Calculating Daily Reference Evapotranspiration in Erhai Irrigated District with the Evaporation Paradox in Consideration

【Background】Evapotranspiration(ET)is an important process in the hydrological cycle and has been increasing,as shown by both field measurements and satellite remote sensing,at 0.63 mm/year over the past 30 years due to the increased leaf area index which resulted in a 0.72 mm increase in transpiration.In contrast,evaporation from soil has been decreasing at 0.32 mm/year,a phenomenon known as evaporation paradox.【Objective】The purpose of this paper is to compare the daily reference evapotranspiration(ET₀)calculated from the Penman-Monteith formula with the evaporation paradox in consideration with data measured from a 20cm pan in the Erhai lake irrigation district after modifying the measurements by a pan conversion coefficient.【Method】Meteorological data measured from Dali weather station at upper stretch of the Erhai Lake irrigation district in central Yunnan plateau was used to analyze the occurrence of the evaporation paradox.We then predicted ET₀ using four models:linear regressions using the meteorological factors only,multiplying measurements from the 20cm pan by a pan conversion coefficient,multiple regression model using the meteorological factors in combination with the pan measurements,and the normal copula model.The ET₀ predicted by these models was compared with that calculated from the Penman-Monteith formula.【Result】①Data measured from the 20 cm evaporation pan showed that the evaporation had been in decline from 1954 to 2018,while ET₀ and temperature had been in increase although the former did not increase as fast as the latter.ET₀ and evaporation measured from the pan trend with time in opposite ways,but the difference between their annual means was not significant.Both seasonal and annual average evaporations showed paradox in the 1960s and the 2000s,while in the 1970s only did the annual average evaporation and seasonal average evaporation in summer,autumn and winter showed paradox.In the 1990s and 2010s,the paradox occurred only seasonally in summer and autumn,respectively.②Without the evaporation paradox,the daily ET₀ predicted by the multiple regression model using both the meteorological factors and the pan measurements was most accurate compared to the standard ET₀,followed by the linear regression model using the meteorological factors only;the worst was that calculated by multiplying the pan measurements by a pan conversion coefficient.Among all data we calculated by the model considering both the pan measurements and the meteorological factors,the fractions that have relative errors less than 15%,20%and 25%were 79.18%~90.16%,89.32%~97.23%,94.79%~98.36%,respectively.③In the presence of evaporation paradox,the evaporation measured from the pan and the ET₀ trended differently with time,and the ET₀ can only be calculated using the copula joint distribution function model.The constructed T-T_max model was most accurate,and among all data calculated using it,the fractions that have relative errors less than 15%,20%and 25%were 73.70%~86.56%,82.51%~92.95%,89.89%~98.52%,respectively.【Conclusion】We used the M-K test to determine the presence of evaporation paradox.We proved that in the absence of the paradox,the multiple regression model considering both meteorological factors and the pan measurements worked better,while in the presence of the paradox,the T-T_max two-dimensional normal copula model was most accurate for predicting ET₀.