基于METRIC模型的江汉平原蒸散量估算适应性分析

蒸散发是整个水文循环中的关键环节,现有传统测量方法虽具有较高的点位精度,但空间代表性不足,无法满足大空间尺度的遥感估算。以半湿润区为主的江汉平原为研究区域,以遥感蒸散模型为核心手段,首次引入METRIC模型,探讨METRIC模型在江汉平原蒸散量估算中的适用性,并应用METRIC模型和SEBAL模型对研究区域进行遥感蒸散量反演,同时利用世界粮农组织(FAO)提供的P-M模型参考蒸散量计算公式,计算了气象站点当日参考蒸散量并进行估算误差对比。结果表明:(1)4日中蒸散量较小的3日METRIC模型估算误差较SEBAL模型更小,平均估算误差降低约9%,仅蒸散量较大的1日SEBAL模型较METRIC模型具有更高的估算精度,这表明蒸散量较小的时间段内,METRIC模型在江汉平原表现出良好的适用性,估算误差较SEBAL模型更小,具有更好的应用价值,而蒸散量较大的时间段内,SEBAL模型能够提供较METRIC模型更高的估算精度,具有更好的应用价值;根据季节交替使用两模型能够有效提高区域遥感蒸散量的估算精度;(2)不同蒸散量时两模型出现估算精度差异的最大影响因素为冷热像元即干湿边界的选取原则,SEBAL模型干湿边界的选取原则以水体作为湿边界,适合土壤蒸发量及植被散发量较小的干旱地区,而METRIC模型干湿边界的选取原则以湿润、温度较低的植被覆盖区域作为湿边界,增加了对植被散发的权重考虑,避免了高植被覆盖区域水分胁迫带来的精度影响,并以DEM数据对区域高程、地形坡度加以修正,可降低高程差所带来的"冷却效应",这些改进使该模型在蒸散量较小的时间段内取得了更高的估算精度;但蒸散量较大时,水体蒸散在总蒸散量中的权重大幅增加,降低了植被散发等对最终估算结果的影响程度,此时SEBAL模型以温度较低水体作为湿边界的选取原则使水体蒸散发在总蒸散量估算中占据更大的权重,相较METRIC模型在蒸散量较大时能够提供更高的估算精度。

Adaptability Analysis of Estimation of Evapotranspiration in Jianghan Plain Based on METRIC Model

Evapotranspiration is a key link in the entire hydrological cycle.Although the existing traditional measurement methods have high point position accuracy,they are insufficiently representative of space and cannot meet large-scale remote sensing estimates.In response to this situation,this paper takes the semi-humid area as the main research area in Jianghan Plain as the study area,and uses the remote sensing eva-potranspiration model as the core method to introduce the METRIC model for the first time to discuss the applicability of the METRIC model in Jianghan Plain.At the same time,the paper applies the METRIC model and the SEBAL model to the inversion of regional remote sensing evapotranspiration,and uses the P-M model reference evapotranspiration calculation formula provided by the World Food and Agriculture Organization to calculate the reference evapotranspiration of the meteorological station on the day for error comparison.The results show that the estimation error of the three-day METRIC model with small evapotranspiration in four days is smaller than that of the SEBAL model,with the average error,which is reduced by about 9%.Only the one-day SEBAL model with larger evapotranspiration has higher accuracy than that of the METRIC model.The results indicate that the METRIC model performs well in Jianghan Plain when the evapotranspiration is smaller,which shows good applicability,with smaller errors than the SEBAL model does,and has higher application value;when the evapotranspiration is larger,the SEBAL model can provide higher estimation accuracy than the METRIC model can and has higher application value;the use of the two models based on seasonal alternation can effectively improve the estimation accuracy of regional remotely sensed evapotranspiration.The results also show that the biggest influencing factor for the diffe-rence in accuracy between the two models for different evapotranspiration is the selection principle of cold and hot pixels,namely,the wet and dry boundary.The dry and wet boundary selection principle of the SEBAL model uses water as the wet boundary,which is suitable for drought where soil evaporation and plant emissions are small.The wet and dry boundary selection principle of the METRIC model uses humid and low temperature vegetation coverage as the wet boundary,which increases the weight of consideration of vegetation emission,avoids the accuracy impact caused by water stress in high vegetation coverage areas,and reduces the"cooling effect"caused by elevation difference by modifying the elevation and terrain slope with DEM data.The above improvements enable it to achieve better accuracy when the evapotranspiration is small;when the evapotranspiration is large,the weight of the water evapotranspiration in the total evapotranspiration is greatly increased.The influence of vegetation emission on the final estimation results is reduced.At this time,the SEBAL model can provide better estimation accuracy when the evapotranspiration is large,which uses the lower temperature water body as the wet boundary to make the water body more weight in the calculation,compared with the METRIC model.