Quantification of actual crop evapotranspiration (ETa) over large areas is a critical issue to manage water resources, particularly in semi-arid regions. In this study, four models driven by high resolution remote sensing data were intercompared and evaluated over an heterogeneous and complex traditional irrigated area located in the piedmont of the High Atlas mountain, Morocco, during the 2017 and 2018 seasons: (1) SAtellite Monitoring of IRrigation (SAMIR) which is a software-based on the FAO-56 dual crop coefficient water balance model fed with Sentinel-2 high-resolution Normalized Difference Vegetation Index (NDVI) to derive the basal crop coefficient ($K_{\mathit{cb}}$); (2) Soil Plant Atmosphere and Remote Sensing Evapotranspiration (SPARSE) which is a surface energy balance model fed with land surface temperature (LST) derived from thermal data provided from Landsat 7 and 8; (3) a modified version of the Shuttleworth–Wallace (SW) model which uses the LST to compute surface resistances and (4) METRIC-GEE which is a version of METRIC model (“Mapping Evapotranspiration at high Resolution with Internalized Calibration”) that operates on the Google Earth Engine platform, also driven by LST. Actual evapotranspiration (ETa) measurements from two Eddy-Covariance (EC) systems and a Large Aperture Scintillometer (LAS) were used to evaluate the four models. One EC was used to calibrate SAMIR and SPARSE (EC1) which were validated using the second one (EC2), providing a Root Mean Square Error (RMSE) and a determination coefficient (R) of 0.53 mm/day (R=0.82) and 0.66 mm/day (R=0.74), respectively. SW and METRIC-GEE simulations were obtained respectively from a previous study and Google Earth Engine (GEE), therefore no calibration was performed in this study. The four models predict well the seasonal course of ETa during two successive growing seasons (2017 and 2018). However, their performances were contrasted and varied depending on the seasons, the water stress conditions and the vegetation development. By comparing the statistical results between the simulation and the measurements of ETa it has been shown that SAMIR and METRIC-GEE are the less scattered and the better in agreement with the LAS measurements (RMSE equal to 0.73 and 0.68 mm/day and R equal to 0.74 and 0.82, respectively). On the other hand, SPARSE is less scattered (RMSE = 0.90 mm/day, R = 0.54) than SW which is slightly better correlated (RMSE = 0.98 mm/day, R = 0.60) with the observations. This study contributes to explore the complementarities between these approaches in order to improve the evapotranspiration mapping monitored with high-resolution remote sensing data

大面积实际作物蒸散量 (ETa) 的量化是管理水资源的一个关键问题，特别是在半干旱地区。在这项研究中，在 2017 年和 2018 年季节，在摩洛哥高阿特拉斯山山前的一个异质和复杂的传统灌溉区，由高分辨率遥感数据驱动的四个模型进行了相互比较和评估：（1）卫星监测灌溉 (SAMIR) 是基于 FAO-56 双作物系数水平衡模型的软件，使用 Sentinel-2 高分辨率归一化差异植被指数 (NDVI) 来推导基础作物系数 (${ķ}_{\mathit{CB}}$); (2) 土壤植物大气和遥感蒸散量 (SPARSE)，它是一种地表能量平衡模型，由Landsat 7 和 8 提供的热数据得出的地表温度 (LST) 提供；(3) Shuttleworth-Wallace (SW) 模型的修改版本，它使用 LST 计算表面电阻和 (4) METRIC-GE​​E，它是 METRIC 模型的一个版本（“使用内部校准绘制高分辨率的蒸散量”）在同样由 LST 驱动的 Google Earth Engine 平台上运行。来自两个涡协方差 (EC) 系统和一个大孔径闪烁仪 (LAS) 的实际蒸散 (ETa) 测量值用于评估四个模型。一个 EC 用于校准 SAMIR 和 SPARSE (EC1)，后者使用第二个 (EC2) 进行了验证，提供了一个均方根误差 (RMSE) 和决定系数 (R) 分别为 0.53 毫米/天 (R=0.82) 和 0.66 毫米/天 (R=0.74)。SW 和 METRIC-GE​​E 模拟分别从先前的研究和谷歌地球引擎 (GEE) 获得，因此在本研究中没有进行校准。这四个模型很好地预测了连续两个生长季节（2017 年和 2018 年）中 ETa 的季节性过程。然而，它们的表现因季节、水分胁迫条件和植被发育而不同。通过比较模拟和 ETa 测量值之间的统计结果，可以看出 SAMIR 和 METRIC-GE​​E 的散射越少，并且与 LAS 测量值的一致性越好（RMSE 等于 0.73 和 0.68 毫米/天，R 等于0.74 和 0.82，分别）。另一方面，SPARSE 比 SW 更分散（RMSE = 0.90 毫米/天，R = 0.54），而 SW 与观测结果的相关性（RMSE = 0.98 毫米/天，R = 0.60）稍好。本研究有助于探索这些方法之间的互补性，以改进利用高分辨率遥感数据监测的蒸散图