高精度的农田蒸散发（ET）对于田间尺度精准的水分平衡量化和水分亏缺研究具有重要意义，对农业精准灌溉和农田水分利用效率提高具有实用价值。本研究利用全球共计28个农田站点的涡动相关系统（EC）通量观测数据，对500 m空间分辨率8天时间分辨率的MOD16和PML-V2两种遥感蒸散发产品，进行了对比分析与精度评估。评估结果表明，PML-V2 ET产品与EC观测ET相比，均方根误差（RMSE）变化范围为3.3~22.4 mm/8d，平均偏差（bias）变化范围为-15.98~13.27 mm/8d；MOD16 ET产品与EC观测ET相比，RMSE变化范围为3.81~21.47 mm/8d，bias变化范围为-16.42~15.05 mm/8d。整体而言，两种产品精度相当，MOD16产品低估8天ET（bias：-2.31 mm/8d，R2：0.452，RMSE：8.82 mm/8d），PML-V2产品略高估8天ET（bias：0.51 mm/8d，R2：0.455，RMSE：8.81 mm/8d）；PML-V2产品在18个站点（64%）上表现更优，但是部分站点上MOD16产品对时序变化细节（如到达波峰的季节、年中的下降/上升趋势）的刻画优于PML-V2产品；PML-V2产品未能捕捉到冬小麦和夏玉米轮种而引起的ET年中先下降后上升的变化趋势，而MOD16产品虽然成功捕捉到ET时序曲线上冬小麦和夏玉米两种作物各自生长期内的波峰，但仍较大程度上低估了冬小麦ET（如栾城站点和禹城站点）；另外，MOD16产品和PML-V2产品均严重低估了水稻ET（如US-Twt站点）。本研究可为农田蒸散发算法发展及其精度验证提供参考。

Crop evapotranspiration (ET) with high precision is of great significance for the accurate quantification of water balance and the study of water deficit in the field-scale, and it has practical value for the precision irrigation of farmland and the improvement of agricultural water use efficiency. It is essential to validate ET before a remotely sensed ET product being used. This study evaluated crop ETs from two remotely sensed products (MOD16 and PML-V2) with 500 m spatial resolution and 8-day temporal resolution by using eddy-covariance (EC) flux observations from 28 flux tower sites cover with crop globally. The results showed that, compared with the observed ET, the root mean square error (RMSE) and bias of the PML-V2 ET products varied from 3.3 to 22.4 mm/8d and from 15.98 to 13.27 mm/8d, respectively, while the RMSE and bias of the MOD16 ET product varied from 3.81 to 21.47 mm/8d and from -16.42 to 15.05 mm/8d, respectively. On the whole, the overall accuracies of these two products were similar, the MOD16 product underestimated the 8-day ET with a bias of -2.31 mm/8d, a R2 of 0.452 and a RMSE of 8.82 mm/8d, while the PML-V2 product slightly overestimated the 8-day ET with a bias of 0.51 mm/8d, a R2 of 0.455 and a RMSE of 8.81 mm/8d. The PML-V2 product performed better at 18 tower sites (almost 64%), but the MOD16 product performed better than the PML-V2 products at some sites in the depiction of details on time-series change (such as the season of reaching the peak during the year, the decreasing and increasing trend in the middle of year). The results showed that the PML-V2 product failed to capture the gradual decrease then increase ET trend in the middle of year which caused by the rotation of winter wheat and summer maize, while the MOD16 product successfully captured the hitting of the two peaks in ET time-series during the two growth seasons of winter wheat and summer maize (such as the Luancheng and Yucheng sites). However, the MOD16 product still underestimated the 8-day ET of winter wheat to a certain degree. Moreover, the results showed that both the MOD16 product and the PML-V2 product seriously underestimated the 8-day ET of paddy with a RMSE of 21.47~22.4 mm/8d and a bias of -16.42~-15.98 mm/8d (such as the US-Twt site). This study could provide reference for the development and validation of ET models for cropland. In the future, more detailed evaluation of land surface heterogeneity needs to be carried out and more products should be taken into consideration. Further detailed evaluation of ET in different crop types is also required.