为探究流域长时序高空间分辨率蒸散发量计算对区域水资源开发利用、水利工程规划设计及农业可持续发 展的重要意义，以河北省邢台市柳林流域为研究对象，基于 Penman-Monteith 模型和蒸渗仪实测蒸散发数据计算 不同时期的流域作物系数（Kc），并建立 Kc与归一化植被指数（normalized?difference?vegetation?index，NDVI）的关系， 利用 250?m 分辨率 NDVI 产品将蒸渗仪测算的蒸散发量升尺度到柳林流域，计算流域各网格 2000—2021 年的蒸 散发量，分析蒸散发量的时空变化规律。结果表明：柳林流域多年平均潜在蒸散发量为 1?135.6mm，呈下降趋势； 多年平均蒸散发量为 591.4mm，呈上升趋势。蒸散发量在空间上西北高东南低，四季蒸散发量空间分布特征与多 年平均蒸散发量一致，且季节上分配不均。基于 NDVI 估算的蒸散发量与水量平衡法计算的蒸散发量 2000—2020 年多年平均相对误差为 7.9%，说明利用 Kc与 NDVI 关系可以较精确地对蒸散发量进行空间尺度提升。

In the catchment scale, evapotranspiration refers to the sum of evaporation from water surface and soil, and transpiration of vegetation in a specific region. Evapotranspiration can reflect heat exchange and water exchange between land and atmosphere, and is an important element in hydrological cycle. Therefore, the study of evapotranspiration variation is of great significance to investigate the law of water cycle, the planning and design of hydraulic engineering, and the efficient utilization of water resources. Due to the short time series of evapotranspiration observation, the small number of observation stations, and the heterogeneity of underlying surface at catchment scale, the spatial-temporal distribution of evapotranspiration can not be obtained using the measured data. Scaling up the observed short-term evapotranspiration data at a small scale is an efficient way to get long-term time series at regional scale. To this end, Liulin watershed was selected as the study area in which a large lysimeter was installed in 2021. Penman-Monteith model is a method for calculating reference crop evapotranspiration (ET0) determined by FAO based on the energy balance and water-air diffusion theory, and was employed in the Liulin watershed. Combined with the measured evapotranspiration (ETc) by the lysimeter from June 2020 to May 2021, the crop coefficient (Kc) was calculated by the equation ETc= Kc ×ET0. Studies have shown that crop coefficient (Kc) has a good linear correlation with normalized vegetation index (NDVI), and this relationship has also been used to study evapotranspiration in ecosystems with different crop types and uneven underlying surface. The linear relationship between Kc obtained by the lysimeter and NDVI was established based on the data from June 2020 to May 2021. Then the Kc with the grid size of 250 m was scaled up to Liulin watershed scale from 2000 to 2021 according to the spatial NDVI distribution from 2000 to 2021. Finally, the evapotranspiration with the same spatial and temporal resolution as NDVI in Liulin watershed was obtained. The annual evapotranspiration was also calculated by water balance equation based on long-term observed precipitation and runoff data, in order to verify the feasibility of the proposed evapotranspiration scaling up method. The results showed that: (1) The average annual potential evapotranspiration of Liulin watershed from 2000 to 2021 was 1135.6 mm with a decreasing trend. The average annual actual evapotranspiration was 591.4 mm with an increasing trend. Both the monthly mean potential evapotranspiration and actual evapotranspiration were unimodal, and the peaks occurred in June and July, respectively. (2) Spatially, the annual average evapotranspiration was high in the northwest and low in the southeast, and the spatial distribution characteristics of evapotranspiration in the four seasons were similar with the annual average. The distribution of evapotranspiration was extremely uneven in the four seasons. In summer, evapotranspiration reached 285.9 m, accounting for 48.3% of the whole year, while in winter, evapotranspiration was only 24.2 mm, accounting for 4.1% of the whole year. (3) The annual average crop coefficient of the basin was 0.52, and the spatial variation ranged from 0.34 to 0.73. The crop coefficient in summer was the largest, reaching 0.70, and that in winter was the smallest (0.20). The actual evapotranspiration calculated by the relation between Kc and NDVI was 43.5 mm higher than that calculated by the water balance method, and the relative error from 2000 to 2020 was 7.9% which was acceptable. This error might come from the errors in the fitting relation of crop coefficient under different meteorological conditions. With the increase of actual monitoring data, the relationship between Kc and NDVI can be refined according to different meteorological conditions. The accuracy of the method to calculate the actual evapotranspiration can be further improved. In general, the scaling up method was feasible, and can be applied to other similar regions, and can be tested in different climate zones.