遥感技术是大尺度上估算地表碳、水和能量等通量的重要信息来源，全球涡度通量观测数据集已广泛用于遥感通量数据产品的生产和评价，评估遥感像元尺度上通量足迹的空间代表性尤为重要。本文选择我国西南典型山地生态系统-王朗山地遥感四川省野外观测研究站（简称王朗站）区域为例，使用二维参数化足迹模型刻画了通量观测足迹的时空变化特征，同时解析了通量观测足迹在多个遥感像元尺度（30 m、60 m、120 m、250 m、500 m、1000 m、1500 m和2000 m）上的空间代表性。结果表明，在通量足迹的空间变化上，王朗站内不同观测塔通量足迹范围跨度较大（10 ~ 103 m）且对称性较低（通常在40%以下），因此在山地生态系统进行遥感模型及产品验证时需要更加关注通量足迹的空间代表性差异。在通量足迹的时序变化上，王朗站内日尺度上的足迹重叠性差异明显（0% ~ 88%），结合时序变化的足迹特征可进一步提升高时间分辨率下的模型验证和产品精度。王朗研究区内落叶阔叶灌丛站点、落叶阔叶林站点和常绿针叶林站点等三个观测塔的高度为10 m、30 m和75 m，其分别在30 m、60 m和1000 m像元尺度取得对通量足迹的最佳空间代表性。由于山区通量观测的高空间代表性局限于高空间分辨率（观测高度较低时）和中低空间分辨率（观测高度较高时）的遥感像元，认知通量足迹在像元尺度上的空间代表性，结合多尺度遥感观测数据和时空尺度扩展方法，可促进山区生态系统参量估算和通量研究。本文可为站点观测尺度扩展、山地生态系统遥感数据产品生产和地球系统模型验证提供参考。

Objective: With the availability of remote sensing images since the 1970s, it is feasible to obtain the spatial-temporal continuum observations of the land surface at the global scale. In this way, remote sensing is an important information source for large-scale estimation of land surface carbon, water, and energy fluxes. Global eddy covariance flux datasets are widely used to evaluate and produce remote sensing flux products. As the tower-based fluxes can only represent the small areas around the tower, there is usually a mismatch between the tower-based fluxes and multi-scale pixels of remote sensing. Thus, it is crucial to evaluate the spatial representativeness of flux footprint at multi-scale pixels. Method: In this paper, we choose Wanglang Mountain Remote Sensing Field Observation and Research Station of Sichuan Province - a typical mountainous ecosystem of southwest China as the study area. This study used a two-dimensional parametric footprint model (Flux Footprint Prediction, FFP) to characterize the spatiotemporal variations and analyze the spatial representativeness of flux footprint at multi-scale pixels (i.e., 30 m, 60 m, 120 m, 250 m, 500 m, 1000 m, 1500 m and 2000 m). In this work, the land cover types and Normalized Difference Vegetation Index (NDVI) were used to characterize the spatial representativeness of footprint among vegetation types and vegetation density at multi-scale pixels, respectively. At the same time, two site-level simple representativeness indices for land cover type and vegetation density were proposed to evaluate the footprint-to-pixel representativeness across flux towers at Wanglang station. Result: Results showed that the footprint fetch varied across flux towers at Wanglang station (10 ~ 103 m), and the footprint at multiple temporal resolutions had a lower symmetry (usually less than 40%). For the temporal variations of footprint, there were more obvious changes for the overlap of footprint at daily scale (0% ~ 88%), and the variations were smaller at monthly scale (usually larger than 83%). As for the three flux towers around Wanglang station, results showed that the station of deciduous broadleaf shrub (with observed height at 10 m), deciduous broadleaf forest (with observed height at 30 m), and evergreen needleleaf forest (with observed height at 75 m) had a best spatial representativeness at the pixel scale of 30 m, 60 m, and 1000 m, respectively. Moreover, compared to vegetation density, the discrepancies of spatial representativeness were more obvious for vegetation cover. Conclusion: It should be more attention to the spatial representativeness differences of footprint while validating remote sensing models and producing flux datasets around mountainous ecosystems. Moreover, it was necessarily to combine the corresponding footprints with tower-based observations to characterize the temporal variations of fluxes when modeling and producing flux products at high temporal resolution (e.g, daily scale). Due to the high spatial representativeness of footprint was limited to the pixels at high (a lower tower) and medium-low (a higher tower) spatial resolution, the estimation of ecosystem parameters and fluxes research over mountainous area could be promoted by cognizing the spatial representativeness of footprint at pixel scales and combing the multi-scale remote sensing observations with the spatial and temporal scaling method.