基于卫星遥感的大尺度植被物候监测对于农业生产管理、气候变化响应等领域具有重要意义。卫星遥感提取的物候参数真实性检验常以近地面数字相机观测为数据源。以往的验证研究中,大多关注观测尺度的差异造成的影响,却忽略了地面相机与卫星观测天顶角的不一致性。本文获取了垂直(PhotoNet)与倾斜(PhenoCam)物候相机在相近纬度、同种植被类型的观测站点数据,分别与Sentinel-2卫星提取的物候参数进行对比,系统性地评估两种地面相机观测角度对卫星物候期验证结果的影响。结果表明,相机的观测角度是卫星遥感物候验证研究中的不确定性原因之一。在多数情况下,相机垂直观测时与卫星遥感提取的物候期表现出更好的一致性,平均相差9天,而倾斜相机与卫星的物候期平均相差可以达到19天。然而,该偏差还与植被类型和植被生长阶段有关,在玉米的凋落期至休眠期,垂直观测的偏差要高于倾斜观测。植被冠层的方向反射特性和相机视场范围内的目标组分差异是引起这种现象的主要原因。本研究证实了地面相机观测角度是卫星物候真实性检验中不可忽视的影响因素。因此,相机野外布设时应充分考虑角度效应带来的验证误差,从而为卫星遥感物候监测提供更可靠的验证数据。

In the fields of agricultural production management and climate change research, it has been well recognized that monitoring large-scale plant phenology with satellite-based remote sensing is of great significance to reveal the process of the interaction of biology and nature environment. In the activities of validation on remotely sensed phenology information, near-surface digital cameras are often employed as main data sources. However, more efforts were focused on the scale difference between ground and remotely sensed data, and rarely on the difference of sensors viewing zenith, i.e., the satellites mainly adopted the near- nadir observation while cameras were mostly inclined arrangement. In order to systematically assess the effects of camera observation angles on the satellite phenological verification results, vertical (PhotoNet) and inclined (PhenoCam) camera observations were acquired at the similar latitude for the same vegetation type, then we compared them with phenological parameters extracted from Sentinel-2 data. For sixteen locations we compared greenness chromatic coordinate (GCC) series derived from digital cameras and Sentinel-2. A double hyperbolic tangent model was fitted for each series. The threshold method was applied to the annual complete modelled data, and the curvature extremum method was used for incomplete data to estimate onset of greenup, maturity of the green canopy, peak of season, end of greenness and dormancy of the green vegetation (OG/MG/PS90/EG/DG). The results showed that the viewing zenith of cameras is one of the uncertain sources to validate phenology information from satellite imagery. In most cases, the vertically observed camera showed better agreement with the phenological parameters extracted by satellite-based method, with an average bias of 9 days, while a larger bias of 19 days was observed for inclined camera observation. Therefore, the two camera observation methods result in the verification deviations of up to 10 days on average. However, the deviations might be vegetation type and growth stage dependently. It was found that the bias of vertical observation was significantly higher than that of inclined observation during the end to dormant period for maize. The different result of the vertical and inclined camera on the similar vegetation can be partly explained from the directional reflection characteristics of vegetation canopy and the difference of target components (e.g., different fraction of soil and vegetation, photosynthetic and non-photosynthetic components of vegetation) within the camera field of view. Our results demonstrate that the viewing zenith angle of the near-surface cameras is an important factor in validation of satellite phenological parameters. In addition, the analysis found that the uncertainty of verification results caused by phenological period extraction method, illumination and satellite observation geometry is limited, which is not the main factor affecting the verification of satellite phenology parameters. As a result, it is suggested that to provide more reliable verification data for satellite remote sensing monitoring, the verification error introduced by angle effect should be fully considered while near surface cameras are deployed in the field.