Cross-calibration is a wildly used approach to radiometrically calibrate satellite sensors to meet uncertainty requirements. However, its accuracy is greatly influenced by bidirectional reflectance distribution function (BRDF) effects. To understand such influences, we analyzed the long-term BRDF characteristics of two commonly used cross-calibration targets (‘bright’ desert and ‘dark’ forest) using long-term (2002 – 2016) Moderate Resolution Imaging Spectroradiometer (MODIS) BRDF products. BRDF characteristics derived by MODIS observations of both land cover types had 365-day cycles. Furthermore, both solar and sensor zenith angles showed more influence on BRDF than relative azimuth angles, and BRDF variations were approximately 10 times sensitive when solar or sensor zenith angles surpassed the boundary of 75°, which is therefore not recommended for selecting image pairs between target and reference satellite sensors for cross-calibration. Due to cloud cover, three prevailing interpolation methods (nearest, linear, and spline interpolation) were also assessed in their capacity to fill temporally missing BRDF products. Linear interpolation was preferred when valid MODIS scenes were more than two in five days, while nearest interpolation outperformed the other two methods for no more than two valid MODIS scenes in five days. In addition, BRDF look up tables (LUTs) were established for radiometrically cross-calibrating satellite sensors when BRDF products were unavailable. The LUTs were further validated with Landsat-8/Operational Land Imager (OLI) top of atmosphere (TOA) radiance simulated by the Second Simulation of the Satellite Signal in the Solar Spectrum model (6S) with BRDF effects corrected using the BRDF LUTs and MODIS BRDF products for typical geometries at two sites. High consistency was achieved with mean biases less than 0.3%, which suggests that BRDF LUTs could be used as alternatives when BRDF products are unavailable. These results provide guidelines to improve the accuracy of cross-calibration including image-matchup selecting criteria, efficient BRDF effect removal, and selection of potential calibration sites.

交叉校准是用于辐射校准卫星传感器以满足不确定性要求的一种广泛使用的方法。但是，其精度受双向反射分布函数（BRDF）的影响很大。为了了解此类影响，我们使用长期（2002年至2016年）中分辨率成像光谱仪（MODIS）BRDF产品分析了两个常用的交叉校准目标（“亮”沙漠和“黑暗”森林）的长期BRDF特性。 。由MODIS观测到的两种土地覆被类型得出的BRDF特征具有365天的周期。此外，太阳和传感器的天顶角对BRDF的影响都比相对方位角更大，当太阳或传感器的天顶角超过75°的边界时，BRDF的变化大约是灵敏度的10倍，因此，建议不要在选择目标和参考卫星传感器之间的图像对进行交叉校准时使用此功能。由于云层覆盖，还评估了三种主要的插值方法（最近，线性和样条插值）在填充暂时丢失的BRDF产品上的能力。当有效的MODIS场景在五天内超过两个时，线性插值是首选，而在五天内不超过两个有效的MODIS场景中，最接近的插值优于其他两种方法。此外，当BRDF产品不可用时，建立了BRDF查找表（LUT），以进行辐射交叉校准的卫星传感器。通过在太阳光谱模型（6S）中对卫星信号的第二次模拟对Landsat-8 / Operational Land Imager（OLI）大气顶部（TOA）辐射进行了进一步的LUT验证，并使用BRDF LUT和MODIS校正了BRDF效应适用于两个地点的典型几何形状的BRDF产品。在平均偏差小于0.3％的情况下实现了高度一致性，这表明在无法提供BRDF产品时，可以将BRDF LUT用作替代方案。这些结果为提高交叉校准的准确性提供了指导，包括图像匹配选择标准，有效的BRDF效果去除以及潜在的校准位点的选择。这表明，在没有BRDF产品的情况下，可以将BRDF LUT用作替代产品。这些结果为提高交叉校准的准确性提供了指导，包括图像匹配选择标准，有效的BRDF效果去除以及潜在的校准位点的选择。这表明，在没有BRDF产品的情况下，可以将BRDF LUT用作替代产品。这些结果为提高交叉校准的准确性提供了指导，包括图像匹配选择标准，有效的BRDF效果去除以及潜在的校准位点的选择。