The semiempirical kernel-driven model is commonly used for global surface reflectance characterization because of its simplicity and underlying physical meaning. However, the current kernel-driven reflectance models assume that the terrain is flat and homogeneous, and can induce significant errors in the surface reflectance estimation and subsequent parameter retrievals over rugged terrain. In this study, an improved topography-coupled kernel-driven (TCKD) reflectance model with the correction of diffuse skylight effects was proposed based on the diffused-equivalent slope model (dESM) and RossThick-LiTransit (RTLT) kernel-driven model. The TCKD model’s accuracy and effectiveness were evaluated using surface reflectance simulated by the radiosity approach and the Moderate Resolution Imaging Spectroradiometer (MODIS) data. Against simulated data, the results show that the TCKD model can accurately capture the distortion of the reflectance shape and hemispherical distribution caused by the topographic effects. Compared to MODIS data, the TCKD model has an overall better performance than the RTLT model across different spatial scales and land cover types. When the mean slope is larger than 35° at the 500-m resolution, the TCKD model’s near-infrared (NIR) root-mean-square error (RMSE) and the regression slope of the fitting line are 0.037 and 0.752, respectively, whereas those of the RTLT model are 0.049 and 0.645. Neglecting the diffuse skylight in the TCKD model can also lead to great bias in the reflectance retrievals. When the mean slope is 31°, as the ratio of diffuse skylight varies from 0 to 1, the NIR RMSE of the TCKD model decreases from 0.012 to 0.005, whereas that increases from 0.012 to around 0.02 if the diffuse skylight effects are neglected. These preliminary results demonstrate that the TCKD model is capable of improving the fitting ability of the kernel-driven model over rugged terrain and provides potentials for better retrieving and interpreting land surface parameters such as land surface albedo in mountainous areas.

中文翻译：

---

陆地表面各向异性反射率的改进地形耦合核驱动模型

半经验核驱动模型因其简单性和潜在物理意义而常用于全局表面反射率表征。然而，当前的内核驱动反射率模型假设地形平坦且均匀，并且会在崎岖地形上的表面反射率估计和随后的参数检索中引起重大错误。在本研究中，基于漫反射等效斜率模型 (dESM) 和 RossThick-LiTransit (RTLT) 核驱动模型，提出了一种改进的地形耦合核驱动 (TCKD) 反射率模型，并修正了漫射天窗效应。TCKD 模型的准确性和有效性是使用辐射方法模拟的表面反射率和中分辨率成像光谱仪 (MODIS) 数据进行评估的。针对模拟数据，结果表明，TCKD模型能够准确捕捉到地形效应引起的反射率形状和半球分布的畸变。与 MODIS 数据相比，TCKD 模型在不同空间尺度和土地覆盖类型上的整体性能优于 RTLT 模型。当500米分辨率下平均斜率大于35°时，TCKD模型的近红外（NIR）均方根误差（RMSE）和拟合线的回归斜率分别为0.037和0.752，而RTLT 模型的值为 0.049 和 0.645。忽略 TCKD 模型中的漫射天窗也会导致反射率反演的巨大偏差。当平均坡度为 31°时，随着漫射天窗的比例从 0 变化到 1，TCKD 模型的 NIR RMSE 从 0.012 减小到 0.005，而从 0 增加。如果忽略漫射天光效果，则为 012 到 0.02 左右。这些初步结果表明，TCKD 模型能够提高核驱动模型对崎岖地形的拟合能力，并为更好地检索和解释山区地表反照率等地表参数提供了潜力。