Airborne laser scanning (ALS) is widely used in studies of Earth surface change and has potential to inform targeted landscape remediation over large areas. Leveraging this capability requires geomorphic change detection methods that exploit the full 3D information contained in ALS point clouds but remains challenging over large areas (i.e., > 10 km2). We developed a methodology for geomorphic change detection over large areas using multitemporal ALS in a multiscale model-to-model cloud comparison (M3C2) framework adapted for volumetric estimation. Time series Sentinel-2 optical observations were used to isolate persistently-bare areas as candidate sites to co-register the ALS point clouds. Geomorphic stability of those sites was determined from coherence change detection using time series Sentinel-1 InSAR, thereby ensuring only geomorphically-stable sites were used for co-registration. Results showed the Sentinel-based co-registration produced a closer vertical alignment (0.00 ± 0.09 m) between ALS point clouds over stable parts of the landscape, while co-registration using an iterative closest-point algorithm contained bias (0.07 ± 0.10 m). The methodology was used to estimate annual sediment yield for a semi-arid catchment in northeastern Australia and results were compared with long-term field-based stream sediment monitoring. The ALS-based geomorphic change detection estimated 2.58 ± 0.54 t·ha−1·a−1 sediment yield and stream sediment monitoring estimated 1.40 t·ha−1·a−1. These similar estimates indicate multitemporal ALS can produce realistic whole-of-catchment sediment yield estimates in ungauged catchments (i.e., with no stream sediment monitoring) and improves the spatial detail of those estimates. Accurately detecting geomorphic change from multitemporal ALS also required a strategy to manage vegetation-related error due to misclassification of ALS point clouds. Combined identification of fine-scale erosion processes and reliable estimation of catchment-scale erosion rates indicates the proposed methodology provides a valuable tool for planning landscape remediation over large areas.

中文翻译：

---

使用 Sentinel-1 InSAR 和 Sentinel-2 光学观测的多时相机载激光扫描优化大面积亚米分辨率 3D 地貌变化检测


机载激光扫描 （ALS） 广泛用于地球表面变化的研究，并有可能为大面积的有针对性的景观修复提供信息。利用此功能需要地貌变化检测方法，这些方法利用 ALS 点云中包含的完整 3D 信息，但在大面积区域（即 > 10 km2）上仍然具有挑战性。我们在适用于体积估计的多尺度模型到模型云比较 （M3C2） 框架中使用多时态 ALS 开发了一种大面积地貌变化检测的方法。时间序列 Sentinel-2 光学观测用于隔离持续裸露的区域作为候选站点，以共同配准 ALS 点云。使用时间序列 Sentinel-1 InSAR 通过相干变化检测确定这些站点的地貌稳定性，从而确保仅使用地貌稳定的站点进行共同配准。结果表明，基于 Sentinel 的共同配准在景观的稳定部分上的 ALS 点云之间产生了更紧密的垂直对齐 （0.00 ± 0.09 m），而使用迭代最近点算法的共同配准包含偏差 （0.07 ± 0.10 m）。该方法用于估计澳大利亚东北部半干旱集水区的年沉积物产量，并将结果与基于长期现场的河流沉积物监测进行比较。基于 ALS 的地貌变化检测估计了 2.58 ± 0.54 t·ha−1·a−1 的沉积物产量，河流沉积物监测估计了 1.40 t·ha−1·a−1。这些类似的估计表明，多时相 ALS 可以在未测量的流域（即没有河流沉积物监测）中产生现实的整个流域沉积物产量估计，并改善这些估计的空间细节。 准确检测多时态 ALS 的地貌变化还需要一种策略来管理由于 ALS 点云错误分类而导致的植被相关误差。对精细侵蚀过程的识别和对流域规模侵蚀速率的可靠估计相结合，表明所提出的方法为规划大面积景观修复提供了有价值的工具。