ABSTRACT Mapping the streambed morphology is crucial for understanding fluvial forms and processes, for advancing both our knowledge and best management practice of riverine systems. It is often done by wading streams but recently topobathymetric Light Detection and Ranging (LiDAR) and imagery captured from airborne platforms are becoming promising and effective surveying methodologies. The recent use of remotely piloted aerial systems (RPAS) combined with structure-from-motion (SfM) photogrammetric processing provides a novel, high-resolution approach to modelling fluvial morphology. Nevertheless, a complicating factor of such data acquisition in fluvial settings is linked to water bodies, whose presence introduces errors and distortions because of light reflection and refraction at the air–water interface. Although proof-of-concept research has shown it is possible to reduce the effects of refraction in certain settings (e.g., clear waters, unbroken surfaces), corresponding validation methods remain limited to point-based assessments of error, these being typically limited to accessible parts of the channel. Here, we provide the first high-resolution, spatially continuous validation of a stream reach bathymetry surveyed with SfM and corrected with an advanced refraction method. This method required only the camera co-ordinates and is available as open source, coded in C++. We used RPAS imagery from a regulated reach of the Palancia River, Spain, where a diversion structure fully dewatered the reach and let us obtain the entire streambed topography, which we used as spatially continuous control topography to compare the bathymetry surveyed from imagery during normal flow conditions. We compared this method with the small angle refraction correction (SARC), which has less data requirements, and analysed the error distribution across habitat types (i.e. pools and riffles). These results showed that our approach had smaller errors than SARC in both habitat types, especially in the riffle. By analysing the relationship between error and channel roughness obtained from our dry-bed model, we found that the greatest errors arise in the peripheral zones of the water surface and in its areas where streambed roughness generates more turbulence. Quantitative validation confirmed the reliability of our method as a relatively low-cost tool for the modelling and management of geomorphology and habitat within small – to medium-sized streams.

中文翻译：

---

中生境单元对淹没河流地形模型的空间验证

摘要 绘制河床形态图对于理解河流形态和过程、推进我们对河流系统的知识和最佳管理实践至关重要。它通常是通过涉水流完成的，但最近，地形测深光探测和测距 (LiDAR) 和从机载平台捕获的图像正在成为有前途和有效的测量方法。最近使用遥控航空系统 (RPAS) 结合运动结构 (SfM) 摄影测量处理提供了一种新颖的高分辨率方法来模拟河流形态。然而，在河流环境中获取此类数据的一个复杂因素与水体有关，由于空气-水界面处的光反射和折射，水体的存在会引入误差和扭曲。尽管概念验证研究表明在某些环境（例如清澈的海水、完整的表面）中可以减少折射的影响，但相应的验证方法仍然仅限于基于点的误差评估，这些通常仅限于可访问的频道的一部分。在这里，我们提供了第一个高分辨率、空间连续的验证，对使用 SfM 进行测量并使用高级折射方法进行校正的河流河段测深进行了验证。这种方法只需要相机坐标，并且可以开源，用 C++ 编码。我们使用了来自西班牙帕兰西亚河受管制河段的 RPAS 图像，在那里一个导流结构使河段完全脱水，让我们获得了整个河床地形，我们将其用作空间连续控制地形来比较在正常流动条件下从图像中测量的水深。我们将此方法与数据要求较少的小角度折射校正（SARC）进行了比较，并分析了栖息地类型（即水池和浅滩）的误差分布。这些结果表明我们的方法在两种栖息地类型中的误差都比 SARC 小，尤其是在浅滩。通过分析从我们的干床模型获得的误差与河道粗糙度之间的关系，我们发现最大的误差出现在水面的外围区域及其河床粗糙度产生更多湍流的区域。