Abstract Although numerous approaches for deriving water depth from bands of remotely-sensed imagery in the visible spectrum exist, digital terrain models for remote tropical carbonate landscapes remain few in number. The paucity is due, in part, to the lack of in situ measurements of pertinent information needed to tune water depth derivation algorithms. In many cases, the collection of the needed ground-truth data is often prohibitively expensive or logistically infeasible. We present an approach for deriving water depths up to 15 m in Case 1 waters, whose inherent optical properties can be adequately described by phytoplankton, using multi-spectral satellite imagery without the need for direct measurement of water depth, bottom reflectance, or water column properties within the site of interest. The reliability of the approach for depths up to 15 m is demonstrated for ten satellite images over five study sites. For this depth range, overall RMSE values range from 0.89 m to 2.62 m when using a chlorophyll concentration equal to 0.2 mg m−3 and a generic seafloor spectrum generated from a spectral library of common benthic constituents. Accuracy of water depth predictions drastically decreases beyond these depths. Sensitivity analyses show that the model is robust to selection of bottom reflectance inputs and sensitive to parameterization of chlorophyll concentration.

中文翻译：

---

在没有地面实况数据的情况下，从珊瑚礁景观中的多光谱图像中光学推导水深的算法

摘要 尽管存在多种从可见光谱中的遥感图像波段推导水深的方法，但用于偏远热带碳酸盐景观的数字地形模型仍然很少。部分原因是缺乏对调整水深推导算法所需的相关信息的原位测量。在许多情况下，收集所需的真实数据通常非常昂贵或在后勤上不可行。我们提出了一种在案例 1 水域中推导水深达 15 m 的方法，其固有的光学特性可以由浮游植物充分描述，使用多光谱卫星图像，无需直接测量水深、底部反射率或水柱感兴趣的站点内的属性。五个研究地点的十张卫星图像证明了该方法在深度达 15 m 时的可靠性。对于此深度范围，当使用等于 0.2 mg m-3 的叶绿素浓度和从常见底栖成分光谱库生成的通用海底光谱时，总体 RMSE 值范围为 0.89 m 至 2.62 m。超过这些深度，水深预测的准确性就会急剧下降。敏感性分析表明，该模型对底部反射输入的选择具有鲁棒性，并且对叶绿素浓度的参数化敏感。2 mg m-3 和从常见底栖成分光谱库生成的通用海底光谱。超过这些深度，水深预测的准确性就会急剧下降。敏感性分析表明，该模型对底部反射输入的选择具有鲁棒性，并且对叶绿素浓度的参数化敏感。2 mg m-3 和从常见底栖成分光谱库生成的通用海底光谱。超过这些深度，水深预测的准确性就会急剧下降。敏感性分析表明，该模型对底部反射输入的选择具有鲁棒性，并且对叶绿素浓度的参数化敏感。