This paper presents the geomorphic classification and bathymetry (at 100 m) of 12 major floodplains along a 900 km reach of the middle-lower Amazon River (Manaus – Monte Alegre; total floodplain area of 5164 km²), based on the inundation frequency (IF) derived from 36 yr of Landsat data. Using a mathematical relationship between IF and surveyed depth at the Curuai floodplain in June 2004, this IF-depth model is applied across floodplains of the middle-lower Amazon to estimate bathymetry. The applicability of the model is justified by similar grain size distribution (dominantly silt and clay) of sediment in the surface water of floodplain lakes (the main materials constructing the floodplains), as well as a hydrogeomorphic classification of this reach: seven floodplains lying on the alluvial plain dominated by the Amazon-flood pulse (Type I), three floodplains incised in valleys dominated by sediment-poor upland tributaries (Type II) and two floodplains of mixed types. This classification is used to distinguish the sensitivity of floodplains to the IF-depth model, where Type I floodplains of higher sensitivity provide a more accurate depth estimation. The bathymetry for both types is validated with field survey data across eight floodplains (covering over 200 km survey distance) collected in June – July 2016 and adjusted by the difference between mean June 2004 and June/July 2016 water level. Correlation coefficients of 0.87 (Type I) and 0.94 (Type II) indicate a strong relationship between estimated and measured depth for both floodplain types, while the root mean square error of 1.03 m (Type I) and 1.02 m (Type II) suggest that bathymetry is estimated to around 1 m error, equivalent to 10.9% and 10.8% of maximum water level variability, respectively. We offer the first field-validated and complete bathymetry map for the middle-lower Amazon River with accuracy that can be used to evaluate the role of floodplains in influencing biodiversity, sedimentation, flood control and biogeochemical cycling.

本文介绍了亚马逊河中下游 900 公里范围内 12 个主要洪泛区的地貌分类和水深测量（在 100 m 处）（马瑙斯 - 蒙特阿莱格里；洪泛区总面积为 5164 平方公里²)，基于来自 36 年 Landsat 数据的淹没频率 (IF)。使用 IF 与 2004 年 6 月 Curuai 洪泛区调查深度之间的数学关系，此 IF 深度模型应用于亚马逊中下游的洪泛区，以估计水深。泛滥平原湖泊（构成泛滥平原的主要材料）表层水中沉积物的相似粒度分布（主要是淤泥和粘土）以及该河段的水文地貌分类证明了该模型的适用性：七个漫滩位于以亚马逊洪水脉动为主的冲积平原（I 型），在以贫沙高地支流为主的山谷中切割的三个洪泛平原（II 型）和两个混合类型的洪泛平原。此分类用于区分洪泛区对 IF 深度模型的敏感性，其中灵敏度较高的 I 类洪泛区提供更准确的深度估计。两种类型的水深测量均通过 2016 年 6 月至 2016 年 7 月收集的八个洪泛区（覆盖超过 200 公里的调查距离）的实地调查数据进行验证，并根据 2004 年 6 月平均水位与 2016 年 6 月/2016 年 7 月水位之间的差异进行调整。0.87（类型 I）和 0.94（类型 II）的相关系数表明两种洪泛区类型的估计深度和测量深度之间存在很强的相关性，而 1.03 m（类型 I）和 1.02 m（类型 II）的均方根误差表明测深估计误差约为 1 m，分别相当于最大水位变化的 10.9% 和 10.8%。