LAGOS-US RESERVOIR: Data module classifying conterminous U.S. lakes 4 hectares and larger as natural lakes or reservoirs

The LAGOS-US RESERVOIR data module (hereafter, RESERVOIR) classifies all 137,465 lakes > 4 hectares in the conterminous U.S. into one of the following three categories using a machine-learning predictive model based on visual interpretation of lake outlines and a classification rule based on lake shape. Natural Lakes (NLs) are defined as lakes that are likely to be entirely or mostly naturally-formed and that do not have large, flow-altering structures on or near them; Reservoir Class A’s (RSVR_A) are defined as lakes that are likely to be either human-made or highly human-altered by the presence of a relatively large water control structure that appears to significantly change the flow of water; and Reservoir Class B’s (RSVR_Bs) are lakes that are likely to be entirely human-made based on isolation from rivers and a highly angular shape that is rarely, if ever, seen in natural lakes also often. We trained the machine learning models on 12,162 manually-classified lakes to assign probabilities of a lake being in 1 of 2 of the categories (NL or RSVR), then we further classified the RSVR classification into either A or B based on NHD Fcodes, isolation, and angularity. The data module includes a detailed User Guide, metadata tables, and a data table that includes information such as location, lake geometry, surface water connectivity class, and official name. Using our definition, our classification indicates that over 46 % of lakes > 4 ha in the conterminous U.S. are reservoir lakes. These data can be combined with other LAGOS-US data modules and U.S. national databases using unique lake identifiers to study both reservoir lakes and natural lakes at broad scales.

LAGOS-US水库数据集模块（以下简称RESERVOIR）采用基于湖泊轮廓目视解译的机器学习预测模型，结合湖泊形态分类规则，将美国本土137465个面积大于4公顷的湖泊划分为以下三类：自然湖泊（Natural Lakes，NLs）指几乎完全或主要由自然作用形成，且湖体及周边无大规模改变水流的构筑物的湖泊；A类水库（Reservoir Class A’s，RSVR_A）指大概率为人工修建，或因存在规模较大的控水构筑物而显著改变水流的高度人工改造湖泊；B类水库（Reservoir Class B’s，RSVR_Bs）指完全人工修建的湖泊，这类湖泊与河流隔绝，且具有高度棱角分明的形态——此类形态在自然湖泊中极为罕见。本数据集基于12162个人工分类湖泊训练机器学习模型，以预测湖泊属于自然湖泊（NL）或水库（RSVR）两类之一的概率；随后再结合美国国家水文数据集（National Hydrography Dataset，NHD）的F代码、水系连通性及形态棱角度，将水库类别进一步划分为A、B两类。该数据集模块包含详细的用户指南、元数据表，以及包含湖泊位置、几何参数、地表水连通性类别、官方名称等信息的数据表。按照本分类标准，美国本土面积大于4公顷的湖泊中，超46%为水库湖泊。本数据集可通过唯一湖泊标识符与其他LAGOS-US数据集模块及美国国家级数据库整合，用于大尺度下的水库湖泊与自然湖泊相关研究。