Web‐based systems allow users to filter data and resources using river network structure on interactive map environments that use server side processing. With the increasing resolution of river networks, optimized methods for storage of river network representation in databases and efficient queries on the river network structure become critical. This paper presents a detailed study of widely used methods for representing generic networks in relational databases and benchmarking common queries on river network data using these methods. The analysis has been applied to a data set consisting of the river network of Iowa, with over 620,000 individual subwatersheds/nodes in the network. For typical river network queries (path to the outlet; contributing watersheds), two of considered data models (Adjacency List and Nested Set) guarantee response times below 1 s. This will allow users to carry out large‐scale analysis and visualizations tasks on the web for hydrological data sets. A new data model, Stream Network, is proposed based on Path Enumeration, stores directly additional hydrological information, enabling, for example, retrieval of the main stream.