Scalable, Adaptive, and Risk-Informed Design of Hydrological Sensor Networks

Stream monitoring networks are essential for understanding and managing Earth’s water resources, yet their deployment is rarely coordinated at the system scale to meet these objectives. We present a data-driven framework for the design of streamflow monitoring networks that improves hydrologic predictions while also accommodating socio-environmental constraints. This approach uses a rank-revealing QR decomposition to isolate monitoring sites that best capture the spatio-temporal structure of hydrologic time series obtained from retrospective simulations. Evaluated using 44 years of reanalysis data, we find that our sensor placement approach enables better reconstructions of streamflow at ungauged locations compared to existing methods. Our approach accommodates incremental expansion of existing gauge networks and integrates operational priorities—such as flood risk—without compromising the accuracy of hydrologic predictions. Demonstrated across diverse hydrological regimes, this framework provides a scalable and robust method for gauge network design that will empower water managers to make more informed decisions.