Traditional operational hydrologic forecasting relies on legacy conceptual watershed models coupled with expert forecaster judgment – a medley of scripted processes and manual adjustments to data, analyses, and products made by trained hydrologists. While this approach benefits from the rich experience of hydrologists, the manual effort and ad hoc workflows limit the capacity to test the potential of new scientific approaches. Because the current practice is not reproducible, it cannot quantify forecast uncertainties, which decision makers increasingly require to balance risks and opportunities during challenging hydrometeorological events. Due to the reliance on hydrologists to adjust models in real-time, this practice is also unable to scale up to meet growing user needs by generating predictions at higher resolutions (i.e., at more locations), higher frequency and for more variables (e.g., showing inundated areas, snowfall or soil moisture).
Recent decades have seen numerous scientific and technological advances to meet these needs. These are now beginning to filter into operational streamflow forecast practice in centers around the world. Technology advances related to computing, data storage, and connectivity provide a foundation for transforming the computational side of streamflow prediction, while high-potential research can be found in the areas of remote sensing, physical, distributed earth system modeling, parameter estimation, data assimilation, verification, statistical post-processing, weather and climate prediction, and uncertainty estimation through the use of ensembles.
Scientists and engineers in RAL’s Hydrometeorological Applications Program at the National Center for Atmospheric Research are undertaking research in new directions to facilitate the transition of these advances into operational streamflow forecasting practice in the US.