Predictability of streamflow across space and time scales

Evidence exists showing that the frequency of flood-related disasters is increasing around the world. Therefore, skillful prediction of future streamflow, particularly flooding, is crucial for flood mitigation efforts and water resources planning and management. However, such prediction systems or methods have several challenges and limitations. The streamflow forecasting systems inherit uncertainties or errors from several real-world hydrologic processes, which often limit our ability to forecast streamflow accurately. The challenge, therefore, is to find the balance between streamflow forecasting accuracy, forecast horizon, and basin size. However, studies providing a comprehensive assessment of the predictability of streamflow across space and time scales are still lacking. The overarching goal of this dissertation is to contribute to the understanding and discussion of uncertainties in streamflow forecasting methods and the resulting streamflow predictability. Achieving this goal requires a holistic framework using a broad spectrum of both data-driven and process-based methods, as well as an investigation of the associated performance of streamflow forecasting methods. The hydrologic forecasting community, as well as water resources planners and managers, often find it useful to have a model that is simple, yet that provides a reasonable performance. This dissertation also illuminates such a framework, which derives its strength from simple hydrologic insights and serves as a bound for any operational process-based or artificial intelligence-based models.