Uncertainty in River Forecasts: Quantification and Implications for Decision- Making in Emergency Management

This dissertation focuses on (river) forecasting, but also includes a study on stormwater treatment. Using forecasts for decision-making is complicated by their inherent uncertainty. An interview-based study qualitatively and a survey empirically investigate forecast use in emergency management. Emergency managers perceive uncertainty as a given rather than as a problem. To cope with the uncertainty, decision-makers gather as much information as possible; forecasts are only one piece of information among many. For decision-making, emergency managers say that they rely more on radar than on river forecasting. However, forecasts play an important role in communication with the public, because they are the official interpretation of the situation. Emergency managers can add a lot of value to those forecasts by combining them with local knowledge, but might not do so because of accountability concerns. Forecasts must have value to emergency managers, because those with more work experience rely more on them than those without. Another study further develops the application of quantile regression to generate probabilistic river forecasts. Compared to existing research, this study includes a larger number of river gages; includes more independent variables; and studies longer lead times. Additionally, it is the first to apply this method to the U.S. American context. It was found that the model has to be customized for each river gage for extremely high event thresholds. For other thresholds and across lead times, a one-size-fits-all model suffices. The model performance is robust to the size of the training dataset, but depends on the year, the river gage, lead time and event threshold that are being forecast. An additional study considers the robustness of stormwater management to the amount of runoff. Impervious surfaces, such as roads and parking lots, can increase the amount of runoff and lead to more pollution reaching streams, rivers, and lakes. Best Management Practices (BMPs) reduce the peak discharge into the storm sewer system and remove pollutants such as sediments, phosphorus and nitrogen from the stormwater runoff. Empirically, it is found that BMP effectiveness decreases sooner, steeper and deeper with increasing sizes of storm events than assumed in current computer models.