End-User Driven Streamflow Forecast Evaluation

Abstract:

High-resolution, continental-scale streamflow forecasts have the potential to provide valuable guidance to a wide range of water managers across the country. However, to realize this potential and increase beneficial use of the forecasts, both forecasters and end-users need an understanding of forecast skill and reliability in locations that stand to benefit from the information. Systematic forecast evaluations are not widely performed in part due to the technical complexity of such an effort, but also due to the fact that the interpretation of ‘skill’ (i.e. what makes a forecast useful or not) varies widely across sectors and end-users. To be truly meaningful, forecast evaluation should use an approach that considers the context of decisions being made and the perspectives of a given end-user. For example, some users have incorporated forecasts into their decision-making process for years, while others previously relied upon very limited information (e.g. rules of thumb). For the former, the use of forecasts is likely well-defined and a set of quantitative metrics for evaluation can be defined. For the latter, how forecasts are used may be less clear, making qualitative evaluation approaches more appropriate. In all cases, to assess the utility of forecasts the nature of the decisions and the baseline decision-making process (i.e. without forecasts) should be considered. Recent efforts to support end-users in using and evaluating forecasts across a range of circumstances and using a range of quantitative and qualitative approaches have demonstrated these points and will be presented.