Which forecast errors matter most? Assessing public priorities in weather forecast accuracy

Abstract. Weather forecasts are widely used for day-to-day decision making, and the recipients of the forecasts rely on accurate information to be able to make the best possible decisions. However, despite advances in forecasting, errors can still occur such an incorrect prediction of timing, location, intensity or persistence of a weather event. Based on the results from two surveys, this work aims to investigate how important the accuracy of each of these forecast attributes is to the general public by asking them to rate forecasts where one of the attributes is not correct. We show that even though all attributes are considered important for a satisfactory and useful forecast, forecasts that underestimate the intensity of an event or assume intermittent rainfall while the event turns out to be continuous receive the lowest ratings. Notably, we also found that survey participants rate forecasts worse if they perceive them as missed events rather than false alarms. Understanding the public's priorities regarding forecast attributes may contribute to forecasters' decisions about which attributes to prioritize or help the development of additional verification methods that take these preferences into account when evaluating the usefulness of forecasts for their users.