The collective value of weather probabilistic forecasts according to public threshold distribution patterns

ABSTRACT The Korea Meteorological Administration has limited understanding of how users interpret probabilistic forecast information and use it in their decision‐making processes. Thus, a survey was conducted among users to find out at which probability threshold for adverse weather they would react and take protective action to minimize the losses. According to previous studies, probability threshold depends on the type of users but they are in general higher than a 50% forecasts chance of adverse weather, even though research has shown that the forecast value is maximized when actions are taken for probability of adverse weather lower than 50%. A collective value score model, defined as a weighted sum of user's satisfaction and threshold distribution in a group of users, is introduced as a measure of total satisfaction of a user group. The collective value score model is applied to a set of precipitation probabilistic forecasts of Seoul, South Korea during 2002–2011. The results show that the collective value score can be improved if the range of users' probability thresholds is widened rather than improving the forecast accuracy.