Anders K. Ångström was known primarily for his contributions to the field of atmospheric radiation. However, his scientific interests encompassed many diverse topics. This paper describes the contents of two early, remarkable, and, until recently, largely unknown papers by Ångström on probability forecasting and the use/value of weather forecasts. These papers, entitled “Sannolikhet och Praktisk Vaderleksprognos” (“Probability and Practical Weather Forecasting”) and “On the Effectivity of Weather Warnings,” were published in 1919 and 1922, respectively.
Noteworthy features of these two papers include 1) a discussion of the sources of uncertainty in weather forecasting, 2) discourses on the problems of estimating probabilities by means of empirical relative frequencies and forecasters' subjective judgments, 3) the use of a Gaussian model to describe the accuracy of minimum temperature forecasts, 4) the identification of the ratio of the protection costa to the “risked value” if protective action is not taken b as a characteristic of users of forecasts, 5) analytical expressions for the economic value of weather warnings, 6) quantitative analysis of the problems faced by forecasters in deciding whether or not to issue weather warnings when they are uncertain about future weather conditions, and 7) arguments concerning the need to obtain estimates of the costs and losses that may be incurred by potential users in order to assess economic effectiveness.
The contents of Ångström's two papers are reviewed and summarized, making extensive use of quotations from the texts. An effort is made to place the papers and their contents in proper historical context. Two topics of current interest, namely, ensemble forecasting and the provision of specialized weather services, are discussed briefly in light of the results presented and issues raised in these papers.
*Swedish Meteorological and Hydrological Institute, Norrköping, Sweden.
+Max-Planck-Institut für Meteorologie, Hamburg, Germany.
**Permanent affiliation: Prediction and Evaluation Systems, Corvallis, Oregon.