9th HEPEX webinar: Five points to remember in risk forecasting

Speaker: Anders Persson

Date and time: Thursday, October 24, 2013 3:00 pm GMT Summer Time (London, GMT+01:00)

Register: https://ecmwf.webex.com/ecmwf/onstage/g.php?t=a&d=954137752

Abstract: The ultimate criterion of a “good” forecast is always the decisions made from it. In uncertainty or probability forecasting this involves:

1. Combat our human tendency of over-confidence, in particular our habit of looking for confirmation of our ideas more than possible refutations
2. Learn not to underestimate the effects of randomness, in particular our tendency to see regularities is random events
3. Probabilities are “objectively” calculated in different systems, but they might not agree. The human forecasters must also with short notices subjectively consider and weight in new or unexpected information from forecasts or observations. Doing this there are several mathematical and psychological traps to avoid
4. Communicating the uncertainty information can be done both quantitatively and qualitatively, in percentage, intervals or odds. The probabilities often make more sense when projected against some comparison, for example the “base rate” (climatology) The spatial and temporal resolutions also matter
5. Drawing the right conclusions from well understood forecast information is not always trivial. Most people have an intuitive feeling for “cost” och “penalty” functions, but not how their formulations might affect the decision making

Good forecasting is not only a matter of “getting it right” but also make the receivers understand it, remember it and be able to draw conclusions from it.

About the speaker: Anders has over 40 years of experience in forecasting and has worked as research meteorologists at ECMWF, SMHI and UK Metoffice. Even though a meteorologists by training, he became acquainted with hydrological forecasting when he assisted in the setting up of such a service at SMHI.