Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 17: Poster I
DY 17.39: Poster
Dienstag, 26. Februar 2008, 16:00–18:00, Poster C
Weather Roulette---Game Theoretical Concepts For Evaluating Forecasts — •Jochen Bröcker1, Mark Roulston2, and Leonard Smith3 — 1Max Planck Institute for the Physics of Complex Systems, Dresden, Germany — 2London School of Economics, London, United Kingdom — 3Met Office, Exeter, United Kingdom
Probabilistic forecasts are (at least in principle) superior to deterministic forecasts, as the former allow for a better assessment of potential risks. In weather and climate for example, probabilistic forecasts have therefore been a longstanding aim. In these sectors, the use of ensembles to convey probabilistic information has become more and more common.
Different from deterministic forecasts, probabilistic forecasts (ensemble forecasts or other) cannot simply be evaluated by just how far they missed truth. But how do we know then if our probabilistic forecast is ``good''? In this contribution, concepts for evaluating probabilistic forecasts are discussed. Using a game theoretic approach, we investigate the income stream the forecast would generate in a lottery--like setup named ``weather roulette''. Thus the value of the forecast is measured directly in terms of money.
We discuss to what extend a good performance in weather roulette portends general virtues of a forecast. It turns out that, when implemented correctly, weather roulette does not encourage hedging. Numerical results are presented using ensemble weather forecasts for temperature and pressure at several locations.