Skill of data based predictions versus dynamical models. A case study on extreme temperature anomalies
Copyright © 2016 American Geophysical Union
Reason for embargo
Under temporary embargo pending publisher permission.
This chapter considers extreme events as short-lived large deviations from a system's normal state. It compares the probabilistic predictions of extreme temperature anomalies issued by two different forecast schemes. One is a dynamical physical weather model, and the other is a simple data model. These two types of predictions are evaluated by proper skill scores and receiver operating characteristic (ROC) analysis, respectively. The chapter considers events as being extreme whenever they are in the uppermost or lowermost range of values for a given quantity. It discusses below extreme temperature anomalies, that is, large deviations of the surface temperature from its climatological average for the corresponding day of the year, which are to a good approximation Gaussian distributed. The chapter considers the performance of predictors for the temperature anomaly to overcome a given threshold on the following day for all possible threshold values. The target is the prediction of weather extremes.
This is the author accepted manuscript. The final version is available from Wiley via the DOI in this record
In: Extreme Events: Observations, Modeling, and Economics, edited by Mario Chavez, Michael Ghil and Jaime Urrutia-Fucugauchi