Kaisa Riikka Ylinen, & Juha Pekka Kilpinen. (2018). Calibrating Ensemble Forecasts to Produce More Reliable Probabilistic Extreme Weather Forecasts. In Kees Boersma, & Brian Tomaszeski (Eds.), ISCRAM 2018 Conference Proceedings – 15th International Conference on Information Systems for Crisis Response and Management (pp. 1089–1097). Rochester, NY (USA): Rochester Institute of Technology.
Abstract: Accurate predictions of severe weather events are extremely important for society, economy, and environment. Due to the fact that weather forecasts are inherently uncertain, it is required to give information about forecast uncertainty to all users providing weather forecasts in probabilistic terms utilizing ensemble forecasts. Since ensemble forecasts tend to be under dispersive and biased, they need to be calibrated with statistical methods. This paper presents a method for the calibration of temperature forecasts using Gaussian regression, and the calibration of wind gust forecasts with a box-cox t-distribution method. Statistical calibration was made for the operational European Centre for Medium-Range Weather Forecasts (ECMWF) ensemble prediction system (ENS) forecasts for lead times from 3 to 360 hours. The verification results showed that calibration improved both temperature and wind gust ensemble forecasts. The probabilistic temperature forecasts were better after calibration over whole lead time scale, but the probabilistic wind gust forecasts up to 240 hours.
|
