Naslov
Wind speed ensemble predictions with an analog-based method in complex terrain

Autori
Odak Plenković, Iris ; Delle Monache, Luca ; Horvath, Kristian ; Hrastinski, Mario ; Bajić, Alica

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Skup
Workshop on Advances in Meso- and Micrometeorology

Mjesto i datum
Donja Stubica, Hrvatska, 03.11.2014. - 04.11.2014

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
short-range predictions; deterministic forecasting; complex terrain; statistical techniques; analog-based method

Sažetak
The chaotic nature of the atmosphere and the imperfections of mesoscale models mean that there is a need to improve them even more, as well as provide information about their uncertainty. Analog ensemble (AE) method can do both. The first step to build AE method is the search for similar (i.e. analogs) past predictions across several variables (e.g., wind speed, wind direction, temperature) to the current prediction. Then, the measurements corresponding to the analogs form the AE. The AE can be used to generate both deterministic and probabilistic short- or medium-range forecasts. The AE was generated by ALADIN mesoscale model run over two nested domain with 8 and 2 km horizontal resolution, respectively and simplified, dynamically downscaled model with 2 km grid spacing (DADA). Several different deterministic AE forecasts were tested (AE mean and median, AE weighted mean, Kalman filter of AE means etc.) at three climatologically different regions across Croatia for point-based wind speed predictions at 10 m height. Results were verified and compared to starting model to address the following questions: • How many analogs is needed to produce the optimal results? • How much these analog – based methods improve starting model and what is the difference between them? • What is the impact of the starting model on the performance of the AE? The verification procedure includes several metrics computed considering wind speed as continuous and categorical predictand. This study shows that deterministic AE predictions, compared to model used to generate it and Kalman filter of a starting model, improve rank correlation coefficient between predictions and measurements and reduce bias and root-mean-square error, especially in complex terrain. They produce similar results regardless of the starting model. Also, probabilistic AE predictions provide reliable information about their uncertainty.

Izvorni jezik
Engleski

Znanstvena područja
Geologija

Napomena
IPA projekt WILL4WIND, IPA2007/HR/16IPO/001-040507

POVEZANOST RADA