Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'The model chain: First steps towards tomorrow's technology' taking place on Thursday, 13 March 2014 at 09:00-10:30. The meet-the-authors will take place in the poster area.

Kristian Horvath Meteorological and Hydrological Service, Croatia
Kristian Horvath (1) F P
(1) Croatian Meteorological and Hydrological Service, Zagreb, Croatia

Printer friendly version: printer.gif Print


Verification of 10-M wind speed obtained by dynamical downcaling in the complex terrain of Croatia


Knowing the wind properties is a key to the assessment of wind energy resources and for wind power forecasting. In the complex terrain and coastal regions, where a significant portion of wind energy arises from regional/local winds, it is beneficial to utilize a chain of numerical models to dynamically refine the associated wind fields. The principal questions we address are i) whether an increase of model chain resolution improves accuracy, and ii) could Computational Fluid Dynamics (CFD)-like, simplified and computationally cheaper meteorological models be used in the model chain for assessment and forecasting of wind properties?


A common method used to estimate wind climate in the complex terrain is dynamical downscaling by Mesoscale Numerical Weather Prediction (MNWP) models, with several refinements depending on the selected resolutions of MNWP model. Croatian Meteorological and Hydrological Service (CMHS) uses ALADIN MNWP model with 8 km horizontal resolution and refined CFD-like model version at 2 km horizontal resolution (DADA) adapted for operational forecast of 10-m wind. The aim of this study is an assessment of the ability of mesoscale models at different resolutions to reproduce relevant wind speed climate in the complex terrain of Croatia by method of statistical verification.

Main body of abstract

When applied to case studies, DADA has proven to be quite successful in weather situations with strong bora wind (20 ms-1 or more). This is particularly important when estimating high wind speeds, where small errors can have a significant impact on the assessment of wind power (proportional to cube of wind speed). About two years ago, trial version of full ALADIN 2 km model was launched at CMHS. Obtained wind speed forecasts from 8 km grid spacing model simulations, 2 km grid spacing model simulations and DADA forecasts were compared with measurements for period of 2010-2012. To avoid interpolation of the model outputs, we have chosen one of four neighboring grid points that best describes the orography around the measuring station. Monthly wind speed averages and wind roses have shown good agreement for majority of analyzed stations and whole three datasets. Statistical verification procedure was performed considering wind speed as continuous and categorical variable. The continuous statistical scores, such as multiplicative mean systematic error (MBIAS), root-mean-square error (RMSE) and mean absolute error (MAE) were calculated and averaged over monthly periods to show their seasonal variability. The same procedure was performed for categorical statistical scores like Heidke skill score (HSS), critical success index (CSI), frequency bias (FB), etc. Based on variety of statistical scores, DADA forecasts have proven to be quite successful in forecasting wind properties at majority of locations. Monthly averages of wind speed are well matched with the measurements, but the variance is somewhat smaller due to overestimation of low speeds.


The results showed that DADA and ALADIN 2 km full version forecasts are of similar quality and that is worth investing in development of downscaling methods which enable saving of time and computing space. As expected, both versions of the 2 km grid model represent an improvement over the 8 km grid spacing model. Finally, we argue that technologically optimal dynamical model chain should have different components of progressively smaller complexity with increasing the model resolution.

Learning objectives
1) Compare different assessment results and evaluate the performance of statistical verification
2) Learn about how an increase of chain of models resolution affects an accuracy of wind forecasts when applied in complex terrain
3) Discuss advantages and disadvantages of CFD-like numerical models used in the model chain for assessment and foresting of wind properties