Sensitivity Studies for a Hybrid Numerical–Statistical Short-Term Wind and Gust Forecast at Three Locations in the Basque Country (Spain)

Abstract

This study evaluates the performance of statistical models applied to the output of numerical models for short-term (1-24 h) hourly wind forecasts at three locations in the Basque Country. The target variables are horizontal wind components and the maximum wind gust at 3 h intervals. Statistical approaches such as persistence, analogues, linear regression, and random forest (RF) are used. The verification statistics used are coefficient of determination (R 2 ) and root mean square error (RMSE). Statistical models use three inputs: (1) Local wind observations; (2) extended EOFs (empirical orthogonal functions) derived from past local observations and ERA-Interim variables in a previous 24-h period covering a domain around the area of study; and (3) wind forecasts provided by ERA-Interim. Results indicate that, for horizons less than 1-4 h, persistence is the best model. For longer predictions, RF provides the best forecasts. For horizontal components at 4-24 h horizons, RF slightly outperformed ERA-Interim wind forecasts. For gust, RF performs better than ERA-Interim for all the horizons. Persistence is the most influential factor for 2-5 h. Beyond this horizon, predictors from the ERA-Interim wind forecasts led the contribution. Hybrid numerical-statistical methods can be used to improve short-term wind forecasts.Atmosphere 2020, 11, 45 2 of 22Wind predictions are issued based on meteorological numerical weather forecast (NWF) systems, in which data assimilation (DA) plays an important role in order to properly correct errors in the forecasting model by the use of observations in the data assimilation stages. In these NWF systems, the data assimilation methods optimally estimate the state of the atmosphere [19,20] and a numerical weather forecast model is used to predict the future states of the atmosphere. Two examples of these advanced NWP systems are the ERA-Interim [21] global atmospheric reanalysis model, from 1979 until 2019 [21], and NCEP re-analyses [22]. They are not real-time operational NWF systems, since the DA and the forecasting models are frozen in time, but they are re-run for the whole period using all the available observations commonly used in numerical weather forecasts in order to avoid, as much as possible, errors in the record due to changes in the structure (architecture, resolution, kind of observations processed) of the archives derived from operational NWF systems. In the case of the ERA-Interim re-analyses, besides analyses, forecasts are also provided starting at the analysis times. These forecasts are the ones used in this paper as a surrogate of an operational model. However, re-analyses are quite coarse-resolution models when compared with operational models of the same generation. In the particular case of the ERA-Interim, the spatial resolution is 0.75 • , too coarse for local wind forecasts.To fill this resolution gap, numerical and statistical downscaling techniques are customarily used. Numerical downscaling increases the resolution by nesting a higher-resolution model, such ...

A Methodology to Design a Wind Transfer Function: Application to the Valdevaqueros Dune (SW Spain)

“…After different linear and polynomial regression trials, the best way to fit the relationship was determined to be orthogonal regression [47][48][49][50][51]. The main difference between the least squares, used for speeds, and the orthogonal regressions, for directions, is the way to minimize the distance to the fitted line (vertical in the first one and perpendicular in the second one).…”