#### DMCA

## Log-normal distribution based EMOS models for probabilistic wind speed forecasting

### Citations

734 |
A description of the fifth generation Penn State/NCAR Mesoscale Model (MM5). National Center for Atmospheric Research Tech
- Grell, Dudhia, et al.
- 1994
(Show Context)
Citation Context ...ained from different runs of the fifth generation Pennsylvania State University–National Center for Atmospheric Research mesoscale model (PSU-NCAR MM5) with initial conditions from different sources (=-=Grell et al., 1995-=-). The EPS covers the Pacific Northwest region of western North America providing forecasts on a 12 km grid. Our data base (identical to the one used in Möller et al. (2013)) contains ensembles of 48... |

372 | Strictly proper scoring rules, prediction, and estimation.
- Gneiting, Raftery
- 2007
(Show Context)
Citation Context ...of the scoring rule to be optimized in order to obtain the estimates of parameters of different EMOS models. For evaluating density forecasts the most popular scoring rules are the logarithmic score (=-=Gneiting and Raftery, 2007-=-), i.e. the negative logarithm of the predictive PDF evaluated at the verifying observation, and the continuous ranked probability score (CRPS; Gneiting and Raftery, 2007; Wilks, 2011). Given a predic... |

286 |
The ECMWF ensemble prediction system; methodology and validation
- Molteni, Buizza, et al.
- 1996
(Show Context)
Citation Context ...lobal ensemble prediction system of the ECMWF consists of 50 exchangeable ensemble members which are generated from random perturbations in initial conditions and stochastic 5physics parametrization (=-=Molteni et al., 1996-=-; Leutbecher and Palmer, 2008). Forecasts of near-surface (10 meter) wind speed for lead times up to 10 days ahead are issued twice a day at 00 UTC and 12 UTC, with a horizontal resolution of about 33... |

196 | Ensemble forecasting at NCEP and the breeding method
- Toth, Kalnay
- 1997
(Show Context)
Citation Context ...tself, a possible solution is to run the models with different initial conditions resulting in an ensemble of forecasts (Leith, 1974). Since its first operational implementation (Buizza et al., 1993; =-=Toth and Kalnay, 1997-=-) the ensemble method has become a widely used technique all over the world. One of the leading organizations issuing ensemble forecasts is the European Centre for Medium-Range Weather Forecasts (ECMW... |

144 | Using Bayesian model averaging to calibrate forecast ensembles,”
- Raftery, Balabdaoui, et al.
- 2003
(Show Context)
Citation Context ... Figure 1) of the raw ensemble is the probability integral transform (PIT) histogram of the predictive distribution. The PIT is the value of the predictive CDF evaluated at the verifying observation (=-=Raftery et al., 2005-=-), and the PIT histogram provides a good measure about the possible improvements of the under-dispersive character of the raw ensemble. The closer the histogram is to the uniform distribution, the bet... |

80 | Calibrated probabilistic forecasting using ensemble model output statistics and minimum CRPS estimation,
- Gneiting, Raftery, et al.
- 2005
(Show Context)
Citation Context ...iate weather quantity is a single parametric density function, where the parameters depend on the ensemble members. In case of temperature and pressure the normal distribution is a reasonable choice (=-=Gneiting et al., 2005-=-), while for non-negative variables such as wind speed, a skewed distribution is required. A popular candidate is the Weibull distribution (see, e.g., Justus et al., 1978), gamma or log-normal distrib... |

64 | Ensemble forecasting.
- Leutbecher, TN
- 2008
(Show Context)
Citation Context ... the TN-LN mixture model is tested on forecasts of maximal wind speed of the eight-member University of Washington Mesoscale Ensemble (UWME, see e.g., Eckel and Mass, 2005) and of the ECMWF ensemble (=-=Leutbecher and Palmer, 2008-=-), and on instantaneous wind speed forecasts produced by the operational Limited Area Model Ensemble Prediction System of the Hungarian Meteorological Service (HMS) called ALADIN-HUNEPS (Hágel, 2010;... |

55 |
Weather forecasting with ensemble methods,”
- Gneiting, Raftery
- 2005
(Show Context)
Citation Context ...classical point forecasts (e.g. ensemble mean or ensemble median) using a forecast ensemble one can also estimate the distribution of a future weather variable which allows probabilistic forecasting (=-=Gneiting and Raftery, 2005-=-). However, the forecast ensemble is usually underdispersive and as a consequence, uncalibrated. This phenomenon has been observed with several operational ensemble prediction systems (see, e.g., Buiz... |

53 |
A comparison of the ECMWF, MSC, and NCEP global ensemble prediction systems
- Buizza, Houtekamer, et al.
- 2005
(Show Context)
Citation Context ...2005). However, the forecast ensemble is usually underdispersive and as a consequence, uncalibrated. This phenomenon has been observed with several operational ensemble prediction systems (see, e.g., =-=Buizza et al., 2005-=-). A possible solution to account for this deficiency is statistical post-processing. From the various modern post-processing techniques (for an overview see, e.g., Williams et al. (2014); Gneiting (2... |

49 | Making and evaluating point forecasts.
- Gneiting
- 2011
(Show Context)
Citation Context ...and means, which are evaluated with the use of mean absolute errors (MAEs) and root mean squared errors (RMSEs). Note that MAE is optimal for the median, while RMSE is optimal for the mean forecasts (=-=Gneiting, 2011-=-; Pinson and Hagedorn, 2012). 4.1 University of Washington Mesoscale Ensemble As the eight members of the UWME are non-exchangeable, the dependencies of location and scale parameters of the TN and GEV... |

37 |
Methods for estimating wind speed frequency distributions.
- Justus, Hargraves, et al.
- 1999
(Show Context)
Citation Context ...ion is a reasonable choice (Gneiting et al., 2005), while for non-negative variables such as wind speed, a skewed distribution is required. A popular candidate is the Weibull distribution (see, e.g., =-=Justus et al., 1978-=-), gamma or log-normal distributions are also in use (Garcia et al., 1988), while Thorarinsdottir and Gneiting (2010) suggested an EMOS model based on truncated normal distribution with a cut-off at z... |

33 |
Computation of optimal unstable structures for a numerical weather prediction model. Tellus 45A
- Buizza, Tribbia, et al.
- 1993
(Show Context)
Citation Context ... prediction process itself, a possible solution is to run the models with different initial conditions resulting in an ensemble of forecasts (Leith, 1974). Since its first operational implementation (=-=Buizza et al., 1993-=-; Toth and Kalnay, 1997) the ensemble method has become a widely used technique all over the world. One of the leading organizations issuing ensemble forecasts is the European Centre for Medium-Range ... |

33 | Probabilistic wind speed forecasting using ensembles and Bayesian model averaging,” - Sloughter, Gneiting, et al. - 2010 |

23 | Comparing Density Forecasts using Threshold- and Quantile-Weighted Scoring Rules”, - Gneiting, Ranjan - 2011 |

16 | Calibrating multimodel forecast ensembles with exchangeable and missing members using Bayesian model averaging. Monthly Weather Review 138:190–202
- Fraley, Raftery, et al.
- 2010
(Show Context)
Citation Context ...erification score (see Section 3.5). If the ensemble can be divided into groups of exchangeable members, ensemble members within a given group will get the same coefficient of the location parameter (=-=Fraley et al., 2010-=-) resulting in a predictive distribution of the form N0 ( a0 + a1 M1∑ `1=1 f1,`1 + · · ·+ am Mm∑ `m=1 fm,`m , b0 + b1S 2 ) , (3.2) where again, S2 denotes the ensemble variance. One might think of tak... |

15 | Z.: Uncertainties in COSMO-DE precipitation forecasts introduced by model perturbations and variation of lateral boundaries, Atmos - Gebhardt, Theis, et al. |

15 | Probabilistic forecasts of wind speed: ensemble model output statistics by using heteroscedastic censored regression,”
- Thorarinsdottir, Gneiting
- 2009
(Show Context)
Citation Context ...ack the maximum likelihood (ML) estimates of the parameters. Short calculation shows that the CRPS corresponding to the CDF G of a TN distribution N 0(µ, σ2) can be given in a closed form (see, e.g., =-=Thorarinsdottir and Gneiting, 2010-=-), namely CRPS (G, x) = σ[Φ(µ/σ)]−2 [x− µ σ Φ ( µ/σ )( 2Φ ( (x− µ)/σ)+ Φ(µ/σ)− 2) + 2ϕ ( (y − µ)/σ)Φ(µ/σ)− 1√ pi Φ (√ 2µ/σ )] . In case of the LN model one faces a similar situation, straightforward c... |

14 |
E.,1974: Theoretical skill of Monte-Carlo forecasts
- Leith
(Show Context)
Citation Context ... reliable initial conditions or from the numerical weather prediction process itself, a possible solution is to run the models with different initial conditions resulting in an ensemble of forecasts (=-=Leith, 1974-=-). Since its first operational implementation (Buizza et al., 1993; Toth and Kalnay, 1997) the ensemble method has become a widely used technique all over the world. One of the leading organizations i... |

13 |
B.: Probabilistic aspects of meteorological and ozone regional ensemble forecasts
- Monache, Hacker, et al.
(Show Context)
Citation Context ...am, each of which has expected relative frequency 1/c, and pi denotes the observed relative frequency in class i, can be used to quantify the deviation of the rank distribution from uniformity (Delle =-=Monache et al., 2006-=-). For the UWME ensemble, ∆ equals 0.6508, and the ensemble range contains the observed maximal wind speed in only 45.24 % of the cases (the nominal value of this coverage equals 7/9, i.e 77.78 %). He... |

8 | EnsembleBMA: An R package for probabilistic forecasting using ensembles and Bayesian model averaging
- Fraley, Raftery, et al.
- 2008
(Show Context)
Citation Context ...troduced by Raftery et al. (2005) and the ensemble model output statistics (EMOS) or non-homogeneous regression technique, suggested by Gneiting et al. (2005), as they are implemented in ensembleBMA (=-=Fraley et al., 2009-=-, 2011) and ensembleMOS packages of R. Both approaches provide estimates of the densities of the predictable weather quantities and once a predictive density is given, a point forecast can be easily d... |

7 | The ARPEGE/ALADIN mesoscale numerical modelling system and its application at the Hungarian Meteorological Service - Horányi, Kertész, et al. - 2006 |

7 | Latest developments around the ALADIN operational short-range ensemble prediction system in Hungary - Horányi, Mile, et al. - 2011 |

6 |
Ensemble Prediction at Météo France (poster introduction by Olivier Riviere
- Descamps, Labadier, et al.
- 2009
(Show Context)
Citation Context ...have their own ensemble prediction systems (EPS), e.g., the COSMO-DE EPS of the German Meteorological Service (DWD; Gebhardt et al., 2011; Bouallègue et al., 2013) or the PEARP EPS of Méteo France (=-=Descamps et al., 2009-=-). Besides calculating the classical point forecasts (e.g. ensemble mean or ensemble median) using a forecast ensemble one can also estimate the distribution of a future weather variable which allows ... |

5 | Statistical post-processing of probabilistic wind speed forecasting in - Baran, Horányi, et al. - 2013 |

5 | Available at: www.stat.washington.edu/research/reports/2008/tr516.pdf - Fraley, Raftery, et al. - 2011 |

4 | Multivariate probabilistic forecasting using ensemble Bayesian model averaging and copulas - Möller, Lenkoski, et al. - 2013 |

3 | Probabilistic wind speed forecasting using Bayesian model averaging with truncated normal components - Baran - 2014 |

2 | Enhancing COSMO-DE ensemble forecasts by inexpensive techniques - Bouallègue, Theis, et al. - 2013 |

2 | Fitting wind speed distributions: A case - Garcia, Torres, et al. - 1998 |

2 | Automated Surface Observing System (ASOS) Users Guide. Available at: http://www.weather.gov/asos/aum-toc.pdf - Pinson, Hagedorn - 1998 |

1 | Comparison of BMA and EMOS statistical calibration methods for temperature and wind speed ensemble weather prediction. Időjárás - Baran, Horányi, et al. - 2014 |

1 | Calibration of medium-range weather forecasts. ECMWF Technical Memorandum No. 719. Available at: old.ecmwf.int/publications/library/ecpublications - Gneiting - 2014 |

1 | Comparison of non-homogeneous regression models for probabilistic wind speed forecasting. Tellus A
- Lerch, Thorarinsdottir
- 2013
(Show Context)
Citation Context ... to quantify the improvement in twCRPS with respect to a reference predictive CDF Fref we make use of the threshold-weighted continuous ranked probability skill score (twCRPSS) defined as (see, e.g., =-=Lerch and Thorarinsdottir, 2013-=-) twCRPSS ( F, x ) := 1− twCRPS ( F, x ) twCRPS ( Fref , x ) . This score is obviously positively oriented, and as a reference we always use the predictive CDF corresponding to the TN model. 11 15 20 ... |