## Probabilistic quantitative precipitation forecasting using Bayesian model averaging. Monthly Weather Review 135 (2007)

### Cached

### Download Links

- [www.stat.washington.edu]
- [www.stat.washington.edu:80]
- [stat.washington.edu]
- [www.stat.washington.edu]
- [www.stat.washington.edu]
- CiteULike

### Other Repositories/Bibliography

Venue: | Monthly Weather Review |

Citations: | 33 - 20 self |

### BibTeX

@ARTICLE{Sloughter07probabilisticquantitative,

author = {J. Mclean Sloughter and Adrian E. Raftery and Tilmann Gneiting and Mark Albright and Jeff Baars},

title = {Probabilistic quantitative precipitation forecasting using Bayesian model averaging. Monthly Weather Review 135},

journal = {Monthly Weather Review},

year = {2007},

volume = {135},

pages = {3209--3220}

}

### Years of Citing Articles

### OpenURL

### Abstract

and useful comments, and for providing data. They are also grateful to Patrick Tewson for implementing the UW Ensemble BMA website. This research was supported by the DoD Multidisciplinary University Research Initiative (MURI) program administered by the Office of Naval Research under Grant N00014-01-10745. Bayesian model averaging (BMA) is a statistical way of postprocessing forecast ensembles to create predictive probability density functions (PDFs) for weather quantities. It represents the predictive PDF as a weighted average of PDFs centered on the individual bias-corrected forecasts, where the weights are posterior probabilities of the models generating the forecasts and reflect the forecasts ’ relative contributions to predictive skill over a training period. It was developed initially for quantities whose PDFs can be approximated by normal distributions, such as temperature and sea-level pressure. BMA does not apply in its original form to precipitation, because the predictive PDF of precipitation is nonnormal in two major ways: it has a positive probability of being equal to zero, and it is skewed. Here we extend BMA to probabilistic quantitative precipitation forecasting. The predictive PDF corresponding to

### Citations

8614 | Maximum likelihood from incomplete data via the EM algorithm
- Dempster, Laird, et al.
- 1977
(Show Context)
Citation Context ... , fKst) is given by (4), with subscripts s and t added to y and fk. This cannot be maximized analytically, and instead we maximize it numerically using the expectation-maximization, or EM algorithm (=-=Dempster et al. 1977-=-; McLachlan and Krishnan 1997). The EM algorithm is a method for finding the maximum likelihood estimator when the problem can be recast in terms of unobserved quantities such that, if we knew what th... |

1065 | Bayes factors
- Kass, Raftery
- 1995
(Show Context)
Citation Context ...we use illustrative examples drawn from these data. Finally, in section 4 we discuss possible improvements to the method. 2 Methods 2.1 Bayesian model averaging Bayesian model averaging (Leamer 1978; =-=Kass and Raftery 1995-=-; Hoeting et al. 1999) was originally developed as a way to combine inferences and predictions from multiple statistical models, and was applied to statistical linear regression and related models in ... |

1016 |
The EM Algorithm and Extensions
- Mclachlan, Krishman
- 1996
(Show Context)
Citation Context ...4), with subscripts s and t added to y and fk. This cannot be maximized analytically, and instead we maximize it numerically using the expectation-maximization, or EM algorithm (Dempster et al. 1977; =-=McLachlan and Krishnan 1997-=-). The EM algorithm is a method for finding the maximum likelihood estimator when the problem can be recast in terms of unobserved quantities such that, if we knew what they were, the estimation probl... |

993 | Finite Mixture Models - McLachlan, Peel - 2000 |

254 |
On the Mathematical Foundations of Theoretical Statistics
- Fisher
- 1922
(Show Context)
Citation Context ..., and two predictor variables, the original forecast and the indicator variable for the forecast being equal to zero. We estimate wk, k = 1, . . . , K; c0; and c1 by the maximum likelihood technique (=-=Fisher 1922-=-) from the training data. The likelihood function is defined as the probability of the training data given the parameters to be estimated, viewed as a function of the parameters. The maximum likelihoo... |

103 |
Specification Searches
- Leamer
- 1978
(Show Context)
Citation Context ...ut the paper we use illustrative examples drawn from these data. Finally, in section 4 we discuss possible improvements to the method. 2 Methods 2.1 Bayesian model averaging Bayesian model averaging (=-=Leamer 1978-=-; Kass and Raftery 1995; Hoeting et al. 1999) was originally developed as a way to combine inferences and predictions from multiple statistical models, and was applied to statistical linear regression... |

99 |
Bayesian Model Averaging: A Tutorial (with discussion). Statist. Sci. 14, 382–417. Corrected version at http://www.stat.washington.edu/www/research/online/hoeting1999.pdf
- Hoeting, Madigan, et al.
- 1999
(Show Context)
Citation Context ...mples drawn from these data. Finally, in section 4 we discuss possible improvements to the method. 2 Methods 2.1 Bayesian model averaging Bayesian model averaging (Leamer 1978; Kass and Raftery 1995; =-=Hoeting et al. 1999-=-) was originally developed as a way to combine inferences and predictions from multiple statistical models, and was applied to statistical linear regression and related models in social and health sci... |

76 | 2005: Using Bayesian Model Averaging to Calibrate Forecast Ensembles - tery, Gneiting, et al. |

73 |
The use of Model Output Statistics (MOS) in objective weather forecasting
- Glahn, Lowry
- 1972
(Show Context)
Citation Context ...rate probabilistic precipitation forecasts based on deterministic forecasts. Regression techniques such as model output statistics (MOS) can be used to generate probabilities of exceeding thresholds (=-=Glahn and Lowry 1972-=-; Klein and Glahn 1974; Bermowitz 1975; Charba 1998; Antolik 2000), or to generate quantiles of expected precipitation (Bremnes 2004). Applequist et al. (2002) found that logistic regression can outpe... |

59 | A method for producing and evaluating probabilistic forecasts from ensemble model integrations - Anderson - 1996 |

36 | Calibrated probabilistic forecasting using ensemble model output statistics and minimum CRPS estimation - Gneiting, Raftery, et al. - 2005 |

35 | 2002: Initial results of a mesoscale short-range ensemble forecasting system over the Pacific Northwest - Grimit, Mass |

30 |
A Model Fitting Analysis of Daily Rainfall Data
- Stern, Coe
- 1984
(Show Context)
Citation Context ...y. The second part of our model specifies the PDF of the amount of precipitation given that it is not zero. Previous authors have fit gamma distributions to precipitation amounts (Coe and Stern 1982; =-=Stern and Coe 1984-=-; Wilks 1990; Hamill and Colucci 1998; Wilson et al. 1999), as they can fit skewed data and are quite flexible, and we also took the gamma distribution as our starting point. The gamma distribution wi... |

25 | 2005: A comparison of the ECMWF, MSC, and NCEP global ensemble prediction systems - Buizza, Houtekamer, et al. |

24 | Calibrated probabilistic quantitative precipitation forecasts based on the MRF ensemble - Eckel, Walters - 1998 |

19 | Calibrated probabilistic mesoscale weather field forecasting: The geostatistical output perturbation method - Gel, Raftery, et al. |

19 | 2001), Quantitative precipitation forecasts over the United States by the ECMWF Ensemble Prediction System - Mullen, Buizza |

17 | 2004: Ensemble re-forecasting: improving medium range forecast skill using retrospective forecasts - Whitaker, Wei |

16 | 1998a: Evaluation of the Eta-RSM ensemble probabilistic precipitation forecasts
- Hamill, Colucci
(Show Context)
Citation Context ...specifies the PDF of the amount of precipitation given that it is not zero. Previous authors have fit gamma distributions to precipitation amounts (Coe and Stern 1982; Stern and Coe 1984; Wilks 1990; =-=Hamill and Colucci 1998-=-; Wilson et al. 1999), as they can fit skewed data and are quite flexible, and we also took the gamma distribution as our starting point. The gamma distribution with shape parameter α and scale parame... |

12 | 2000: An overview of the National Weather Service’s centralized statistical quantitative precipitation forecasts - Antolik |

12 |
Forecasting local weather by means of model output statistics
- Klein, Glahn
- 1974
(Show Context)
Citation Context ...cipitation forecasts based on deterministic forecasts. Regression techniques such as model output statistics (MOS) can be used to generate probabilities of exceeding thresholds (Glahn and Lowry 1972; =-=Klein and Glahn 1974-=-; Bermowitz 1975; Charba 1998; Antolik 2000), or to generate quantiles of expected precipitation (Bremnes 2004). Applequist et al. (2002) found that logistic regression can outperform standard regress... |

11 | 2004: Probabilistic forecasts of precipitation in terms of quantiles using NWP model output - Bremnes |

11 |
Fitting models to daily rainfall data
- Coe, RD
- 1982
(Show Context)
Citation Context ...forecast for that day. The second part of our model specifies the PDF of the amount of precipitation given that it is not zero. Previous authors have fit gamma distributions to precipitation amounts (=-=Coe and Stern 1982-=-; Stern and Coe 1984; Wilks 1990; Hamill and Colucci 1998; Wilson et al. 1999), as they can fit skewed data and are quite flexible, and we also took the gamma distribution as our starting point. The g... |

10 | 2002: Comparison of Methodologies for Probabilistic Quantitative Precipitation Forecasting - Applequist, al |

10 | A 30% chance of rain tomorrow”: How does the public understand probabilistic weather forecasts? Risk Analysis - Gigerenzer, Hertwig, et al. - 2005 |

10 | Probabilistic quantitative precipitation forecasts for river forecasting - Krzysztofowicz, Drake - 1992 |

9 |
Maximum likelihood estimation for the gamma distribution using data containing zeros
- Wilks
- 1990
(Show Context)
Citation Context ...f our model specifies the PDF of the amount of precipitation given that it is not zero. Previous authors have fit gamma distributions to precipitation amounts (Coe and Stern 1982; Stern and Coe 1984; =-=Wilks 1990-=-; Hamill and Colucci 1998; Wilson et al. 1999), as they can fit skewed data and are quite flexible, and we also took the gamma distribution as our starting point. The gamma distribution with shape par... |

6 | Generic probability distribution of rainfall in space: the bivariate model - HD, Krzysztofowicz |

4 |
An application of model output statistics to forecasting quantitative precipitation
- Bermowitz
- 1975
(Show Context)
Citation Context ...ased on deterministic forecasts. Regression techniques such as model output statistics (MOS) can be used to generate probabilities of exceeding thresholds (Glahn and Lowry 1972; Klein and Glahn 1974; =-=Bermowitz 1975-=-; Charba 1998; Antolik 2000), or to generate quantiles of expected precipitation (Bremnes 2004). Applequist et al. (2002) found that logistic regression can outperform standard regression, and Hamill ... |

4 | 2000: Simulation of precipitation fields from Probabilistic Quantitative Precipitation Forecast - Perica, Seo, et al. |

4 |
A strategy for verifying weather element forecasts from an ensemble prediction system
- WILSON, BURROWS, et al.
- 1999
(Show Context)
Citation Context ...amount of precipitation given that it is not zero. Previous authors have fit gamma distributions to precipitation amounts (Coe and Stern 1982; Stern and Coe 1984; Wilks 1990; Hamill and Colucci 1998; =-=Wilson et al. 1999-=-), as they can fit skewed data and are quite flexible, and we also took the gamma distribution as our starting point. The gamma distribution with shape parameter α and scale parameter β has the PDF g(... |

2 | Observations QC summary page — http://www.atmos.washington.edu/ mm5rt/qc obs/qc obs stats.html - Baars - 2005 |

2 |
The LAMP QPF products. Part I: Model development
- Charba
- 1998
(Show Context)
Citation Context ...istic forecasts. Regression techniques such as model output statistics (MOS) can be used to generate probabilities of exceeding thresholds (Glahn and Lowry 1972; Klein and Glahn 1974; Bermowitz 1975; =-=Charba 1998-=-; Antolik 2000), or to generate quantiles of expected precipitation (Bremnes 2004). Applequist et al. (2002) found that logistic regression can outperform standard regression, and Hamill et al. (2004)... |

1 | 2005: Effective mesoscale, short-range ensemble forecasting - Mass |