BMA Model Fit to Precipitation Data

Share:

Description

The ensembleBMAgamma0 model fit with a 30 day training period to the precipitation data set from
http://www.stat.washington.edu/MURI, which gives daily daily 48 hour forecasts of 24 hour accumulated precipitation over the US Pacific Northwest region from December 12, 2002 through March 31, 2005 on a 9 member version of the University of Washington mesoscale ensemble (Grimit and Mass 2002; Eckel and Mass 2005). Precipitation amounts are quantized to hundredths of an inch.

Format

A list with the following arguments:

dateTable

A named vector in which the names are the dates and the entries are the number of observations for each date.

trainingRule

The training rule used to compute the model fits.

prob0coefs

The coefficients in the logistic regression for probability of zero precipitation.

biasCoefs

The coefficients in the linear regression for bias correction.

varCoefs

The variance coefficients of the models.

weights

The BMA weights for the models.

power

An scalar value giving the power by which the forecasts are transformed for the BMA fitting.

References

E. P. Grimit and C. F. Mass, Initial results of a mesoscale short-range ensemble forecasting system over the Pacific Northwest, Weather and Forecasting 17:192–205, 2002.

F. A. Eckel and C. F. Mass, Effective mesoscale, short-range ensemble forecasting, Weather and Forecasting 20:328–350, 2005.

J. M. Sloughter, A. E. Raftery, T. Gneiting and C. Fraley, Probabilistic quantitative precipitation forecasting using Bayesian model averaging, Monthly Weather Review 135:3309–3320, 2007.

C. Fraley, A. E. Raftery, T. Gneiting and J. M. Sloughter, ensembleBMA: An R Package for Probabilistic Forecasting using Ensembles and Bayesian Model Averaging, Technical Report No. 516R, Department of Statistics, University of Washington, 2007 (revised 2010).

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
## Not run:  # R check

  data(prcpFit)

  modelParameters(prcpFit, date = "20030113")

  data(prcpGrid) 

  prcpGridData <- ensembleData(forecasts = prcpGrid[,1:9], 
                               latitude = prcpGrid[,"latitude"],
                               longitude = prcpGrid[,"longitude"],
                               forecsatHour = 48,
                               initializationTime = "00")

# probability of precipitation
  1 - cdf( prcpFit, prcpGridData, value = 0)

# probability of precipitation above 0.25 in
  1 - cdf( prcpFit, prcpGridData, date = "20030115", value = 25)
  

## End(Not run)

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.