SpatGEVBMA.wrapper: Wrapper for spatial.gev.bma with netcdf files

In NorskRegnesentral/SpatGEVBMA: Hierarchical spatial generalized extreme value (GEV) modeling with Bayesian Model Averaging (BMA)

Wrapper for spatial.gev.bma with netcdf files

Description

This is a wrapper for running the spatial.gev.bma function in the SpatGEV package and imputing the results on a spatial grid with covariate values, based on a netcdf-file for the covariate grid, a spreadsheet with responses at observation stations and their spatial location in a text-file. The output is a netcdf-file with user specified posterior quantiles for the user specified return period, in addition to spatial maps of these and an optional Inter Quartile Range (IQR) uncertainty map. In addtion the input variables are written to file.

The grid covariates values are placed in netcdf-files in the folder 'covariate.folder'. The spatial coordinate system can be whatever you like as long as it is decimal spaced (not minutes and seconds) and the same coordinate system is used for the location of the stations (see below). The coordinates must either be specified by captial 'X' and 'Y' (set 'coordinate.type="XY"') or 'Lat' and 'Lon', with capital Ls (set 'coordinate.type="LatLon"'). All netcdf-files must give covariate values at the same spatial locations (there is currently no check for this)

The response data for stations are gathered in a spreadsheet located at 'station.annualMax.file', at spreadsheet named or numbered 'station.annualMax.sheet'. The first column of this spreadsheet contains the observation year, while first row contains the station number (starting from column 2)

The spatial locations for the station are given in a seperate table formatted txt-file with the first row containing the column names. A column named 'Stnr' must be present to denote the station number corresponding to those in the spreadsheet. If 'coordinate.type="XY"', then there must be columns 'X' and 'Y' specifying their spatial location in the same coordinate system as used in the netcdf covariate files. If 'coordinate.type="LatLon"', there must be columns 'Lat' and 'Lon' corresponding to the format in the netcdf-file.

All stations needs to be within the area covered by the grid of covariate values.

Results are imputed only in the locations where all covariates are available.

Usage

SpatGEVBMA.wrapper(covariates.folder, station.annualMax.file,station.annualMax.sheet = 1,
	station.locations.file, output.path = getwd(), output.folder.name = "SpatGEV.res",
	return.period = 20, post.quantiles = c(0.025,0.5,0.975), show.uncertainty = TRUE, 
	coordinate.type = "XY", table.format = "html", mcmc.reps = 10^5, 
	burn.in = round(mcmc.reps*0.2), cores = 1, annualMax.name = NULL, 
    create.tempfiles = FALSE, keep.temp.files = FALSE, save.all.output = TRUE, 
	testing = FALSE)

Arguments

covariates.folder	Path to folder with covariate files in netcdf-format (see above)
station.annualMax.file	File name of spreadsheet annualMax file (see above)
station.annualMax.sheet	The sheet name or index containing the station annualMax to be read (exactly 1 number)
station.locations.file	File name of table formatted textfile including the spatial locations of the stations
output.path	Path to the where the result folder should be stored
output.folder.name	Name of result folder
return.period	Return period to impute results for (single number or a vector of numbers)
post.quantiles	Vector of quantiles for which the posterior should be evaluated
show.uncertainty	Logical indicating whether an IQR uncertainty plot should also be provided
coordinate.type	Character indicating the type/name of coordinate system being used, either "XY" or "LatLon" (see above)
transform.output	Character specifying whether and how the output should be transformed. NULL corresponds to no transformation. "UTM_QQ_to_LatLon" transforms from UTM QQ (insert number) to LatLon
table.format	Character indicating the format for the covariate effect summary tables. Either "html" or "latex".
mcmc.reps	Number of MCMC runs for fitting the model with the station data. Should typically be at least be 10^5
burn.in	The length of the initial burn-in period which is removed
cores	The number of cores on the computer used for the imputation. Using detectCores()-1 is good for running locally.
annualMax.name	Name of annualMax data used in output plots and netcdf files. If NULL, then the name of the specified sheet is used.
create.tempfiles	Logical indicating whether temporary files should be saved in a Temp folder to perform debugging and check intermediate variables/results if the function crashes
keep.temp.files	Logical indicating whether the temporary files (if written) should be kept or deleted on function completion
save.all.output	Logical indicating whether all R objects should be save to file upon function completion. Allocates approx 2.5 Gb per return period for all of Norway.
testing	Variable indicating whether the run is a test or not. FALSE indicates no testing, a positive number indicates the number of locations being imputed

Value

This returns posteriors in netcdf file format, in addition to corresponding image plots.

Author(s)

Martin Jullum <[e-mail]>

NorskRegnesentral/SpatGEVBMA documentation built on July 22, 2023, 9:59 a.m.