multi.focal.function cuts out square or circular moving windows from a stack of grids (matrices) and applies a user-defined matrix function that takes multiple arguments to this data. multi.local.function is a more efficiently coded special case of moving windows of size 0, i.e. functions applied to individual grid cells of a stack of grids. This is especially useful for applying predict methods of statistical models to a stack of grids containing the explanatory variables (see Examples and grid.predict). The function is suitable for large grid files as it can process them row by row; but it may be slow because one call to the focal function is generated for each grid cell.
multi.focal.function(in.grids, in.grid.prefix, in.factor.grid, out.grid.prefix, path = NULL, in.path = path, out.path = path, fun, in.varnames, out.varnames, radius = 0, is.pixel.radius = TRUE, na.strings = "NA", valid.ranges, nodata.values = c(), out.nodata.value, search.mode = c("circle", "square"), digits = 4, hdr.digits = 10, dec = ".", quiet = TRUE, nlines = Inf, mw.to.vector = FALSE, mw.na.rm = FALSE, pass.location = FALSE, ...)
multi.local.function(in.grids, in.grid.prefix, out.grid.prefix, path = NULL, in.path = path, out.path = path, fun, in.varnames, out.varnames, na.strings = "NA", valid.ranges, nodata.values = c(), out.nodata.value, digits = 4, hdr.digits = 10, dec = ".", quiet = TRUE, nlines = Inf, na.action = stats::na.exclude, pass.location = FALSE, ...)in.path; in.grid.prefix will be used as a prefix to the file name if specified; default file extension: .asc-) will separate the prefix and the in.varnames-) will separate the prefix and the out.varnamesin.grids and write output grid files; see also in.path and out.path, which overwrite path if they are specifiedin.grids (defaults to path)pathfun is expected to accept named arguments with the names given by in.varnames; grid.predict is a wrapper function that can be used for applying a model's predict method to a stack of grids; see Details. In multi.local.function, fun must be able to process arguments that are vectors of equal length (e.g., a vector of 50 slope angles, another vector of 50 elevation values, etc.).in.grids; if missing, same as in.grids; if specified, must have the same length and order as in.gridsfun; if missing, multi.focal.function will try to determine the varnames from fun itself, or or from a call to fun if this is a function (see Details)is.pixel.radius and search.mode; note that all data within distance <=radius< code=""> will be included in the moving window, not . TRUE (default), the radius will be interpreted as a (possibly non-integer) number of pixels; if FALSE, it is interpreted as a radius measured in the grid (map) units.scanlength(in.grids) with numeric vector of length 2, specifying minimum and maximum valid values read from input file; all values or >valid.ranges[[i]][1] will be converted to NA. NA, in addition to the nodata value specified in the grid headerNAs in the output file(s); if missing, use the same nodata value as specified in the header of the input grid file"circle" (default) for a circular search window, or "square" for a squared one.write.ascii.grid.header).FALSE, gives some output ("*") after every 10th line of the grid file and when the job is done.NAs be removed from moving window prior to passing the data to fun? Only applicable when mw.to.vector=TRUE.fun? If TRUE, two additional arguments named arguments x and y are passed to fun; NOTE: This currently only works for radius=0, otherwise a warning is produced and pass.location is reset to FALSE.NA values are omitted from the stack of input variables; use na.exclude (default) or na.pass if fun can handle NA values correctlyfun; local.function: arguments to be passed to focal.function.multi.focal.function returns the character vector of output file names.
multi.local.function is probably most useful for applying the predict method of a fitted model to a grids representing the predictor variables. An example is given below and in more detail in Brenning (2008) (who used multi.focal.function for the same purpose); see also grid.predict.multi.local.function is essentially the same as multi.focal.function for radius=0, but coded MUCH more efficiently. (The relevant code will eventually migrate into multi.focal.function as well, but requires further testing.) Applying a GAM to the data set of Brenning (2008) takes about 1/100th the time with multi.local.function compared to multi.focal.function.
multi.focal.function extends focal.function by allowing multiple input grids to be passed to the focal function fun operating on moving windows. It passes square matrices of size 2*radius+1 to the function fun if mw.to.vector=FALSE (default), or a vector of length <=(2*radius+1)^2< code=""> if mw.to.vector=TRUE; one such matrix or vector per input grid will be passed to fun as an argument whose name is specified by in.varnames.
These matrices or vectors will contain the content of the moving window, which may possibly contain NAs even if the in.grid has no nodata values, e.g. due to edge effects. If search.mode="circle", values more than radius units (pixels or grid units, depending on is.pixel.radius) away from the center pixel / matrix entry will be set to NA. In addition, valid.range, nodata.values, and the nodata values specified in the in.grid are checked to assign further NAs to pixels in the moving window. Finally, if in.factor.grid specifies zones, all pixels in the moving window that belong to a different zone than the center pixel are set to NA, or, in other words, zone boundaries are used as breaklines.
The function fun should return a single numeric value or a numeric vector, such as a regression result or a vector of class probabilities returned by a soft classifier. In addition to the named arguments receiving the moving window data, fun may have additional arguments; the ... argument of focal.function is passed on to fun. grid.predict uses this feature.
Optionally, fun should support the following feature: If no argument is passed to it, then it should return a character vector giving variable names to be used for naming the output grids.
For the input files, .asc is used as the default file extension, if it is not specified by the user.
See focal.function for details.
focal.function, grid.predict
## Not run:
# # Assume that d is a data.frame with point observations
# # of a numerical response variable y and predictor variables
# # a, b, and c.
# # Fit a generalized additive model to y,a,b,c.
# # We want to model b and c as nonlinear terms:
# require(gam)
# fit <- gam(y ~ a + s(b) + s(c), data = d)
# multi.local.function(in.grids = c("a", "b", "c"),
# out.varnames = "pred",
# fun = grid.predict, fit = fit )
# # Note that the 'grid.predict' uses by default the
# # predict method of 'fit'.
# # Model predictions are written to a file named pred.asc
# ## End(Not run)
## Not run:
# # A fake example of a logistic additive model:
# require(gam)
# fit <- gam(cl ~ a + s(b) + s(c), data = d, family = binomial)
# multi.local.function(in.grids = c("a", "b", "c"),
# out.varnames = "pred",
# fun = grid.predict, fit = fit,
# control.predict = list(type = "response") )
# # 'control.predict' is passed on to 'grid.predict', which
# # dumps its contents into the arguments for 'fit''s
# # 'predict' method.
# # Model predictions are written to a file named pred.asc
# ## End(Not run)
Run the code above in your browser using DataLab