Skip to content

Conversation

cmarmstrong
Copy link

@cmarmstrong cmarmstrong commented Jun 28, 2019

Because vgmArea loops over source/target features, it's easy to parallelize. I've used this function for the past year or so with no issues, so I thought it's time to share it. It saves me bundles of time.

Basic usage is:

## form := formula object
## src/tgt := Spatial* objects
## vgm := gstat::vgm object
## nnodes := passed to parallel::makeCluster
function(form, src, tgt, vgm, nnodes, ...) {
    v <- gstat::variogram(form, data=src)
    vgm <- gstat::fit.variogram(v, vgm)
    krg <- gstat::krige0(form, src, tgt, gstat::parVgmArea, vgm=vgm, nnodes=nnodes, ...)
}

and a recreation of the example from demo/a2p.R:

Rprof()
# import NC SIDS data:
library(sp)
library(maptools)
fname = system.file("shapes/sids.shp", package="maptools")[1]
nc = readShapePoly(fname, proj4string = 
	CRS("+proj=longlat +datum=NAD27 +ellps=clrk66"))

# reproject to UTM17, so we can use Euclidian distances:
library(rgdal)
nc = spTransform(nc, CRS("+proj=utm +zone=17 +datum=WGS84 +ellps=WGS84"))

# create a target (newdata) grid, and plot:
grd = spsample(nc, "regular", n = 1000)
class(grd)
plot(nc, axes = TRUE)
points(grd, pch = 3)

library(gstat) # replace this with devtools::load_all('/path/to/gstat/branch')

# area-to-point kriging:
kr = krige0(SID74 ~ 1, nc, grd, parVgmArea, ndiscr = 9, 
	vgm = vgm(1, "Exp", 1e5, 0), # point variogram,
	nnodes=4)
out = SpatialPixelsDataFrame(grd, data.frame(pred = kr))

pl0 = spplot(nc["SID74"], main = "areas")
pl1 = spplot(out, sp.layout = list("sp.polygons", nc, first=F,col='grey'), 
    main = "points on a grid")
print(pl0, split = c(1,1,1,2), more = TRUE)
print(pl1, split = c(1,2,1,2), more = FALSE)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant