distSG: Calculate across-network distance for a set of sparse points

In SpatialGraph: The SpatialGraph Class and Utilities

Calculate across-network distance for a set of sparse points

Description

This function obtains the across-network distance for a set of sparse points, by using the distance slot in a SpatialGraph. The calculation is supported by a previously calculated between vertex distance matrix [via a call to the library igraph by the function distSGv]. The SpatialGraph is considered as undirected for distance calculation. If euc=TRUE [default], the distance between two points is defined within this function as the maximum of both the minimum along-network distance and the Euclidean distance. The distance itself between the points in x,y and the network is neglected in the function for the along-network distance. Both, x and y, are SpatialPointsDataFrame objects, which must contain at least the fields eID and chain, which describe their relationship with the SpatialGraph object defined by SG. These can be obtained with either the function pointsSLDFchain or pointsToLines (the latter is faster, but depends on GEOS)

Usage

 distSG(SG, x, y = NULL, euc = TRUE, wei = NULL, getpath = FALSE)

Arguments

SG	SpatialGraph
x	SpatialPointsDataFrame
y	SpatialPointsDataFrame
euc	boolean scalar, whether to use Euclidean distance as minimum threshold for resulting distances
wei	if not null, field in SG@e with a variable to obtain a state-related weight. See details below.
getpath	if TRUE (and wei != NULL), eID identifiers for each path from x to y elements is returned

Details

The application of state-related weights in this version is a simple state-dependent weight matrix related to some field in SG@e [i.e. the edges in the input SpatialGraph]. The only current calculation evaluates the path between queried points (x,y), and along the path, for every junction and jump into a new edge, the ratio for the evaluated state variable (taken as the highest value divided by the lowest value) between the two edges at the junction is obtained. Currently a maximum ratio equal to 10.0 is hard-coded. The product of ratios along the path gives the weight.

Value

If wei=NULL, a matrix of distances between x and y. If wei is not NULL, a list with a distance matrix and weight matrix (plus a matrix with eID identifiers for the path, if getpath=TRUE) is returned.

Author(s)

Javier Garcia-Pintado, e-mail: [e-mail]

Examples

  if (1 > 2) { # not run
    dem <- readGDAL(file.path(system.file('external',package='hydrosim'),
                    'watershed1','IDRISI_maps','dem','dem.rst'))   # SpatialGridDataFrame
    plotGmeta(layer=dem, xlim=662500 + 2500 * c(-1,+1),
              ylim=4227500 + 2500 * c(-1,1), zlim='strloc', as.na=0)

    # generate some crossing lines
    zz <- list()
    zz[[1]] <- digitGmeta(layer=dem, type='Lines', ID=1)
    zz[[2]] <- digitGmeta(layer=dem, type='Lines', ID=2)
    zz[[3]] <- digitGmeta(layer=dem, type='Lines', ID=3)
    SL <- SpatialLines(zz)
    SG <- sl2sg(SL, getpath=TRUE)
    points(SG@v, cex=2)                    # plot SpatialGraph vertices

    apath <- SG@path[[1,2]]                # iteratively plot a path as an example
    for (iv in 1:length(apath$v)) {
      points(SG@v[apath$v[iv],], cex=2,pch=2)
      if (iv == length(apath$v))
        break
      lines(SG@e[apath$e[iv],],col='blue',lwd=2,lty=2)
      Sys.sleep(1)
    }

    # sample a few points [as a matrix] close to some edges
    xy    <- digit()                    # sample locations
    xych  <- pointsToLines(xy, SG@e)    # SpatialPointsDataFrame mapping
    points(xy, col='blue', pch=3)
    points(xych, col='darkgreen', pch=19)

    # along-network distance
    xyndis <- distSG(SG, xych)

    # state-dependent weighted along-network distance
    SG@e@data$wxs <- 3+round(runif(nrow(SG@e@data)),2)         # [m2] foo wetted cross-section areas
    SG@e@data

    xywdis <- distSG(SG, xych, wei='wxs')
    xywdis <- xywdis$dis * xywdis$wei       # Schur weight application into distance estimation
  }

SpatialGraph documentation built on Sept. 28, 2023, 5:08 p.m.