Least-cost path in meter

I wrote something to compute the along-path distance in meters using the
shortestPath function.
It works but it takes too much time. Advice to improve efficiency?

# I create an empty matrix of size n*n (n = number of individuals)
APD <- matrix(nrow=dim(xy)[1],ncol=dim(xy)[1])

# A loop by row and column
# I compute the shortestPath for all pairs of individuals
for (m in 1:dim(xy)[1]) {

     for(j in 1:m) {

path1 <- c("NA")

A <- cbind(xy[m,1],xy[m,2])
B <- cbind(xy[j,1],xy[j,2])


path1 <- try(shortestPath(tr1, A, B, output="SpatialLines"), silent = TRUE)

#try function in cases where individuals have the same location and thus
the shortestPath function does not work
if (class(path1)=="try-error") { sumdis <- 0 } else {

#I get the coordinates of the SpatialLines object created
   coo <- as.data.frame(coordinates(path1))
   dis <- c()
#I compute the distance between each successive paires of coordinates
   for (k in 1:c(dim(coo)[1]-1)) {
   dis <- c(dis, dist(rbind(coo[k,],coo[k+1,])))
}
#I sum the distances to get the along-path distance in meters
sumdis <- sum(dis)

}
#I fill the matrix
APD[m,j] <- sumdis

}
}





2013/11/6 Jeremy Larroque <larroque.jeremy at gmail.com>

Hi,



setting all the connection to 1 give me the straight line distance between
points, not the distance along the least-cost distance path.

Maybe a way to get the least-cost path in meters would be to count the
number of transitions and multiply it by the resolution of the raster?



Thanks,



Jeremy


2013/10/24 Jacob van Etten <jacobvanetten at yahoo.com>

Hi Jeremy,

Here goes a complete example. The function geoCorrection does the work.
It divides any value in the transition matrix by the distance between those
cells (e.g. cell centres). By setting all connections to 1, we get
conductance=1/distance

###

library("gdistance")

r <- raster(system.file("external/maungawhau.grd",
                         package="gdistance"))
plot(r)
r #units are in meters
tr1 <- transition(r,function(x){1},8) #gives all connection value 1
tr1 <- geoCorrection(tr1) #conductance=1/distance is equivalent to
resistance=distance

tr2 <- transition(r,function(x){1},16) #more precision
tr2 <- geoCorrection(tr2)

A <- c(2667670,6479000)
B <- c(2667800,6479400)
cc <- rbind(A,B)
points(cc)

costDistance(tr1, cc) #454
costDistance(tr2, cc) #431

#Pythagoras to check
sqrt((B[1] - A[1])^2 + (B[2] - A[2])^2) #421

###

Using more connections (16 instead of 8) gives a better result in this
case.

The method also works for latlon grids. See ?geoCorrection.

Jacob.


   On Thursday, 24 October 2013, 7:09, Jeremy Larroque <
larroque.jeremy at gmail.com> wrote:
  Hi everyone,



I'm using the gdistance package to compute least-cost distance, i.e. the
cumulative cost of the least-cost path, on a raster file between points
using the costDistance function.

But does anyone have an idea of how to calculate the distance in meter
along the least-cost path?



Thanks,


Jeremy

     [[alternative HTML version deleted]]

	[[alternative HTML version deleted]]