Appendix 1: R script detailing the approach used in this study.
# Simplified version of the script used in this study. Provide an overview of the general method employed.
#------------------------------------------------------------------------
# create and plot background matrix with artificial barrier in middle
m <- matrix(1, nrow=10, ncol=10) ; m
m[,5] <- 4

library(raster)
r <- raster(m)
plot(r)

# create and plot transition matrix
library(gdistance)
t <- transition(r, transitionFunction=mean, 4, symm=TRUE, intervalBreaks=3)
plot(raster(t))

# create and plot sampling points and genetic data associated.
# (x coordinates, y coordinates, genetic data for 3 loci)
matG2 <- matrix(c(0.21, 0.22, 0.82, 0.23, 0.81, 0.83, 0.81, 0.21, 0.50, 0.51, 0.23, 0.83, 0, 0, 2, 0, 1, 1, 1, 2, 1, 1,
1, 0, 2, 1, 0, 1, 0, 0), ncol=5)
xcoord<- matG2[, 1] ; ycoord <- matG2[, 2]
P<-cbind(xcoord,ycoord)
points(P)


# construction of moving windows (sampling areas)
library("ade4") ; library("vegan")

# Define the extent of sampling areas and the interval wanted
Min <- 0.7 # Minimum radius of areas wanted
Max <- 0.9 # Maximum radius of areas wanted
Step <- 0.1 # interval wanted

# Loops to test correlations in sampling area at multiple scales and locations
resultsfinal <- cbind(1,1,1,1,1)
colnames(resultsfinal) <- c("xcoord","Ycoord", "Radius", "MantelR", "Pval")
for(Radius in seq(Min, Max, by = Step)){
     results = NULL
     for(i in 1:length(xcoord)){
         Xcircle <- ( xcoord [i] + Radius*cos(seq(0,2*pi,length.out=100)))
         Ycircle <- ( ycoord [i] + Radius*sin(seq(0,2*pi,length.out=100)))
         polygon(Xcircle, Ycircle)

         # extract individuals data in each sampling are constructed
         expr <- point.in.polygon(xcoord,ycoord,Xcircle,Ycircle)
         xcoord[expr==1]
         ycoord[expr==1]
         coordPoly <- cbind (xcoord[expr==1],ycoord[expr==1])

         # sort data and compute matrix of basic pairwise euclidian distances (not used further in this example)
         CoordOrder<- coordPoly[order(coordPoly[,1],decreasing=FALSE),]
         locOrder<-data.frame(CoordOrder)
         DisGeoEucl<-dist(locOrder, method = "euclidean", diag = TRUE, upper = TRUE)

         # compute corresponding matrix of genetic distances
         listcoord = (1:6)[expr==1]
         Genet = NULL ## fichier vide pour collage des données

         for(h in listcoord){
             tmp <- matG2[(matG2[, 1]==xcoord[h])and(matG2 [, 2]== ycoord[h]), ]
             Genet = rbind(Genet,tmp)
                        }
    GenetOrder<- Genet[order(Genet[,1],decreasing=FALSE),]
    GenetOrderSanscoord <- GenetOrder[,-c(1,2)]
    MatdistGenet<- vegdist(GenetOrderSanscoord, method="bray", binary=FALSE, diag=FALSE, upper=TRUE, na.rm = TRUE)
    MatdistGenet <- as.dist(MatdistGenet)

    # Compute matrix landscape "resistance" distances based on raster
    spatiallocX <-  locOrder[,1] ## extraction des colonnes pour repasser en spatial
    spatiallocY <-  locOrder[,2]
    SpaLoc <- SpatialPoints(cbind(spatiallocX, spatiallocY))
    Resdis<- commuteDistance(t, SpaLoc)
    Resdis<-as.dist(Resdis, diag = TRUE, upper=TRUE)

    # simple mantels test between genetic and landscape "resistance" distances
    MantelpRes <- mantel.rtest(MatdistGenet, Resdis, nrepet = 99)
    results <- rbind (results, cbind (xcoord [i], ycoord [i],Radius, MantelpRes[2], MantelpRes[4]))
     }
     resultsfinal <- rbind(resultsfinal,results)
 }

# display result file with for each individual: x and y coordinates, radius of sampling area, mantel output and associated p-value
Resultsfinal