MCP home range overlap - r

Question

I am trying to estimate home range overlap among a batch of animals that I have estimated home ranges for through both the MCP and kernelUD methods in the adehabitatHR package. I realize there is a function for calculating overlap with the kernel method, but can anyone tell me how best to calculate overlap for MCP?
I gather the rgeos package has a gIntersection function that I have begun to play around with.
Has anyone come up with a relatively straightforward method/code - whether in adehabitatHR, regeos, or elsewhere?

Answer 1

Maybe something like this helps:
library(adehabitatHR)
library(rgeos)
## Use some simulated data
ani1 <- SpatialPoints(matrix(rnorm(200, mean=2), ncol=2))
ani2 <- SpatialPoints(matrix(rnorm(200, mean=1), ncol=2))
## Function to calculate overlaps
gOverlap <- function(hr1, hr2, ...) {
a <- gIntersection(hr1, hr2, ...)
if (is.null(a)) {
return(0)
}
gArea(a, byid=TRUE) / gArea(hr1, byid=TRUE)
}
## Calcualte homeranges
hr1 <- mcp(ani1)
hr2 <- mcp(ani2)
## Calculate HR overlap
gOverlap(hr1, hr2)

Answer 2

Running overlap for single pairs at once can be very inefficient when you have many animals or many collaring sessions. I'm sure there is a cleaner way to do this (and I welcome suggestions on refinements!), but this worked for me:
#Relevant libraries (I think they're all here!)
library(adehabitatHR)
library(geosphere)
library(rgeos)
library(nlme)
#Define an overlap function for a single dyad
dyad.overlap <- function(cp,dyads.df){
p1<-subset(cp, cp#data$id==dyads.df[1])
p2<-subset(cp, cp#data$id==dyads.df[2])
Overlap<- ifelse(is.null(gIntersection(p1,p2)), 0, gArea(gIntersection(p1,p2)) / gArea(p1)) #puts 0 if no overlap between the dyad
return(Overlap)
}
#Define a function for overlap of all animals from a given time period
mcp.overlaps <- function(period.df){
period.df$Animal<-factor(period.df$Animal) #remove any ids not relevant to that period
count <- length(unique(period.df$Animal)) #identify number of individuals for the period
anim <- unique(period.df$Animal) #identify names of individuals/periods
xy<-SpatialPointsDataFrame(period.df[c("Easting","Southing")], data=data.frame(id=period.df$Animal)) #create SPDF
proj4string(xy)<-CRS("+init=epsg:32750") #define projection
cp <- adehabitatHR::mcp(xy, percent=95) #create Spatial Polygons Data Frame of the 95% MCPs
writeOGR(obj=cp, dsn="tempdir", layer=paste(period.df$file_folder[i],"95 Percent MCPs"), driver="ESRI Shapefile") #export shp of 95% MCPs for period, if desired
dyads<-(combn(anim, 2, simplify=T)) #all possible dyad combinations
dyads.df<-data.frame(A1=dyads[1,1:count],A2=dyads[2,1:count]) #creates a data frame of the pairs
dyads.df$Overlap<-apply(dyads.df, 1, dyad.overlap, cp=cp)
dyads.df$Period<-period.df$file_folder[1]
return(dyads.df)
}
#Now run the overlap for each time period included in the data frame
All.Overlaps<-do.call(rbind.data.frame, gapply(df, groups=df$file_folder, FUN=mcp.overlaps))
Of course, if you only have a single time period (e.g., a year of data and looking at annual home ranges), all you have to do is:
All.Overlaps <- mcp.overlaps(df)
This assumes that all data is in a single data frame (df), with the variables defined as following (adjust names to suit your data):
file_folder: collaring period/session of interest
Animal: unique identifier for an individual
Easting/Southing: your lat long coordinates (be sure to adjust the projection as needed)

Answer 3

I have modified a little the Johannes' answer, maybe it helps a little more:
library(adehabitatHR)
library(rgeos)
data(puechabonsp)
rel <- puechabonsp$relocs
cp <- mcp(rel[,1])
## Set Up a matrix which will store the results
mat <- matrix(NA,4,4,dimnames=list(c("1","2","3","4"),
c("1","2","3","4")))
## Set Up the loop
gOverlap <- function(hr, number, matrix){
for(i in c(1:number)){
for(j in c(1:number)){
a <- gIntersection(hr[i,], hr[j,])
if (is.null(a)){
matrix[i, j] <- 0
} else{matrix[i, j] <- gArea(a)}
}
}
return(matrix)
}
##Test the function
gOverlap(cp, 4, mat)
##Plotting to corroborate
plot(cp)
The output is a matrix with every intersection, and to obtain the total overlap you only have to sum row or columns without the self-intersection. Finally, you can add or delete individuals just modifying the initial matrix.
All the best,