plotKML (version 0.2-4)

SpatialMaxEntOutput-class: A class for outputs of analysis produced using the dismo package (MaxEnt)

Description

A class containing input and output data produced by running the maxent (Maximum Entropy) species distribution modeling algorithm. Object of this class can be directly visualized in Google Earth by using the plotKML-method.

References

• Hijmans, R.J, Elith, J., (2012)http://cran.r-project.org/web/packages/dismo/vignettes/sdm.pdf{Species distribution modeling with R}. CRAN, Vignette for the dismo package, 72 p.
• dismo package (http://CRAN.R-project.org/package=dismo)
• rgbif package (http://CRAN.R-project.org/package=rgbif)

See Also

dismo::maxent, rgbif::taxonsearch, GSIF::MaxEnt

Examples

# load occurences:
data(bigfoot)
aea.prj <- "+proj=aea +lat_1=29.5 +lat_2=45.5 +lat_0=23 +lon_0=-96 
+x_0=0 +y_0=0 +ellps=GRS80 +datum=NAD83 +units=m +no_defs"
coordinates(bigfoot) <- ~Lon+Lat
proj4string(bigfoot) <- CRS("+proj=latlon +datum=WGS84")
bigfoot.aea <- spTransform(bigfoot, CRS(aea.prj))
# Load the covariates:
data(USAWgrids)
gridded(USAWgrids) <- ~s1+s2
proj4string(USAWgrids) <- CRS(aea.prj)
# USA borders:
library(maps)
data(R_pal)
state.m = map('state', plot=FALSE, fill=TRUE)
IDs <- sapply(strsplit(state.m$names, ":"), function(x) x[1])
library(maptools)
states <- as(map2SpatialPolygons(state.m, IDs=IDs), "SpatialLines")
proj4string(states) <- CRS("+proj=latlon +datum=NAD83")
states.aea <- spTransform(states, CRS(aea.prj))
# subset to area of interest:
sel.ov <- !is.na(overlay(USAWgrids, bigfoot.aea))
# run species distribution modelling:
sel.grids <- c("globedem","nlights03","sdroads","gcarb","twi","globcov")
library(dismo)
# MaxEnt requires the maxent.jar file to run:
jar <- paste(system.file(package="dismo"), "/java/maxent.jar", sep='')
if (file.exists(jar)) {
  predictors <- stack(USAWgrids[sel.grids])
  # prepare the occurrence-only records:
  xy <- data.frame(bigfoot.aea@coords[sel.ov,])
  # fit a MaxEnt model (can take few minutes!):
  me <- maxent(predictors, xy, factors='globcov') # takes 1-2 mins;
  # predict distribution of Big Foot:
  bigfoot.pr <- predict(me, predictors, progress='window')
  par(mfrow=c(1,2), mar=c(.5,.5,3.5,0.5), oma=c(0,0,0,0))
  image(bigfoot.pr, col=R_pal[["blue_grey_red"]], main='MaxEnt predictions', 
     axes = FALSE, xlab="", ylab="")
  rx <- range(bigfoot.pr@data@values, na.rm=TRUE)
  rx <- rev(as.character(round(c(rx[1],NA,mean(rx),NA,rx[2]), 2))) 
  legend("topleft", rx, fill=rev(R_pal[["blue_grey_red"]][c(1,5,10,15,20)]), 
     horiz=FALSE, cex=.6, bty="n")
  lines(states.aea)
  points(bigfoot.aea[sel.ov,], pch="+", cex=.7)
  # run cross-validation
  fold <- kfold(xy, k=5)
  # randomly take 20  xy.test <- xy[fold == 1,]
  bg <- randomPoints(predictors, 1000)  
  ev <- evaluate(me, p=xy.test, a=bg, x=predictors)
  # this allows estimation of the threshold probability:
  threshold <- ev@t[which.max(ev@TPR + ev@TNR)]
  # which allows us to derive the potential homerange for this species:
  image(bigfoot.pr > threshold, col=c("wheat","green"), main='presence/absence', 
      axes = FALSE, xlab="", ylab="")
  lines(states.aea)
  dev.off()
  # prepare data for plotKML:
  hr <- calc(bigfoot.pr, fun=function(x){ifelse(x>threshold, 1, NA)})
  bigfoot.hr <- rasterToPolygons(hr, dissolve=TRUE) # takes 3-4 mins!
  occ <- SpatialPoints(bigfoot.aea[sel.ov,]); proj4string(occ) = CRS(aea.prj)
  # create an object of type "SpatialMaxEntOutput":
  bigfoot.smo <- new("SpatialMaxEntOutput", sciname = "Sasquatch", occurrences = occ, 
     TimeSpan.begin = min(bigfoot@data[sel.ov,"DATE"]), 
     TimeSpan.end = max(bigfoot@data[sel.ov,"DATE"]), 
     maxent = me, sp.domain = bigfoot.hr, predicted = bigfoot.pr)
  # str(bigfoot.smo, max.level=2)
  # plot the whole SDM project in Google Earth:
  data(R_pal)
  icon = "http://plotkml.r-forge.r-project.org/bigfoot.png"
  plotKML(bigfoot.smo, colour_scale = R_pal[["bpy_colors"]], shape = icon)
}