R/seedDispersalLANDIS.R
LANDISDisp.RdSimulate seed dispersal using user defined function. This is a "receiving pixel" focused
dispersal approach.
It is the "potentially receiving" cell that looks around itself for potential seed sources.
If it finds a single seed source, that passes the probability function described by the
dispersalFn.
If this passes a comparison to a uniform random draw, then the receiving cell is deemed to have
a "successful" dispersal for that species.
This function can therefore only be used for a relatively specific situation
where there is a yes/no returned for each potential receiving cell, i.e., not abundance.
This function is also not cumulative, i.e,. there is no higher abundance of seeds received if
a receiving cell has lots of seed sources around it vs. a single seed source.
The difference will come with a higher probability of successfully receiving a "seed".
LANDISDisp(
dtSrc,
dtRcv,
pixelGroupMap,
speciesTable,
dispersalFn = Ward,
b = 0.01,
k = 0.95,
plot.it = FALSE,
successionTimestep,
verbose = getOption("LandR.verbose", TRUE),
...
)data.table
data.table
A RasterLayer with pixel values equal to a pixel group
number that corresponds exactly to pixelGroup column in cohortData.
A species traits table, with at least the following columns:
speciesCode an character representation of species;
Area a character of geographical area within a species range for which trait values are relevant;
longevity species longevity in years, sexualmature age in years at sexual maturity;
shadetolerance numeric value between 1-5 of relative shade tolerance (with respect to other species);
seeddistance_eff the "effective" seed dispersal distance;
seeddistance_max a numeric with the maximum seed dispersal distance;
mortalityshape and growthcurve: growth curve shape parameters.
Other columns (e.g. fire-related traits) can also be included depending on LandR modules in use.
Please see the LANDIS-II Biomass Succession Extension v3.2.1 manual (Scheller and Miranda 2015)
for further detail.
An expression that can take a "dis" argument. See details. Default is "Ward" (temporarily unused, as it is hard coded inside Rcpp function)
LANDIS Ward seed dispersal calibration coefficient (set to 0.01 in LANDIS)
LANDIS Ward seed dispersal the probability that seed will disperse within the effective distance (e.g., 0.95)
Deprecated. If TRUE, then plot the raster at every interaction,
so one can watch the LANDISDisp event grow.
The time between successive seed dispersal events. In LANDIS-II, this is called "Succession Timestep".
Controls message output. Defaults to getOption("LandR.verbose")
Additional parameters. Currently none.
A numeric vector of raster pixel indices, in the same resolution and extent as
seedSrc raster.
dispersalFn (temporarily unused as code is converted to Rcpp -- the
default dispersalFn is hard coded within the spiralSeedDispersal
function that uses C++) must be an expression that returns a probability
distribution. Because it is a dispersal kernel, it must be a probability
distribution. The expression that can take an argument named "dis" (without
quotes) as this will be calculated internally and represents the distance
from the initial (receiving) pixel and all active pixels within that cluster
of active pixels. SpaDES includes the Ward() kernel as
defined in the LANDIS-II documentation.
if (require("googledrive")) {
seed <- sample(1e6, 1)
set.seed(seed)
library(data.table)
# keep this here for interactive testing with a larger raster
rasterTemplate <- LandR:::rasterRead(terra::ext(0, 2500, 0, 2500), res = 100)
# make a pixelGroupMap
pgs <- 4 # make even just because of approach below requires even
pixelGroupMap <- SpaDES.tools::randomPolygons(rasterTemplate, numTypes = pgs)
pixelGroupMap[1:100] <- NA # emulate a mask at the start
# Make a receive pixels table -- need pixelGroup and species
nSpecies <- 3
maxNSpeciesPerPixel <- min(5, nSpecies)
rcvSpByPG <- lapply(seq_len(pgs / 2), function(pg) {
data.table(speciesCode = sample(nSpecies, size = sample(maxNSpeciesPerPixel, 1)))
})
seedReceive <- rbindlist(rcvSpByPG, idcol = "pixelGroup")
# Make a source pixels table -- need pixelGroup and species
srcSpByPG <- lapply(seq_len(pgs / 2), function(pg) {
data.table(speciesCode = sample(nSpecies, size = sample(maxNSpeciesPerPixel, 1)))
})
seedSource <- rbindlist(srcSpByPG, idcol = "pixelGroup")
# make source pixels not same pixelGroups as receive
seedSource[, pixelGroup := pixelGroup + pgs / 2]
# Get a species table -- if using in Canada, can use this
speciesTable <- getSpeciesTable(dPath = tempdir())
speciesTable <- speciesTable[Area == "BSW"]
speciesTable[, speciesCode := as.factor(LandisCode)]
speciesTable[, seeddistance_eff := SeedEffDist]
speciesTable[, seeddistance_max := SeedMaxDist]
speciesTable <- speciesTable
speciesTable <- data.table(speciesTable)[, speciesCode := seq_along(LandisCode)]
seedReceiveFull <- speciesTable[seedReceive, on = "speciesCode"]
output <- LANDISDisp(
dtRcv = seedReceiveFull, plot.it = interactive(),
dtSrc = seedSource,
speciesTable = speciesTable,
pixelGroupMap,
verbose = TRUE,
successionTimestep = 10
)
# Summarize
output[, .N, by = speciesCode]
## Plot the maps
if (interactive()) {
library(quickPlot)
clearPlot()
spMap <- list()
spMap$pixelGroupMap <- pixelGroupMap
for (sppp in unique(output$speciesCode)) {
spppChar <- paste0("Sp_", sppp)
spMap[[spppChar]] <- LandR:::rasterRead(pixelGroupMap)
ss <- unique(seedSource[speciesCode == sppp], on = c("pixelGroup", "speciesCode"))
spMap[[spppChar]][pixelGroupMap[] %in% ss$pixelGroup] <- 1
receivable <- LandR:::rasterRead(pixelGroupMap)
srf <- unique(seedReceiveFull[speciesCode == sppp], on = c("pixelGroup", "speciesCode"))
receivable[pixelGroupMap[] %in% srf$pixelGroup] <- 1
forest <- which(!is.na(pixelGroupMap[]))
src <- which(!is.na(spMap[[spppChar]][]))
recvable <- which(!is.na(receivable[]))
rcvd <- output[speciesCode == sppp]$pixelIndex
spMap[[spppChar]][forest] <- 0
spMap[[spppChar]][recvable] <- 2
spMap[[spppChar]][src] <- 1
spMap[[spppChar]][rcvd] <- 3
spMap[[spppChar]][intersect(src, rcvd)] <- 4
levels(spMap[[spppChar]]) <- data.frame(ID = 0:4,
type = c("OtherForest", "Source", "Didn't receive",
"Received", "Src&Rcvd"))
}
Plot(spMap, cols = "Set2")
# A summary
rr <- apply(rast(spMap)[[-1]][] + 1, 2, tabulate)
rownames(rr) <- raster::levels(spMap[[2]])[[1]][,"type"][1:NROW(rr)]
# next line only works if there are some places that are both source and potential to receive
# rr <- rbind(rr, propSrcRcved = round(rr[5,]/ (rr[5,]+rr[2,]), 2))
}
}
#> Loading required package: googledrive
#> No cachePath supplied and getOption('reproducible.cachePath') is
#> inside a temporary directory;
#> this will not persist across R
#> sessions.
#> Running `prepInputs`
#> Running `preProcess`
#> Preparing: speciesTraits.csv
#> ...downloading...
#> Downloading
#> https://raw.githubusercontent.com/dcyr/LANDIS-II_IA_generalUseFiles/master/speciesTraits.csv
#> ...
#>
|
| | 0%
|
|======================================================================| 100%
#> Hardlinked version of file(s) created:
#> ... no copy/copies made.
#> alsoExtract is unspecified; assuming that all files must be
#> extracted
#> Appending checksums to CHECKSUMS.txt. If you see this message
#> repeatedly, you can specify targetFile (and optionally alsoExtract)
#> so it knows what to look for.
#> Running `process` (i.e., loading file into R)
#> targetFile located at /tmp/RtmpU0Xt8U/speciesTraits.csv
#> Loading object into R
#> Saved! Cache file: 9325be38bac7e8b3.rds; fn:
#> data.table::fread
#> Saved! Cache file: ff83cfcdb8675c83.rds; fn: prepInputs