An R package to improve the usability of datasets with predictors for species distribution modelling (SDM).
Installation:
install.packages("sdmpredictors")
# or for the latest dev version
devtools::install_github("lifewatch/sdmpredictors")Example 1: Create SDM for Dictyota diemensis in Australia Note that this requires the ZOON, ggplot2, cowplot and marinespeed packages to be installed.
library(sdmpredictors)
library(zoon)
# Inspect the available datasets and layers
datasets <- list_datasets(terrestrial = FALSE, marine = TRUE)
View(datasets)
layers <- list_layers(datasets)
View(layers)
# Load equal area rasters and crop with the extent of the Baltic Sea
layercodes <- c("MS_biogeo05_dist_shore_5m", "MS_bathy_5m",
"BO_sstrange", "BO_sstmean", "BO_salinity")
env <- load_layers(layercodes, equalarea = TRUE)
australia <- raster::crop(env, extent(106e5,154e5, -52e5, -13e5))
plot(australia)
# Compare statistics between the original and the Australian bathymetry
View(rbind(layer_stats("MS_bathy_5m"),
calculate_statistics("Bathymetry Australia",
raster(australia, layer = 2))))
# Compare correlations between predictors, globally and for Australia
prettynames <- list(BO_salinity="Salinity", BO_sstmean="SST (mean)",
BO_sstrange="SST (range)", MS_bathy_5m="Bathymetry",
MS_biogeo05_dist_shore_5m = "Shore distance")
p1 <- plot_corr(layers_correlation(layercodes), prettynames)
australian_correlations <- pearson_correlation_matrix(australia)
p2 <- plot_correlation(australian_correlations, prettynames)
cowplot::plot_grid(p1, p2, labels=c("A", "B"), ncol = 2, nrow = 1)
print(correlation_groups(australian_correlations))
# Fetch occurrences and prepare for ZOON
occ <- marinespeed::get_occurrences("Dictyota diemensis")
points <- SpatialPoints(occ[,c("longitude", "latitude")],
lonlatproj)
points <- spTransform(points, equalareaproj)
occfile <- tempfile(fileext = ".csv")
write.csv(cbind(coordinates(points), value=1), occfile)
# Create SDM with ZOON
workflow(
occurrence = LocalOccurrenceData(
occfile, occurrenceType="presence",
columns = c("longitude", "latitude", "value")),
covariate = LocalRaster(stack(australia)),
process = OneHundredBackground(seed = 42),
model = LogisticRegression,
output = PrintMap)
# Layer citations
print(layer_citations(layercodes))Example 2: view marine datasets, layers and load a few of them by name
library(sdmpredictors)
# exploring the marine datasets
datasets <- list_datasets(terrestrial = FALSE, marine = TRUE)
View(datasets)
browseURL(datasets$url[1])
# exploring the layers
layers <- list_layers(datasets)
View(layers)
# download specific layers to the current directory
rasters <- load_layers(c("BO_calcite", "BO_chlomean", "MS_bathy_5m"), datadir = ".")Example 3: looking up statistics and correlations for marine annual layers:
datasets <- list_datasets(terrestrial = FALSE, marine = TRUE)
layers <- list_layers(datasets)
# filter out monthly layers
layers <- layers[is.na(layers$month),]
stats <- layer_stats(layers)
View(stats)
correlations <- layers_correlation(layers)
View(correlations)
# create groups of layers where no layers in one group
# have a correlation > 0.7 with a layer from another group
groups <- correlation_groups(correlations, max_correlation=0.7)
# inspect groups
# heatmap plot for larger groups (if gplots library is installed)
for(group in groups) {
group_correlation <- as.matrix(correlations[group, group, drop=FALSE])
if(require(gplots) && length(group) > 4){
heatmap.2(abs(group_correlation)
,main = "Correlation"
,col = "rainbow"
,notecol="black" # change font color of cell labels to black
,density.info="none" # turns off density plot inside color legend
,trace="none" # turns off trace lines inside the heat map
,margins = c(12,9) # widens margins around plot
)
} else {
print(group_correlation)
}
}See the quickstart vignette for more information
vignette("quickstart", package = "sdmpredictors")