ensemble.blockCV: Spatially or environmentally separated folds for...

In BiodiversityR: Package for Community Ecology and Suitability Analysis

Spatially or environmentally separated folds for cross-validation via blockCV::spatialBlock or blockCV::envBlock

Description

The functions internally calls blockCV::spatialBlock and blockCV::envBlock. Syntax is very similar to that of BiodiversityR::ensemble.calibrate.weights.

Usage

ensemble.spatialBlock(x = NULL, p = NULL, 
    a = NULL, an = 1000, EPSG=NULL,
    excludep = FALSE, target.groups = FALSE, k = 4,
    factors = NULL,
    theRange = NULL, return.object = FALSE, ...)

ensemble.envBlock(x = NULL, p = NULL, 
    a = NULL, an = 1000, EPSG=NULL,
    excludep = FALSE, target.groups = FALSE, k = 4,
    factors = NULL,
    return.object = FALSE, ...)

Arguments

x	RasterStack object (stack) containing all layers that correspond to explanatory variables
p	presence points used for calibrating the suitability models, typically available in 2-column (lon, lat) dataframe; see also prepareData and extract
a	background points used for calibrating the suitability models, typically available in 2-column (lon, lat) dataframe; see also prepareData and extract
an	number of background points for calibration to be selected with randomPoints in case argument a is missing
EPSG	EPSG number (see https://spatialreference.org/) to be assigned internally to the coordinate reference system of the locations via st_crs. Although the function internally first assigns the coordinate reference from the RasterStack x via crs, this method fails in some situations as in the example shown below. In such cases, manually assigning the EPSG could resolve this problem.
excludep	parameter that indicates (if TRUE) that presence points will be excluded from the background points; see also randomPoints
target.groups	Parameter that indicates (if TRUE) that the provided background points (argument a) represent presence points from a target group sensu Phillips et al. 2009 (these are species that are all collected or observed using the same methods or equipment). Setting the parameter to TRUE results in selecting the centres of cells of the target groups as background points, while avoiding to select the same cells twice. Via argument excludep, it is possible to filter out cells with presence observations (argument p).
k	Integer value. The number of desired folds for cross-validation. The default is k = 4. The interpretation of the argument is exactly the same as in ensemble.calibrate.models and kfold.
factors	vector that indicates which variables are factors; see also prepareData
theRange	Numeric value of the specified range by which blocks are created and training/testing data are separated. This distance should be in metres. See also spatialBlock.
return.object	If TRUE, then also return ('block.object') the complete result of spatialBlock or envBlock. In addtion (if TRUE), return the species data ('speciesData') that was created for blockCV. To visualize these results, see below or from foldExplorer.
...	Other arguments to pass to spatialBlock or envBlock, such numLimit (The minimum number of points in each fold for training-presence, training-absence, testing-presence and testing-absence) and iteration (The number of attempts to create folds that fulfil the numLimit requirement).

Details

The functions internally call spatialBlock or envBlock.

The result of the function includes a list (k) with following elements. This list can be directly imported into ensemble.calibrate.weights, but only elements groupp and groupa will be used.

- p : Presence locations, created by ensemble.calibrate.models where points with missing data were excluded and possibly points were added for missing factor levels

- a : Background locations, created by ensemble.calibrate.models where points with missing data were excluded and possibly points were added for missing factor levels

- groupp : k-fold identities for the presence locations

- groupa : k-fold identities for the background locations

Optionally the function also returns elements block.object and speciesData. These can be used to visualize data with foldExplorer.

Value

The function returns a list with the following elements:.

k	A list with data on folds that can be directly used by ensemble.calibrate.weights.
block.object	the results of spatialBlock or envBlock
speciesData	a SpatialPointsDataFrame with species data

Author(s)

Roeland Kindt (World Agroforestry Centre)

References

Roberts et al., 2017. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography. 40: 913-929.

Examples

## Not run: 

library(blockCV)
library(sf)

# get predictor variables
library(dismo)
predictor.files <- list.files(path=paste(system.file(package="dismo"), '/ex', sep=''),
    pattern='grd', full.names=TRUE)
predictors <- stack(predictor.files)
# subset based on Variance Inflation Factors
predictors <- subset(predictors, subset=c("bio5", "bio6", 
    "bio16", "bio17"))
predictors
predictors@title <- "base"

# presence points
presence_file <- paste(system.file(package="dismo"), '/ex/bradypus.csv', sep='')
pres <- read.table(presence_file, header=TRUE, sep=',')[, -1]

# choose background points
background <- randomPoints(predictors, n=1000, p=pres, excludep=T, extf=1.00)
background <- data.frame(background)
colnames(background)=c('lon', 'lat')

# spatial blocking with square blocks of 1000 km and minimum 20 points in each categor
# fails if EPSG is not assigned
block.data <- ensemble.spatialBlock(x=predictors, p=pres, a=background, 
    EPSG=NULL,
    showBlocks=F, theRange=1000000, k=4, numLimit=20, iteration=1000, return.object=T)

block.data <- ensemble.spatialBlock(x=predictors, p=pres, a=background, 
    EPSG=4326,
    showBlocks=F, theRange=1000000, k=4, numLimit=20, iteration=1000, return.object=T)

# explore the results
foldExplorer(blocks=block.data$block.object, rasterLayer=predictors, 
    speciesData=block.data$speciesData)

# apply in calibration of ensemble weights
# make sure that folds apply to subset of points
p.spatial <- block.data$k$p 
a.spatial <- block.data$k$a
k.spatial <- block.data$k

ensemble.w1 <- ensemble.calibrate.weights(x=predictors, 
    p=p.spatial, a=a.spatial, k=k.spatial, 
    species.name="Bradypus",
    SINK=FALSE, PROBIT=TRUE,
    MAXENT=0, MAXNET=1, MAXLIKE=1, GBM=1, GBMSTEP=0, RF=0, CF=1,
    GLM=1, GLMSTEP=0, GAM=1, GAMSTEP=0, MGCV=0, MGCVFIX=0, 
    EARTH=0, RPART=0, NNET=1, FDA=0, SVM=0, SVME=0, GLMNET=0,
    BIOCLIM.O=1, BIOCLIM=1, DOMAIN=0, MAHAL=0, MAHAL01=0,
    ENSEMBLE.tune=TRUE,
    ENSEMBLE.best=0, ENSEMBLE.exponent=c(1, 2, 3),
    ENSEMBLE.min=0.7,
    Yweights="BIOMOD", 
    formulae.defaults=TRUE)

# confirm that correct folds were used
all.equal(ensemble.w1$groupp, block.data$k$groupp)
all.equal(ensemble.w1$groupa, block.data$k$groupa)

# environmental blocking with minimum 5 points in each category
block.data2 <- ensemble.envBlock(x=predictors, p=pres, a=background, 
    factors="biome",
    k=4, numLimit=5, return.object=T)

# explore the results
foldExplorer(blocks=block.data2$block.object, rasterLayer=predictors, 
    speciesData=block.data2$speciesData)

# apply in calibration of ensemble weights
# make sure that folds apply to subset of points
p.env <- block.data2$k$p 
a.env <- block.data2$k$a
k.env <- block.data2$k

ensemble.w2 <- ensemble.calibrate.weights(x=predictors, 
    p=p.env, a=a.env, k=k.env, 
    species.name="Bradypus",
    SINK=FALSE, PROBIT=TRUE,
    MAXENT=0, MAXNET=1, MAXLIKE=1, GBM=1, GBMSTEP=0, RF=0, CF=1,
    GLM=1, GLMSTEP=0, GAM=1, GAMSTEP=0, MGCV=0, MGCVFIX=0, 
    EARTH=0, RPART=0, NNET=1, FDA=0, SVM=0, SVME=0, GLMNET=0,
    BIOCLIM.O=1, BIOCLIM=1, DOMAIN=0, MAHAL=0, MAHAL01=0,
    ENSEMBLE.tune=TRUE,
    ENSEMBLE.best=0, ENSEMBLE.exponent=c(1, 2, 3),
    ENSEMBLE.min=0.7,
    factors="biome",
    Yweights="BIOMOD", 
    formulae.defaults=TRUE)

# confirm that correct folds were used
all.equal(ensemble.w2$groupp, block.data2$k$groupp)
all.equal(ensemble.w2$groupa, block.data2$k$groupa)


## End(Not run)

BiodiversityR documentation built on June 22, 2024, 11:57 a.m.