stack_modelling: Build an SSDM that assembles multiple algorithms and species.

In hugocalcad/LigthSSDM: Optimized SSDM Package for large dataset

Description

This is a function to build an SSDM that assembles multiple algorithm and species. The function takes as inputs an occurrence data frame made of presence/absence or presence-only records and a raster object for data extraction and projection. The function returns an S4 Stacked.SDM class object containing the local species richness map, the between-algorithm variance map, and all evaluation tables coming with (model evaluation, algorithm evaluation, algorithm correlation matrix and variable importance), and a list of ensemble SDMs for each species (see ensemble_modelling).

Usage

stack_modelling(algorithms, Occurrences, Env, Xcol = "Longitude",
  Ycol = "Latitude", Pcol = NULL, Spcol = "SpeciesID", rep = 10,
  name = NULL, save = FALSE, path = getwd(), PA = NULL,
  cv = "holdout", cv.param = c(0.7, 1), thresh = 1001,
  axes.metric = "Pearson", uncertainty = TRUE, tmp = FALSE,
  ensemble.metric = c("AUC"), ensemble.thresh = c(0.75),
  weight = TRUE, method = "pSSDM", metric = "SES", rep.B = 1000,
  range = NULL, endemism = c("WEI", "Binary"), verbose = TRUE,
  GUI = FALSE, cores = 1, folder_tmp = NULL, ...)

Arguments

algorithms	character. Choice of the algorithm(s) to be run (see details below).
Occurrences	data frame. Occurrence table (can be processed first by load_occ).
Env	raster object. Raster object of environmental variables (can be processed first by load_var).
Xcol	character. Name of the column in the occurrence table containing Latitude or X coordinates.
Ycol	character. Name of the column in the occurrence table containing Longitude or Y coordinates.
Pcol	character. Name of the column in the occurrence table specifying whether a line is a presence or an absence. A value of 1 is presence and value of 0 is absence. If NULL presence-only dataset is assumed.
Spcol	character. Name of the column containing species names or IDs.
rep	integer. Number of repetitions for each algorithm.
name	character. Optional name given to the final Ensemble.SDM produced.
save	logical. If set to true, the SSDM is automatically saved.
path	character. If save is true, the path to the directory in which the ensemble SDM will be saved.
PA	list(nb, strat) defining the pseudo-absence selection strategy used in case of presence-only dataset. If PA is NULL, recommended PA selection strategy is used depending on the algorithm (see details below).
cv	character. Method of cross-validation used to evaluate the ensemble SDM (see details below).
cv.param	numeric. Parameters associated with the method of cross-validation used to evaluate the ensemble SDM (see details below).
thresh	numeric. A single integer value representing the number of equal interval threshold values between 0 and 1 (see optim.thresh).
axes.metric	Metric used to evaluate variable relative importance (see details below).
uncertainty	logical. If set to true, generates an uncertainty map and an algorithm correlation matrix.
tmp	logical. If set to true, the habitat suitability map of each algorithms is saved in a temporary file to release memory. But beware: if you close R, temporary files will be deleted To avoid any loss you can save your SSDM with save.model. Depending on number, resolution and extent of models, temporary files can take a lot of disk space. Temporary files are written in R environment temporary folder.
ensemble.metric	character. Metric(s) used to select the best SDMs that will be included in the ensemble SDM (see details below).
ensemble.thresh	numeric. Threshold(s) associated with the metric(s) used to compute the selection.
weight	logical. Choose whether or not you want the SDMs to be weighted using the selection metric or, alternatively, the mean of the selection metrics.
method	character. Define the method used to create the local species richness map (see details below).
metric	character. Metric used to compute the binary map threshold (see details below.)
rep.B	integer. If the method used to create the local species richness is the random bernoulli (Bernoulli), rep.B parameter defines the number of repetitions used to create binary maps for each species.
range	integer. Set a value of range restriction (in pixels) around presences occurrences on habitat suitability maps (all further points will have a null probability, see Crisp et al (2011) in references). If NULL, no range restriction will be applied.
endemism	character. Define the method used to create an endemism map (see details below).
verbose	logical. If set to true, allows the function to print text in the console.
GUI	logical. Don't take that argument into account (parameter for the user interface).
cores	integer. Specify the number of CPU cores used to do the computing. You can use detectCores) to automatically used all the available CPU cores.
folder_tmp	character. carpete temporal para el ENV.
...	additional parameters for the algorithm modelling function (see details below).

Details

algorithms

'all' allows you to call directly all available algorithms. Currently, available algorithms include Generalized linear model (GLM), Generalized additive model (GAM), Multivariate adaptive regression splines (MARS), Generalized boosted regressions model (GBM), Classification tree analysis (CTA), Random forest (RF), Maximum entropy (MAXENT), Artificial neural network (ANN), and Support vector machines (SVM). Each algorithm has its own parameters settable with the ... (see each algorithm section below to set their parameters).

"PA"

list with two values: nb number of pseudo-absences selected, and strat strategy used to select pseudo-absences: either random selection or disk selection. We set default recommendation from Barbet-Massin et al. (2012) (see reference).

cv

Cross-validation method used to split the occurrence dataset used for evaluation: holdout data are partitioned into a training set and an evaluation set using a fraction (cv.param[1]) and the operation can be repeated (cv.param[2]) times, k-fold data are partitioned into k (cv.param[1]) folds being k-1 times in the training set and once the evaluation set and the operation can be repeated (cv.param[2]) times, LOO (Leave One Out) each point is successively taken as evaluation data.

metric

Choice of the metric used to compute the binary map threshold and the confusion matrix (by default SES as recommended by Liu et al. (2005), see reference below): Kappa maximizes the Kappa, CCR maximizes the proportion of correctly predicted observations, TSS (True Skill Statistic) maximizes the sum of sensitivity and specificity, SES uses the sensitivity-specificity equality, LW uses the lowest occurrence prediction probability, ROC minimizes the distance between the ROC plot (receiving operating curve) and the upper left corner (1,1).

axes.metric

Choice of the metric used to evaluate the variable relative importance (difference between a full model and one with each variable successively omitted): Pearson (computes a simple Pearson's correlation r between predictions of the full model and the one without a variable, and returns the score 1-r: the highest the value, the more influence the variable has on the model), AUC, Kappa, sensitivity, specificity, and prop.correct (proportion of correctly predicted occurrences).

ensemble.metric

Ensemble metric(s) used to select SDMs: AUC, Kappa, sensitivity, specificity, and prop.correct (proportion of correctly predicted occurrences).

method

Choice of the method used to compute the local species richness map (see Calabrese et al. (2014) and D'Amen et al (2015) for more informations, see reference below): pSSDM sum probabilities of habitat suitability maps, Bernoulli drawing repeatedly from a Bernoulli distribution, bSSDM sum the binary map obtained with the thresholding (depending on the metric, see metric parameter), MaximumLikelihood adjust species richness of the model by linear regression, PRR.MEM model richness with a macroecological model (MEM) and adjust each ESDM binary map by ranking habitat suitability and keeping as much as predicted richness of the MEM, PRR.pSSDM model richness with a pSSDM and adjust each ESDM binary map by ranking habitat suitability and keeping as much as predicted richness of the pSSDM

endemism

Choice of the method used to compute the endemism map (see Crisp et al. (2001) for more information, see reference below): NULL No endemism map, WEI (Weighted Endemism Index) Endemism map built by counting all species in each cell and weighting each by the inverse of its range, CWEI (Corrected Weighted Endemism Index) Endemism map built by dividing the weighted endemism index by the total count of species in the cell. First string of the character is the method either WEI or CWEI, and in those cases second string of the vector is used to precise range calculation, whether the total number of occurrences 'NbOcc' whether the surface of the binary map species distribution 'Binary'.

...

See algorithm in detail section

Value

an S4 Stacked.SDM class object viewable with the plot.model function.

Generalized linear model (GLM)

Uses the glm function from the package 'stats', you can set the following parameters (see glm for more details):

test

character. Test used to evaluate the SDM, default 'AIC'.

epsilon

numeric. Positive convergence tolerance eps ; the iterations converge when |dev - dev_old|/(|dev| + 0.1) < eps. By default, set to 10e-08.

maxit

numeric. Integer giving the maximal number of IWLS (Iterative Weighted Last Squares) iterations, default 500.

Generalized additive model (GAM)

Uses the gam function from the package 'mgcv', you can set the following parameters (see gam for more details):

test

character. Test used to evaluate the model, default 'AIC'.

epsilon

numeric. This is used for judging conversion of the GLM IRLS (Iteratively Reweighted Least Squares) loop, default 10e-08.

maxit

numeric. Maximum number of IRLS iterations to perform, default 500.

Multivariate adaptive regression splines (MARS)

Uses the earth function from the package 'earth', you can set the following parameters (see earth for more details):

degree

integer. Maximum degree of interaction (Friedman's mi) ; 1 meaning build an additive model (i.e., no interaction terms). By default, set to 2.

Generalized boosted regressions model (GBM)

Uses the gbm function from the package 'gbm,' you can set the following parameters (see gbm for more details):

trees

integer. The total number of trees to fit. This is equivalent to the number of iterations and the number of basis functions in the additive expansion. By default, set to 2500.

final.leave

integer. minimum number of observations in the trees terminal nodes. Note that this is the actual number of observations not the total weight. By default, set to 1.

algocv

integer. Number of cross-validations, default 3.

thresh.shrink

integer. Number of cross-validation folds to perform. If cv.folds>1 then gbm, in addition to the usual fit, will perform a cross-validation. By default, set to 1e-03.

Classification tree analysis (CTA)

Uses the rpart function from the package 'rpart', you can set the following parameters (see rpart for more details):

final.leave

integer. The minimum number of observations in any terminal node, default 1.

algocv

integer. Number of cross-validations, default 3.

Random Forest (RF)

Uses the randomForest function from the package 'randomForest', you can set the following parameters (see randomForest for more details):

trees

integer. Number of trees to grow. This should not be set to a too small number, to ensure that every input row gets predicted at least a few times. By default, set to 2500.

final.leave

integer. Minimum size of terminal nodes. Setting this number larger causes smaller trees to be grown (and thus take less time). By default, set to 1.

Maximum Entropy (MAXENT)

Uses the maxent function from the package 'dismo'. Make sure that you have correctly installed the maxent.jar file in the folder ~\R\library\version\dismo\java available at https://www.cs.princeton.edu/~schapire/maxent/ (see maxent for more details).

Artificial Neural Network (ANN)

Uses the nnet function from the package 'nnet', you can set the following parameters (see nnet for more details):

maxit

integer. Maximum number of iterations, default 500.

Support vector machines (SVM)

Uses the svm function from the package 'e1071', you can set the following parameters (see svm for more details):

epsilon

float. Epsilon parameter in the insensitive loss function, default 1e-08.

algocv

integer. If an integer value k>0 is specified, a k-fold cross-validation on the training data is performed to assess the quality of the model: the accuracy rate for classification and the Mean Squared Error for regression. By default, set to 3.

Warning

Depending on the raster object resolution the process can be more or less time and memory consuming.

References

M. D'Amen, A. Dubuis, R. F. Fernandes, J. Pottier, L. Pelissier, & A Guisan (2015) "Using species richness and functional traits prediction to constrain assemblage predicitions from stacked species distribution models" Journal of Biogeography 42(7):1255-1266 http://doc.rero.ch/record/235561/files/pel_usr.pdf

M. Barbet-Massin, F. Jiguet, C. H. Albert, & W. Thuiller (2012) "Selecting pseudo-absences for species distribution models: how, where and how many?" Methods Ecology and Evolution 3:327-338 http://onlinelibrary.wiley.com/doi/10.1111/j.2041-210X.2011.00172.x/full

J.M. Calabrese, G. Certain, C. Kraan, & C.F. Dormann (2014) "Stacking species distribution models and adjusting bias by linking them to macroecological models." Global Ecology and Biogeography 23:99-112 http://portal.uni-freiburg.de/biometrie/mitarbeiter/dormann/calabrese2013globalecolbiogeogr.pdf

M. D. Crisp, S. Laffan, H. P. Linder & A. Monro (2001) "Endemism in the Australian flora" Journal of Biogeography 28:183-198 http://biology-assets.anu.edu.au/hosted_sites/Crisp/pdfs/Crisp2001_endemism.pdf

C. Liu, P. M. Berry, T. P. Dawson, R. & G. Pearson (2005) "Selecting thresholds of occurrence in the prediction of species distributions." Ecography 28:85-393 http://www.researchgate.net/publication/230246974_Selecting_Thresholds_of_Occurrence_in_the_Prediction_of_Species_Distributions

See Also

modelling to build simple SDMs.

Examples

## Not run: 
# Loading data
data(Env)
data(Occurrences)

# SSDM building
SSDM <- stack_modelling(c('CTA', 'SVM'), Occurrences, Env, rep = 1,
                       Xcol = 'LONGITUDE', Ycol = 'LATITUDE',
                       Spcol = 'SPECIES')

# Results plotting
plot(SSDM)

## End(Not run)

hugocalcad/LigthSSDM documentation built on June 22, 2019, 12:43 a.m.