,
Arnaud Barras [aut] ,
Veronika Braunisch [aut] ,
Conservation Biology - University of Bern [fnd]| Title: | Species Distribution Model Selection |
|---|---|
| Description: | User-friendly framework that enables the training and the evaluation of species distribution models (SDMs). The package implements functions for data driven variable selection and model tuning and includes numerous utilities to display the results. All the functions used to select variables or to tune model hyperparameters have an interactive real-time chart displayed in the 'RStudio' viewer pane during their execution. |
| Authors: | Sergio Vignali [aut, cre] ,
Arnaud Barras [aut] ,
Veronika Braunisch [aut] ,
Conservation Biology - University of Bern [fnd] |
| Maintainer: | Sergio Vignali <[email protected]> |
| License: | GPL-3 |
| Version: | 1.3.1 |
| Built: | 2024-11-12 06:13:37 UTC |
| Source: | https://github.com/consbiol-unibern/sdmtune |
The function add the presence locations to the background. This is equivalent
to the Maxent argument addsamplestobackground=true.
addSamplesToBg(x, all = FALSE)addSamplesToBg(x, all = FALSE)
x |
SWD object. |
all |
logical. If |
An object of class SWD.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") data # Add presence locations with values not included in the backgrounds to the # background locations new_data <- addSamplesToBg(data) new_data # Add all the presence locations to the background locations, even if they # have values already included in the backgrounds new_data <- addSamplesToBg(data, all = TRUE) new_data# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") data # Add presence locations with values not included in the backgrounds to the # background locations new_data <- addSamplesToBg(data) new_data # Add all the presence locations to the background locations, even if they # have values already included in the backgrounds new_data <- addSamplesToBg(data, all = TRUE) new_data
Compute the Akaike Information Criterion corrected for small samples size (Warren and Seifert, 2011).
aicc(model, env)aicc(model, env)
model |
SDMmodel object. |
env |
rast containing the environmental variables. |
The function is available only for Maxent and Maxnet methods.
The computed AICc
Sergio Vignali
Warren D.L., Seifert S.N., (2011). Ecological niche modeling in Maxent: the importance of model complexity and the performance of model selection criteria. Ecological Applications, 21(2), 335–342.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Compute the AICc aicc(model, env = predictors)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Compute the AICc aicc(model, env = predictors)
This Class represents an Artificial Neural Network model object and hosts all the information related to the model.
## S4 method for signature 'ANN' show(object)## S4 method for signature 'ANN' show(object)
object |
ANN object |
See nnet for the meaning of the slots.
sizeinteger. Number of the units in the hidden layer.
decaynumeric. Weight decay.
rangnumeric. Initial random weights.
maxitinteger. Maximum number of iterations.
modelnnet. The randomForest model object.
Sergio Vignali
Compute the AUC using the Man-Whitney U Test formula.
auc(model, test = NULL)auc(model, test = NULL)
model |
An SDMmodel or SDMmodelCV object. |
test |
SWD object when |
For SDMmodelCV objects, the function computes the mean
of the training AUC values of the k-folds. If test = TRUE it computes the
mean of the testing AUC values for the k-folds. If test is an
SWD object, it computes the mean AUC values for the provided
testing dataset.
The value of the AUC.
Sergio Vignali
Mason, S. J. and Graham, N. E. (2002), Areas beneath the relative operating characteristics (ROC) and relative operating levels (ROL) curves: Statistical significance and interpretation. Q.J.R. Meteorol. Soc., 128: 2145-2166.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute the training AUC auc(model) # Compute the testing AUC auc(model, test = test) # Same example but using cross validation instead of training and testing # datasets folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Compute the training AUC auc(model) # Compute the testing AUC auc(model, test = TRUE) # Compute the AUC for the held apart testing dataset auc(model, test = test)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute the training AUC auc(model) # Compute the testing AUC auc(model, test = test) # Same example but using cross validation instead of training and testing # datasets folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Compute the training AUC auc(model) # Compute the testing AUC auc(model, test = TRUE) # Compute the AUC for the held apart testing dataset auc(model, test = test)
This Class represents a Boosted Regression Tree model objects and hosts all the information related to the model.
## S4 method for signature 'BRT' show(object)## S4 method for signature 'BRT' show(object)
object |
BRT object |
See gbm for the meaning of the slots.
distributioncharacter. Name of the used distribution.
n.treesinteger. Maximum number of grown trees.
interaction.depthinteger. Maximum depth of each tree.
shrinkagenumeric. The shrinkage parameter.
bag.fractionnumeric. Random fraction of data used in the tree expansion.
modelgbm. The Boosted Regression Tree model object.
Sergio Vignali
The function checks if Maxent is correctly installed.
checkMaxentInstallation(verbose = TRUE)checkMaxentInstallation(verbose = TRUE)
verbose |
logical, if |
In order to have Maxent correctly configured is necessary that:
Java is installed;
the package "rJava" is installed;
the file "maxent.jar" is in the correct folder.
TRUE if Maxent is correctly installed, FALSE otherwise.
Sergio Vignali
checkMaxentInstallation()checkMaxentInstallation()
This function combines cross-validation models by retraining a new model with all presence and absence/background locations and the same hyperparameters.
combineCV(model)combineCV(model)
model |
SDMmodelCV object. |
This is an utility function to retrain a model with all data after, for example, the hyperparameters tuning (gridSearch, randomSearch or optimizeModel) to avoid manual setting of the hyperparameters in the train function.
An SDMmodel object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, folds = folds) # Define the hyperparameters to test h <- list(reg = 1:2, fc = c("lqp", "lqph")) # Run the function using the AUC as metric output <- gridSearch(model, hypers = h, metric = "auc") output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ] # Combine cross validation models for output with highest test AUC idx <- which.max(output@results$test_AUC) combined_model <- combineCV(output@models[[idx]]) combined_model# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, folds = folds) # Define the hyperparameters to test h <- list(reg = 1:2, fc = c("lqp", "lqph")) # Run the function using the AUC as metric output <- gridSearch(model, hypers = h, metric = "auc") output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ] # Combine cross validation models for output with highest test AUC idx <- which.max(output@results$test_AUC) combined_model <- combineCV(output@models[[idx]]) combined_model
Computes Confusion Matrixes for threshold values varying from 0 to 1.
confMatrix(model, test = NULL, th = NULL, type = NULL)confMatrix(model, test = NULL, th = NULL, type = NULL)
model |
SDMmodel object. |
test |
SWD testing locations, if not provided it uses the training dataset. |
th |
numeric vector. If provided it computes the evaluation at the given
thresholds. Default is |
type |
character. The output type used for "Maxent" and "Maxnet" methods, possible values are "cloglog" and "logistic". |
For models trained with the Maxent method the argument type can be:
"raw", "logistic" and "cloglog".
For models trained with the Maxnet method the argument type can be:
"link", "exponential", "logistic" and "cloglog", see maxnet
for more details.
The Confusion Matrix for all the used thresholds.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Get the confusion matrix for thresholds ranging from 0 to 1 cm <- confMatrix(model, type = "cloglog") head(cm) tail(cm) # Get the confusion matrix for a specific threshold confMatrix(model, type = "logistic", th = 0.6)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Get the confusion matrix for thresholds ranging from 0 to 1 cm <- confMatrix(model, type = "cloglog") head(cm) tail(cm) # Get the confusion matrix for a specific threshold confMatrix(model, type = "logistic", th = 0.6)
Utility that prints the name of correlated variables and the relative correlation coefficient value.
corVar( bg, method = "spearman", cor_th = NULL, order = TRUE, remove_diagonal = TRUE )corVar( bg, method = "spearman", cor_th = NULL, order = TRUE, remove_diagonal = TRUE )
bg |
SWD object with the locations used to compute the correlation between environmental variables. |
method |
character. The method used to compute the correlation matrix. |
cor_th |
numeric. If provided it prints only the variables whose correlation coefficient is higher or lower than the given threshold. |
order |
logical. If |
remove_diagonal |
logical. If |
A data.frame with the variables and their correlation.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations bg_coords <- terra::spatSample(predictors, size = 10000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) # Create SWD object bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") # Get the correlation among all the environmental variables corVar(bg, method = "spearman") # Get the environmental variables that have a correlation greater or equal to # the given threshold corVar(bg, method = "pearson", cor_th = 0.8)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations bg_coords <- terra::spatSample(predictors, size = 10000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) # Create SWD object bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") # Get the correlation among all the environmental variables corVar(bg, method = "spearman") # Get the environmental variables that have a correlation greater or equal to # the given threshold corVar(bg, method = "pearson", cor_th = 0.8)
Run the Jackknife test for variable importance removing one variable at time.
doJk( model, metric, variables = NULL, test = NULL, with_only = TRUE, env = NULL, return_models = FALSE, progress = TRUE )doJk( model, metric, variables = NULL, test = NULL, with_only = TRUE, env = NULL, return_models = FALSE, progress = TRUE )
model |
SDMmodel or SDMmodelCV object. |
metric |
character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc". |
variables |
vector. Variables used for the test, if not provided it takes all the variables used to train the model. |
test |
SWD. If provided it reports the result also for the testing dataset. Not used for aicc and SDMmodelCV. |
with_only |
logical. If |
env |
rast containing the environmental variables, used only with "aicc". |
return_models |
logical. If |
progress |
logical If |
A data frame with the test results. If return_model = TRUE it
returns a list containing the test results together with the models.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric AUC doJk(model, metric = "auc", variables = c("bio1", "bio12"), test = test) # The same without testing dataset doJk(model, metric = "auc", variables = c("bio1", "bio12")) # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric TSS but without running the test for one single # variable doJk(model, metric = "tss", variables = c("bio1", "bio12"), test = test, with_only = FALSE) # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric AICc but without running the test for one single # variable doJk(model, metric = "aicc", variables = c("bio1", "bio12"), with_only = FALSE, env = predictors) # Execute the Jackknife test for all the environmental variables using the # metric AUC and returning all the trained models jk <- doJk(model, metric = "auc", test = test, return_models = TRUE) jk$results jk$models_without jk$models_withonly# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric AUC doJk(model, metric = "auc", variables = c("bio1", "bio12"), test = test) # The same without testing dataset doJk(model, metric = "auc", variables = c("bio1", "bio12")) # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric TSS but without running the test for one single # variable doJk(model, metric = "tss", variables = c("bio1", "bio12"), test = test, with_only = FALSE) # Execute the Jackknife test only for the environmental variables "bio1" and # "bio12", using the metric AICc but without running the test for one single # variable doJk(model, metric = "aicc", variables = c("bio1", "bio12"), with_only = FALSE, env = predictors) # Execute the Jackknife test for all the environmental variables using the # metric AUC and returning all the trained models jk <- doJk(model, metric = "auc", test = test, return_models = TRUE) jk$results jk$models_without jk$models_withonly
Returns the name of all function arguments that can be tuned for a given model.
getTunableArgs(model)getTunableArgs(model)
model |
SDMmodel or SDMmodelCV object. |
character vector.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxnet model and get tunable hyperparameters model <- train(method = "Maxnet", data = data, fc = "l") getTunableArgs(model)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxnet model and get tunable hyperparameters model <- train(method = "Maxnet", data = data, fc = "l") getTunableArgs(model)
Given a set of possible hyperparameter values, the function trains models with all the possible combinations of hyperparameters.
gridSearch( model, hypers, metric, test = NULL, env = NULL, save_models = TRUE, interactive = TRUE, progress = TRUE )gridSearch( model, hypers, metric, test = NULL, env = NULL, save_models = TRUE, interactive = TRUE, progress = TRUE )
model |
SDMmodel or SDMmodelCV object. |
hypers |
named list containing the values of the hyperparameters that should be tuned, see details. |
metric |
character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc". |
test |
SWD object. Testing dataset used to evaluate the model, not used with aicc and SDMmodelCV objects. |
env |
rast containing the environmental variables, used only with "aicc". |
save_models |
logical. If |
interactive |
logical. If |
progress |
logical. If |
To know which hyperparameters can be tuned you can use the output
of the function getTunableArgs. Hyperparameters not included in the
hypers argument take the value that they have in the passed model.
An interactive chart showing in real-time the steps performed by the algorithm is displayed in the Viewer pane.
SDMtune object.
Sergio Vignali
randomSearch and optimizeModel.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = 1:2, fc = c("lqp", "lqph")) # Run the function using the AUC as metric output <- gridSearch(model, hypers = h, metric = "auc", test = test) output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ] # Run the function using the AICc as metric and without saving the trained # models, helpful when numerous hyperparameters are tested to avoid memory # problems output <- gridSearch(model, hypers = h, metric = "aicc", env = predictors, save_models = FALSE) output@results# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = 1:2, fc = c("lqp", "lqph")) # Run the function using the AUC as metric output <- gridSearch(model, hypers = h, metric = "auc", test = test) output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ] # Run the function using the AICc as metric and without saving the trained # models, helpful when numerous hyperparameters are tested to avoid memory # problems output <- gridSearch(model, hypers = h, metric = "aicc", env = predictors, save_models = FALSE) output@results
This Class represents a MaxEnt model objects and hosts all the information related to the model.
## S4 method for signature 'Maxent' show(object)## S4 method for signature 'Maxent' show(object)
object |
Maxent object |
resultsmatrix. The result that usually MaxEnt provide as a csv file.
regnumeric. The value of the regularization multiplier used to train the model.
fccharacter. The feature class combination used to train the model.
iternumeric. The number of iterations used to train the model.
extra_argscharacter. Extra arguments used to run MaxEnt.
lambdasvector. The lambdas parameters of the model.
coeffdata.frame. The lambda coefficients of the model.
formulaformula. The formula used to make prediction.
lpnnumeric. Linear Predictor Normalizer.
dnnumeric. Density Normalizer.
entropynumeric. The entropy value.
min_maxdata.frame. The minimum and maximum values of the continuous variables, used for clamping.
Sergio Vignali
Returns the value of the thresholds generated by the MaxEnt software.
maxentTh(model)maxentTh(model)
model |
SDMmodel object trained using the "Maxent" method. |
data.frame with the thresholds.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l") maxentTh(model)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l") maxentTh(model)
Shows the percent contribution and permutation importance of the environmental variables used to train the model.
maxentVarImp(model)maxentVarImp(model)
model |
SDMmodel or SDMmodelCV object trained using the "Maxent" method. |
When an SDMmodelCV object is passed to the function, the output is the average of the variable importance of each model trained during the cross validation.
A data frame with the variable importance.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model # The next line checks if Maxent is correctly configured but you don't need # to run it in your script model <- train(method = "Maxent", data = data, fc = "l") maxentVarImp(model)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model # The next line checks if Maxent is correctly configured but you don't need # to run it in your script model <- train(method = "Maxent", data = data, fc = "l") maxentVarImp(model)
This Class represents a Maxnet model objects and hosts all the information related to the model.
## S4 method for signature 'Maxnet' show(object)## S4 method for signature 'Maxnet' show(object)
object |
Maxnet object |
regnumeric. The value of the regularization multiplier used to train the model.
fccharacter. The feature class combination used to train the model.
modelmaxnet. The maxnet model object.
Sergio Vignali
Merge two SWD objects.
mergeSWD(swd1, swd2, only_presence = FALSE)mergeSWD(swd1, swd2, only_presence = FALSE)
swd1 |
SWD object. |
swd2 |
SWD object. |
only_presence |
logical If |
In case the two SWD objects have different columns, only the common columns are used in the merged object.
The SWD object is created in a way that the presence locations are always before than the absence/background locations.
The merged SWD object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split only presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Merge the training and the testing datasets together merged <- mergeSWD(train, test, only_presence = TRUE) # Split presence and absence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2) train <- datasets[[1]] test <- datasets[[2]] # Merge the training and the testing datasets together merged <- mergeSWD(train, test)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split only presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Merge the training and the testing datasets together merged <- mergeSWD(train, test, only_presence = TRUE) # Split presence and absence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2) train <- datasets[[1]] test <- datasets[[2]] # Merge the training and the testing datasets together merged <- mergeSWD(train, test)
Make a report that shows the main results.
modelReport( model, folder, test = NULL, type = NULL, response_curves = FALSE, only_presence = FALSE, jk = FALSE, env = NULL, clamp = TRUE, permut = 10, verbose = TRUE )modelReport( model, folder, test = NULL, type = NULL, response_curves = FALSE, only_presence = FALSE, jk = FALSE, env = NULL, clamp = TRUE, permut = 10, verbose = TRUE )
model |
SDMmodel object. |
folder |
character. The name of the folder in which to save the output. The folder is created in the working directory. |
test |
SWD object with the test locations. |
type |
character. The output type used for "Maxent" and "Maxnet" methods, possible values are "cloglog" and "logistic". |
response_curves |
logical, if |
only_presence |
logical, if |
jk |
logical, if |
env |
rast. If provided it computes and adds a prediction map to the output. |
clamp |
logical for clumping during prediction, used for response curves and for the prediction map. |
permut |
integer. Number of permutations. |
verbose |
logical, if |
The function produces a report similar to the one created by MaxEnt software. See terra documentation to see how to pass factors.
Sergio Vignali
# If you run the following examples with the function example(), # you may want to set the argument ask like following: example("modelReport", # ask = FALSE) # Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Create the report ## Not run: modelReport(model, type = "cloglog", folder = "my_folder", test = test, response_curves = TRUE, only_presence = TRUE, jk = TRUE, env = predictors, permut = 2) ## End(Not run)# If you run the following examples with the function example(), # you may want to set the argument ask like following: example("modelReport", # ask = FALSE) # Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Create the report ## Not run: modelReport(model, type = "cloglog", folder = "my_folder", test = test, response_curves = TRUE, only_presence = TRUE, jk = TRUE, env = predictors, permut = 2) ## End(Not run)
The function uses a Genetic Algorithm implementation to optimize the model hyperparameter configuration according to the chosen metric.
optimizeModel( model, hypers, metric, test = NULL, pop = 20, gen = 5, env = NULL, keep_best = 0.4, keep_random = 0.2, mutation_chance = 0.4, interactive = TRUE, progress = TRUE, seed = NULL )optimizeModel( model, hypers, metric, test = NULL, pop = 20, gen = 5, env = NULL, keep_best = 0.4, keep_random = 0.2, mutation_chance = 0.4, interactive = TRUE, progress = TRUE, seed = NULL )
model |
SDMmodel or SDMmodelCV object. |
hypers |
named list containing the values of the hyperparameters that should be tuned, see details. |
metric |
character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc". |
test |
SWD object. Testing dataset used to evaluate the model, not used with aicc and SDMmodelCV objects. |
pop |
numeric. Size of the population. |
gen |
numeric. Number of generations. |
env |
rast containing the environmental variables, used only with "aicc". |
keep_best |
numeric. Percentage of the best models in the population to be retained during each iteration, expressed as decimal number. |
keep_random |
numeric. Probability of retaining the excluded models during each iteration, expressed as decimal number. |
mutation_chance |
numeric. Probability of mutation of the child models, expressed as decimal number. |
interactive |
logical. If |
progress |
logical. If |
seed |
numeric. The value used to set the seed to have consistent results. |
To know which hyperparameters can be tuned you can use the output
of the function getTunableArgs. Hyperparameters not included in the
hypers argument take the value that they have in the passed model.
An interactive chart showing in real-time the steps performed by the algorithm is displayed in the Viewer pane.
Part of the code is inspired by this post.
SDMtune object.
Sergio Vignali
gridSearch and randomSearch.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, val = 0.2, test = 0.2, only_presence = TRUE, seed = 61516) train <- datasets[[1]] val <- datasets[[2]] # Train a model model <- train("Maxnet", data = train) # Define the hyperparameters to test h <- list(reg = seq(0.2, 5, 0.2), fc = c("l", "lq", "lh", "lp", "lqp", "lqph")) # Run the function using as metric the AUC ## Not run: output <- optimizeModel(model, hypers = h, metric = "auc", test = val, pop = 15, gen = 2, seed = 798) output@results output@models output@models[[1]] # Best model ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, val = 0.2, test = 0.2, only_presence = TRUE, seed = 61516) train <- datasets[[1]] val <- datasets[[2]] # Train a model model <- train("Maxnet", data = train) # Define the hyperparameters to test h <- list(reg = seq(0.2, 5, 0.2), fc = c("l", "lq", "lh", "lp", "lqp", "lqph")) # Run the function using as metric the AUC ## Not run: output <- optimizeModel(model, hypers = h, metric = "auc", test = val, pop = 15, gen = 2, seed = 798) output@results output@models output@models[[1]] # Best model ## End(Not run)
Plot a correlation matrix heat map with the value of the correlation coefficients according with the given method. If cor_th is passed then it prints only the coefficients that are higher or lower than the given threshold.
plotCor(bg, method = "spearman", cor_th = NULL, text_size = 3)plotCor(bg, method = "spearman", cor_th = NULL, text_size = 3)
bg |
SWD object used to compute the correlation matrix. |
method |
character. The method used to compute the correlation matrix. |
cor_th |
numeric. If provided it prints only the coefficients that are higher or lower than the given threshold. |
text_size |
numeric, used to change the size of the text. |
A ggplot object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) # Create SWD object bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") # Plot heat map plotCor(bg, method = "spearman") # Plot heat map showing only values higher than given threshold and change # text size plotCor(bg, method = "spearman", cor_th = 0.8, text_size = 4)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) # Create SWD object bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") # Plot heat map plotCor(bg, method = "spearman") # Plot heat map showing only values higher than given threshold and change # text size plotCor(bg, method = "spearman", cor_th = 0.8, text_size = 4)
Plot the Jackknife Test for variable importance.
plotJk(jk, type = c("train", "test"), ref = NULL)plotJk(jk, type = c("train", "test"), ref = NULL)
jk |
data.frame with the output of the doJk function. |
type |
character, "train" or "test" to plot the result of the test on the train or testing dataset. |
ref |
numeric. The value of the chosen metric for the model trained using all the variables. If provided it plots a vertical line showing the reference value. |
A ggplot object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Execute the Jackknife test for all the environmental variables using the # metric AUC jk <- doJk(model, metric = "auc", test = test) # Plot Jackknife test result for training plotJk(jk, type = "train", ref = auc(model)) #' # Plot Jackknife test result for testing plotJk(jk, type = "test", ref = auc(model, test = test))# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "lq") # Execute the Jackknife test for all the environmental variables using the # metric AUC jk <- doJk(model, metric = "auc", test = test) # Plot Jackknife test result for training plotJk(jk, type = "train", ref = auc(model)) #' # Plot Jackknife test result for testing plotJk(jk, type = "test", ref = auc(model, test = test))
Plot a presence absence map using the given threshold.
plotPA( map, th, colors = NULL, hr = FALSE, filename = "", overwrite = FALSE, wopt = list(), ... )plotPA( map, th, colors = NULL, hr = FALSE, filename = "", overwrite = FALSE, wopt = list(), ... )
map |
rast object with the prediction. |
th |
numeric. The threshold used to convert the output in a presence/absence map. |
colors |
vector. Colors to be used, default is |
hr |
logical. If |
filename |
character. If provided the raster map is saved in a file. It must include the extension. |
overwrite |
logical. If |
wopt |
list. Writing options passed to writeRaster. |
... |
Unused arguments. |
A ggplot object.
Sergio Vignali
map <- terra::rast(matrix(runif(400, 0, 1), nrow = 20, ncol = 20)) plotPA(map, th = 0.8) # Custom colors plotPA(map, th = 0.5, colors = c("#d8b365", "#018571")) ## Not run: # Save the file. The following command will save the map in the working # directory. Note that `filename` must include the extension. plotPA(map, th = 0.7, filename = "my_map.tif") ## End(Not run)map <- terra::rast(matrix(runif(400, 0, 1), nrow = 20, ncol = 20)) plotPA(map, th = 0.8) # Custom colors plotPA(map, th = 0.5, colors = c("#d8b365", "#018571")) ## Not run: # Save the file. The following command will save the map in the working # directory. Note that `filename` must include the extension. plotPA(map, th = 0.7, filename = "my_map.tif") ## End(Not run)
Plot Prediction output.
plotPred(map, lt = "", colorramp = NULL, hr = FALSE)plotPred(map, lt = "", colorramp = NULL, hr = FALSE)
map |
rast object with the prediction. |
lt |
character. Legend title. |
colorramp |
vector. A custom colour ramp given as a vector of colours
(see example), default is |
hr |
logical. If |
A ggplot object.
Sergio Vignali
map <- terra::rast(matrix(runif(400, 0, 1), nrow = 20, ncol= 20)) plotPred(map, lt = "Habitat suitability \ncloglog") # Custom colors plotPred(map, lt = "Habitat suitability", colorramp = c("#2c7bb6", "#ffffbf", "#d7191c"))map <- terra::rast(matrix(runif(400, 0, 1), nrow = 20, ncol= 20)) plotPred(map, lt = "Habitat suitability \ncloglog") # Custom colors plotPred(map, lt = "Habitat suitability", colorramp = c("#2c7bb6", "#ffffbf", "#d7191c"))
Plot the Response Curve of the given environmental variable.
plotResponse( model, var, type = NULL, only_presence = FALSE, marginal = FALSE, fun = mean, rug = FALSE, color = "red" )plotResponse( model, var, type = NULL, only_presence = FALSE, marginal = FALSE, fun = mean, rug = FALSE, color = "red" )
model |
SDMmodel or SDMmodelCV object. |
var |
character. Name of the variable to be plotted. |
type |
character. The output type used for "Maxent" and "Maxnet" methods, possible values are "cloglog" and "logistic". |
only_presence |
logical. If |
marginal |
logical. If |
fun |
function used to compute the level of the other variables for marginal curves. |
rug |
logical. If |
color |
The color of the curve, default is "red". |
Note that fun is not a character argument, you must use mean and not
"mean".
If you want to modify the plot, first you have to assign the output of the function to a variable, and then you have two options:
Modify the ggplot object by editing the theme or adding additional
elements
Get the data with ggplot2::ggplot_build() and then build your own
plot (see examples)
A ggplot object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "lq") # Plot cloglog response curve for a continuous environmental variable (bio1) plotResponse(model, var = "bio1", type = "cloglog") # Plot marginal cloglog response curve for a continuous environmental # variable (bio1) plotResponse(model, var = "bio1", type = "cloglog", marginal = TRUE) # Plot logistic response curve for a continuous environmental variable # (bio12) adding the rugs and giving a custom color plotResponse(model, var = "bio12", type = "logistic", rug = TRUE, color = "blue") # Plot response curve for a categorical environmental variable (biome) giving # a custom color plotResponse(model, var = "biome", type = "logistic", color = "green") # Modify plot # Change y axes limits my_plot <- plotResponse(model, var = "bio1", type = "cloglog") my_plot + ggplot2::scale_y_continuous(limits = c(0, 1)) # Get the data and create your own plot: df <- ggplot2::ggplot_build(my_plot)$data[[1]] plot(df$x, df$y, type = "l", lwd = 3, col = "blue", xlab = "bio1", ylab = "cloglog output") # Train a model with cross validation folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "lq", folds = folds) # Plot cloglog response curve for a continuous environmental variable (bio17) plotResponse(model, var = "bio1", type = "cloglog") # Plot logistic response curve for a categorical environmental variable # (biome) giving a custom color plotResponse(model, var = "biome", type = "logistic", color = "green")# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "lq") # Plot cloglog response curve for a continuous environmental variable (bio1) plotResponse(model, var = "bio1", type = "cloglog") # Plot marginal cloglog response curve for a continuous environmental # variable (bio1) plotResponse(model, var = "bio1", type = "cloglog", marginal = TRUE) # Plot logistic response curve for a continuous environmental variable # (bio12) adding the rugs and giving a custom color plotResponse(model, var = "bio12", type = "logistic", rug = TRUE, color = "blue") # Plot response curve for a categorical environmental variable (biome) giving # a custom color plotResponse(model, var = "biome", type = "logistic", color = "green") # Modify plot # Change y axes limits my_plot <- plotResponse(model, var = "bio1", type = "cloglog") my_plot + ggplot2::scale_y_continuous(limits = c(0, 1)) # Get the data and create your own plot: df <- ggplot2::ggplot_build(my_plot)$data[[1]] plot(df$x, df$y, type = "l", lwd = 3, col = "blue", xlab = "bio1", ylab = "cloglog output") # Train a model with cross validation folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "lq", folds = folds) # Plot cloglog response curve for a continuous environmental variable (bio17) plotResponse(model, var = "bio1", type = "cloglog") # Plot logistic response curve for a categorical environmental variable # (biome) giving a custom color plotResponse(model, var = "biome", type = "logistic", color = "green")
Plot the ROC curve of the given model and print the AUC value.
plotROC(model, test = NULL)plotROC(model, test = NULL)
model |
SDMmodel object. |
test |
SWD object. The testing dataset. |
A ggplot object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Plot the training ROC curve plotROC(model) # Plot the training and testing ROC curves plotROC(model, test = test)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Plot the training ROC curve plotROC(model) # Plot the training and testing ROC curves plotROC(model, test = test)
Plot the variable importance as a bar plot.
plotVarImp(df, color = "grey")plotVarImp(df, color = "grey")
df |
data.frame. A data.frame containing the the name of the variables as first column and the value of the variable importance as second column. |
color |
character. The colour of the bar plot. |
A ggplot object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Compute variable importance vi <- varImp(model, permut = 1) # Plot variable importance plotVarImp(vi) # Plot variable importance with custom color plotVarImp(vi, color = "red")# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a model model <- train(method = "Maxnet", data = data, fc = "l") # Compute variable importance vi <- varImp(model, permut = 1) # Plot variable importance plotVarImp(vi) # Plot variable importance with custom color plotVarImp(vi, color = "red")
Predict the output for a new dataset from a trained ANN model.
## S4 method for signature 'ANN' predict(object, data, type, clamp)## S4 method for signature 'ANN' predict(object, data, type, clamp)
object |
ANN object. |
data |
data.frame with the data for the prediction. |
type |
Not used. |
clamp |
Not used. |
Used by the predict,SDMmodel-method, not exported.
A vector with the predicted values.
Sergio Vignali
Predict the output for a new dataset from a trained BRT model.
## S4 method for signature 'BRT' predict(object, data, type, clamp)## S4 method for signature 'BRT' predict(object, data, type, clamp)
object |
BRT object. |
data |
data.frame with the data for the prediction. |
type |
Not used. |
clamp |
Not used. |
Used by the predict,SDMmodel-method, not exported.
The function uses the number of tree defined to train the model and the "response" type output.
A vector with the predicted values.
Sergio Vignali
Predict the output for a new dataset from a trained Maxent model.
## S4 method for signature 'Maxent' predict(object, data, type = c("cloglog", "logistic", "raw"), clamp = TRUE)## S4 method for signature 'Maxent' predict(object, data, type = c("cloglog", "logistic", "raw"), clamp = TRUE)
object |
Maxent object. |
data |
data.frame with the data for the prediction. |
type |
character. MaxEnt output type, possible values are "cloglog", "logistic" and "raw". |
clamp |
logical for clumping during prediction. |
Used by the predict,SDMmodel-method, not exported.
The function performs the prediction in R without calling the MaxEnt Java software. This results in a faster computation for large datasets and might result in a slightly different output compared to the Java software.
A vector with the prediction
Sergio Vignali
Wilson P.D., (2009). Guidelines for computing MaxEnt model output values from a lambdas file.
Predict the output for a new dataset from a trained Maxnet model.
## S4 method for signature 'Maxnet' predict( object, data, type = c("link", "exponential", "cloglog", "logistic"), clamp = TRUE )## S4 method for signature 'Maxnet' predict( object, data, type = c("link", "exponential", "cloglog", "logistic"), clamp = TRUE )
object |
Maxnet object. |
data |
data.frame with the data for the prediction. |
type |
character. Maxnet output type, possible values are "link", "exponential", "cloglog" and "logistic". |
clamp |
logical for clumping during prediction. |
Used by the predict,SDMmodel-method, not exported.
A vector with the predicted values.
Sergio Vignali
Predict the output for a new dataset from a trained RF model.
## S4 method for signature 'RF' predict(object, data, type, clamp)## S4 method for signature 'RF' predict(object, data, type, clamp)
object |
RF object. |
data |
data.frame with the data for the prediction. |
type |
Not used. |
clamp |
Not used. |
Used by the predict,SDMmodel-method, not exported.
A vector with the predicted probabilities of class 1.
Sergio Vignali
Predict the output for a new dataset given a trained SDMmodel model.
## S4 method for signature 'SDMmodel' predict( object, data, type = NULL, clamp = TRUE, filename = "", overwrite = FALSE, wopt = list(), extent = NULL, ... )## S4 method for signature 'SDMmodel' predict( object, data, type = NULL, clamp = TRUE, filename = "", overwrite = FALSE, wopt = list(), extent = NULL, ... )
object |
SDMmodel object. |
data |
|
type |
character. Output type, see details, used only for Maxent and Maxnet methods. |
clamp |
logical for clumping during prediction, used only for Maxent and Maxnet methods. |
filename |
character. If provided the raster map is saved in a file. It must include the extension. |
overwrite |
logical. If |
wopt |
list. Writing options passed to writeRaster. |
extent |
ext object, if provided it restricts the prediction to the given extent. |
... |
Additional arguments to pass to the predict function. |
filename, and extent are arguments used only when the prediction is run for a rast object.
For models trained with the Maxent method the argument type can be:
"raw", "logistic" and "cloglog". The function performs the prediction in
R without calling the MaxEnt Java software. This results in a faster
computation for large datasets and might result in a slightly different
output compared to the Java software.
For models trained with the Maxnet method the argument type can be:
"link", "exponential", "logistic" and "cloglog", see maxnet
for more details.
For models trained with the ANN method the function uses the "raw" output type.
For models trained with the RF method the output is the probability of class 1.
For models trained with the BRT method the function uses the number of trees defined to train the model and the "response" output type.
A vector with the prediction or a rast object if data is a raster rast.
Sergio Vignali
Wilson P.D., (2009). Guidelines for computing MaxEnt model output values from a lambdas file.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Make cloglog prediction for the test dataset predict(model, data = test, type = "cloglog") # Make logistic prediction for the whole study area predict(model, data = predictors, type = "logistic") ## Not run: # Make logistic prediction for the whole study area and save it in a file. # Note that the filename must include the extension. The function saves the # file in your working directory predict(model, data = predictors, type = "logistic", filename = "my_map.tif") ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Make cloglog prediction for the test dataset predict(model, data = test, type = "cloglog") # Make logistic prediction for the whole study area predict(model, data = predictors, type = "logistic") ## Not run: # Make logistic prediction for the whole study area and save it in a file. # Note that the filename must include the extension. The function saves the # file in your working directory predict(model, data = predictors, type = "logistic", filename = "my_map.tif") ## End(Not run)
Predict the output for a new dataset given a trained SDMmodelCV model. The output is given as the provided function applied to the prediction of the k models.
## S4 method for signature 'SDMmodelCV' predict( object, data, fun = "mean", type = NULL, clamp = TRUE, filename = "", overwrite = FALSE, wopt = list(), extent = NULL, progress = TRUE, ... )## S4 method for signature 'SDMmodelCV' predict( object, data, fun = "mean", type = NULL, clamp = TRUE, filename = "", overwrite = FALSE, wopt = list(), extent = NULL, progress = TRUE, ... )
object |
SDMmodelCV object. |
data |
data.frame, SWD or raster rast with the data for the prediction. |
fun |
character. Function used to combine the output of the k models.
Note that fun is a character argument, you must use |
type |
character. Output type, see details, used only for Maxent and Maxnet methods. |
clamp |
logical for clumping during prediction, used only for Maxent and Maxnet methods. |
filename |
character. If provided the raster map is saved in a file. It must include the extension. |
overwrite |
logical. If |
wopt |
list. Writing options passed to writeRaster. |
extent |
ext object, if provided it restricts the prediction to the given extent. |
progress |
logical. If |
... |
Additional arguments to pass to the predict function. |
filename, and extent are arguments used only when the prediction is run for a rast object.
When a character vector is passed to the fun argument, than all the
given functions are applied and a named list is returned, see examples.
When filename is provided and the fun argument contains more than one
function name, the saved files are named as filename_fun, see example.
For models trained with the Maxent method the argument type can be:
"raw", "logistic" and "cloglog". The function performs the prediction in
R without calling the MaxEnt Java software. This results in a faster
computation for large datasets and might result in a slightly different
output compared to the Java software.
For models trained with the Maxnet method the argument type can be:
"link", "exponential", "logistic" and "cloglog", see maxnet
for more details.
For models trained with the ANN method the function uses the "raw" output type.
For models trained with the RF method the output is the probability of class 1.
For models trained with the BRT method the function uses the number of trees defined to train the model and the "response" output type.
A vector with the prediction or a rast object if data is a rast or a list in the case of multiple functions.
Sergio Vignali
Wilson P.D., (2009). Guidelines for computing MaxEnt model output values from a lambdas file.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Make cloglog prediction for the whole study area and get the result as # average of the k models predict(model, data = predictors, fun = "mean", type = "cloglog") # Make cloglog prediction for the whole study area, get the average, standard # deviation, and maximum values of the k models, and save the output in three # files. # The following commands save the output in the working directory. Note that # the `filename` must include the extension ## Not run: maps <- predict(model, data = predictors, fun = c("mean", "sd", "max"), type = "cloglog", filename = "prediction.tif") # In this case three files are created: prediction_mean.tif, # prediction_sd.tif and prediction_max.tif plotPred(maps$mean) plotPred(maps$sd) plotPred(maps$max) # Make logistic prediction for the whole study area, given as standard # deviation of the k models, and save it in a file predict(model, data = predictors, fun = "sd", type = "logistic", filename = "my_map.tif") ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Make cloglog prediction for the whole study area and get the result as # average of the k models predict(model, data = predictors, fun = "mean", type = "cloglog") # Make cloglog prediction for the whole study area, get the average, standard # deviation, and maximum values of the k models, and save the output in three # files. # The following commands save the output in the working directory. Note that # the `filename` must include the extension ## Not run: maps <- predict(model, data = predictors, fun = c("mean", "sd", "max"), type = "cloglog", filename = "prediction.tif") # In this case three files are created: prediction_mean.tif, # prediction_sd.tif and prediction_max.tif plotPred(maps$mean) plotPred(maps$sd) plotPred(maps$max) # Make logistic prediction for the whole study area, given as standard # deviation of the k models, and save it in a file predict(model, data = predictors, fun = "sd", type = "logistic", filename = "my_map.tif") ## End(Not run)
Given the coordinates, the species' name and the environmental variables, the function creates an SWD object (sample with data).
prepareSWD( species, env, p = NULL, a = NULL, categorical = NULL, verbose = TRUE )prepareSWD( species, env, p = NULL, a = NULL, categorical = NULL, verbose = TRUE )
species |
character. The name of the species. |
env |
rast containing the environmental variables used to extract the values at coordinate locations. |
p |
data.frame. The coordinates of the presence locations. |
a |
data.frame. The coordinates of the absence/background locations. |
categorical |
vector indicating which of the environmental variable are categorical. |
verbose |
logical, if |
The SWD object is created in a way that the presence locations are always before than the absence/background locations.
An SWD object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create the SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") data# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create the SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") data
Create random folds for cross validation.
randomFolds(data, k, only_presence = FALSE, seed = NULL)randomFolds(data, k, only_presence = FALSE, seed = NULL)
data |
SWD object that will be used to train the model. |
k |
integer. Number of fold used to create the partition. |
only_presence |
logical, if |
seed |
integer. The value used to set the seed for the fold partition. |
When only_presence = FALSE, the proportion of presence and absence
is preserved.
list with two matrices, the first for the training and the second for
the testing dataset. Each column of one matrix represents a fold with
TRUE for the locations included in and FALSE excluded from the partition.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting presence and absence locations folds <- randomFolds(data, k = 4) # Create 4 random folds splitting only the presence locations folds <- randomFolds(data, k = 4, only_presence = TRUE)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Create 4 random folds splitting presence and absence locations folds <- randomFolds(data, k = 4) # Create 4 random folds splitting only the presence locations folds <- randomFolds(data, k = 4, only_presence = TRUE)
The function performs a random search in the hyperparameters space, creating a population of random models each one with a random combination of the provided hyperparameters values.
randomSearch( model, hypers, metric, test = NULL, pop = 20, env = NULL, interactive = TRUE, progress = TRUE, seed = NULL )randomSearch( model, hypers, metric, test = NULL, pop = 20, env = NULL, interactive = TRUE, progress = TRUE, seed = NULL )
model |
SDMmodel or SDMmodelCV object. |
hypers |
named list containing the values of the hyperparameters that should be tuned, see details. |
metric |
character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc". |
test |
SWD object. Test dataset used to evaluate the model, not used with aicc and SDMmodelCV objects. |
pop |
numeric. Size of the population. |
env |
rast containing the environmental variables, used only with "aicc". |
interactive |
logical. If |
progress |
logical. If |
seed |
numeric. The value used to set the seed to have consistent results. |
To know which hyperparameters can be tuned you can use the output
of the function getTunableArgs. Hyperparameters not included in the
hypers argument take the value that they have in the passed model.
An interactive chart showing in real-time the steps performed by the algorithm is displayed in the Viewer pane.
SDMtune object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = seq(0.2, 3, 0.2), fc = c("lqp", "lqph", "lh")) # Run the function using as metric the AUC output <- randomSearch(model, hypers = h, metric = "auc", test = test, pop = 10, seed = 25) output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ]# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = seq(0.2, 3, 0.2), fc = c("lqp", "lqph", "lh")) # Run the function using as metric the AUC output <- randomSearch(model, hypers = h, metric = "auc", test = test, pop = 10, seed = 25) output@results output@models # Order results by highest test AUC output@results[order(-output@results$test_AUC), ]
Remove variables whose importance is less than the given threshold. The function removes one variable at time and after trains a new model to get the new variable contribution rank. If use_jk is TRUE the function checks if after removing the variable the model performance decreases (according to the given metric and based on the starting model). In this case the function stops removing the variable even if the contribution is lower than the given threshold.
reduceVar( model, th, metric, test = NULL, env = NULL, use_jk = FALSE, permut = 10, use_pc = FALSE, interactive = TRUE, verbose = TRUE )reduceVar( model, th, metric, test = NULL, env = NULL, use_jk = FALSE, permut = 10, use_pc = FALSE, interactive = TRUE, verbose = TRUE )
model |
SDMmodel or SDMmodelCV object. |
th |
numeric. The contribution threshold used to remove variables. |
metric |
character. The metric used to evaluate the models, possible
values are: "auc", "tss" and "aicc", used only if use_jk is |
test |
SWD object containing the test dataset used to
evaluate the model, not used with aicc, and if |
env |
rast containing the environmental variables, used only with "aicc". |
use_jk |
Flag to use the Jackknife AUC test during the variable
selection, if |
permut |
integer. Number of permutations, used if |
use_pc |
logical. If |
interactive |
logical. If |
verbose |
logical. If |
An interactive chart showing in real-time the steps performed by the algorithm is displayed in the Viewer pane.
The model trained using the selected variables.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a Maxnet model model <- train(method = "Maxnet", data = train, fc = "lq") # Remove all variables with permuation importance lower than 2% output <- reduceVar(model, th = 2, metric = "auc", test = test, permut = 1) # Remove variables with permuation importance lower than 3% only if testing # TSS doesn't decrease ## Not run: output <- reduceVar(model, th = 3, metric = "tss", test = test, permut = 1, use_jk = TRUE) # Remove variables with permuation importance lower than 2% only if AICc # doesn't increase output <- reduceVar(model, th = 2, metric = "aicc", permut = 1, use_jk = TRUE, env = predictors) # Train a Maxent model model <- train(method = "Maxent", data = train, fc = "lq") # Remove all variables with percent contribution lower than 2% output <- reduceVar(model, th = 2, metric = "auc", test = test, use_pc = TRUE) ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a Maxnet model model <- train(method = "Maxnet", data = train, fc = "lq") # Remove all variables with permuation importance lower than 2% output <- reduceVar(model, th = 2, metric = "auc", test = test, permut = 1) # Remove variables with permuation importance lower than 3% only if testing # TSS doesn't decrease ## Not run: output <- reduceVar(model, th = 3, metric = "tss", test = test, permut = 1, use_jk = TRUE) # Remove variables with permuation importance lower than 2% only if AICc # doesn't increase output <- reduceVar(model, th = 2, metric = "aicc", permut = 1, use_jk = TRUE, env = predictors) # Train a Maxent model model <- train(method = "Maxent", data = train, fc = "lq") # Remove all variables with percent contribution lower than 2% output <- reduceVar(model, th = 2, metric = "auc", test = test, use_pc = TRUE) ## End(Not run)
This Class represents a Random Forest model objects and hosts all the information related to the model.
## S4 method for signature 'RF' show(object)## S4 method for signature 'RF' show(object)
object |
RF object |
See randomForest for the meaning of the slots.
mtryinteger. Number of variable randomly sampled.
ntreeinteger. Number of grown trees.
nodesizeinteger. Minimum size of terminal nodes.
modelrandomForest. The randomForest model object.
Sergio Vignali
This Class represents an SDMmodel object and hosts all the information related to the model.
## S4 method for signature 'SDMmodel' show(object)## S4 method for signature 'SDMmodel' show(object)
object |
SDMmodel object |
dataSWD object. The data used to train the model.
modelSergio Vignali
Converts an SDMmodel object containing a Maxent model into a dismo MaxEnt object.
SDMmodel2MaxEnt(model)SDMmodel2MaxEnt(model)
model |
SDMmodel object to be converted. |
The converted dismo MaxEnt object.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l") dismo_model <- SDMmodel2MaxEnt(model) dismo_model# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l") dismo_model <- SDMmodel2MaxEnt(model) dismo_model
This Class represents an SDMmodel model object with replicates and hosts all the models trained during the cross validation.
## S4 method for signature 'SDMmodelCV' show(object)## S4 method for signature 'SDMmodelCV' show(object)
object |
SDMmodelCV object |
modelslist. A list containing all the models trained during the cross validation.
dataSWD object. Full dataset used to make the partitions.
foldslist with two matrices, the first for the training and the second
for the testing dataset. Each column of one matrix represents a fold with
TRUE for the locations included in and FALSE excluded from the partition.
Sergio Vignali
Class used to save the results of one of the following functions: gridSearch, randomSearch or optimizeModel.
Plot an SDMtune object. Use the interactive argument to create an interactive chart.
## S4 method for signature 'SDMtune' show(object) ## S4 method for signature 'SDMtune,missing' plot(x, title = "", interactive = FALSE)## S4 method for signature 'SDMtune' show(object) ## S4 method for signature 'SDMtune,missing' plot(x, title = "", interactive = FALSE)
object |
SDMtune object |
x |
SDMtune object. |
title |
character. The title of the plot. |
interactive |
logical, if TRUE plot an interactive chart. |
If interactive = FALSE the function returns a
ggplot object otherwise it returns an SDMtuneChart
object that contains the path of the temporary folder where the necessary
files to create the chart are saved. In both cases the objects are returned
as invisible.
resultsdata.frame. Results with the evaluation of the models.
modelslist. List of SDMmodel or SDMmodelCV objects.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = 1:5, fc = c("lqp", "lqph")) # Run the gridSearch function using as metric the AUC output <- gridSearch(model, hypers = h, metric = "auc", test = test) output # Plot the output plot(output, title = "My experiment") # Plot the interactive chart p <- plot(output, title = "My experiment", interactive = TRUE) # Print the temporary folder that stores the files used to create the chart str(p)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Define the hyperparameters to test h <- list(reg = 1:5, fc = c("lqp", "lqph")) # Run the gridSearch function using as metric the AUC output <- gridSearch(model, hypers = h, metric = "auc", test = test) output # Plot the output plot(output, title = "My experiment") # Plot the interactive chart p <- plot(output, title = "My experiment", interactive = TRUE) # Print the temporary folder that stores the files used to create the chart str(p)
Object similar to the MaxEnt SWD format that hosts the species name, the coordinates of the locations and the value of the environmental variables at the location places.
## S4 method for signature 'SWD' show(object)## S4 method for signature 'SWD' show(object)
object |
SWD object |
The object can contains presence/absence, presence/background, presence only or absence/background only data. Use the prepareSWD function to create the object.
speciescharacter. Name of the species.
coordsdata.frame. Coordinates of the locations.
datadata.frame. Value of the environmental variables at location sites.
panumeric. Vector with 1 for presence and 0 for absence/background
locations.
Sergio Vignali
Save an SWD object as csv file.
swd2csv(swd, file_name)swd2csv(swd, file_name)
swd |
SWD object. |
file_name |
character. The name of the file in which to save the object, see details. |
The file_name argument should include the extension (i.e. my_file.csv).
If file_name is a single name the function saves the presence
absence/background locations in a single file, adding the column pa with
1s for presence and 0s for absence/background locations. If file_name is a
vector with two names, it saves the object in two files: the first name
is used for the presence locations and the second for the absence/background
locations.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") ## Not run: # The following commands save the output in the working directory # Save the SWD object as a single csv file swd2csv(data, file_name = "train_data.csv") # Save the SWD object in two separate csv files swd2csv(data, file_name = c("presence.csv", "absence.csv")) ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") ## Not run: # The following commands save the output in the working directory # Save the SWD object as a single csv file swd2csv(data, file_name = "train_data.csv") # Save the SWD object in two separate csv files swd2csv(data, file_name = c("presence.csv", "absence.csv")) ## End(Not run)
Remove all but one location per raster cell. The function removes NAs and if more than one location falls within the same raster cell it selects randomly one.
thinData(coords, env, x = "x", y = "y", verbose = TRUE, progress = TRUE)thinData(coords, env, x = "x", y = "y", verbose = TRUE, progress = TRUE)
coords |
data.frame or matrix with the coordinates, see details. |
env |
rast containing the environmental variables. |
x |
character. Name of the column containing the x coordinates. |
y |
character. Name of the column containing the y coordinates. |
verbose |
logical, if |
progress |
logical, if |
coords and env must have the same coordinate reference system.
The coords argument can contain several columns. This is useful if the
user has information related to the coordinates that doesn't want to loose
with the thinning procedure. The function expects to have the x coordinates
in a column named "x", and the y coordinates in a column named "y". If this
is not the case, the name of the columns containing the coordinates can be
specified using the arguments x and y.
a matrix or a data frame with the thinned locations.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations, by sampling also on areas with NA values bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", xy = TRUE, values = FALSE) nrow(bg_coords) # Thin the locations # The function will remove the coordinates that have NA values for some # predictors. Note that the function expects to have the coordinates in two # columns named "x" and "y" colnames(bg_coords) thinned_bg <- thinData(bg_coords, env = predictors) nrow(thinned_bg) # Here we sample only on areas without NA values and then we double the # coordinates bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) thinned_bg <- thinData(rbind(bg_coords, bg_coords), env = predictors) nrow(thinned_bg) # In case of a dataframe containing more than two columns (e.g. a dataframe # with the coordinates plus an additional column with the age of the species) # and custom column names, use the function in this way age <- sample(c(1, 2), size = nrow(bg_coords), replace = TRUE) data <- cbind(age, bg_coords) colnames(data) <- c("age", "X", "Y") thinned_bg <- thinData(data, env = predictors, x = "X", y = "Y") head(data)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare background locations, by sampling also on areas with NA values bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", xy = TRUE, values = FALSE) nrow(bg_coords) # Thin the locations # The function will remove the coordinates that have NA values for some # predictors. Note that the function expects to have the coordinates in two # columns named "x" and "y" colnames(bg_coords) thinned_bg <- thinData(bg_coords, env = predictors) nrow(thinned_bg) # Here we sample only on areas without NA values and then we double the # coordinates bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) thinned_bg <- thinData(rbind(bg_coords, bg_coords), env = predictors) nrow(thinned_bg) # In case of a dataframe containing more than two columns (e.g. a dataframe # with the coordinates plus an additional column with the age of the species) # and custom column names, use the function in this way age <- sample(c(1, 2), size = nrow(bg_coords), replace = TRUE) data <- cbind(age, bg_coords) colnames(data) <- c("age", "X", "Y") thinned_bg <- thinData(data, env = predictors, x = "X", y = "Y") head(data)
Compute three threshold values: minimum training presence, equal training sensitivity and specificity and maximum training sensitivity plus specificity together with fractional predicted area and the omission rate. If a test dataset is provided it returns also the equal test sensitivity and specificity and maximum test sensitivity plus specificity thresholds and the p-values of the one-tailed binomial exact test.
thresholds(model, type = NULL, test = NULL)thresholds(model, type = NULL, test = NULL)
model |
SDMmodel object. |
type |
character. The output type used for "Maxent" and "Maxnet" methods, possible values are "cloglog" and "logistic". |
test |
SWD testing locations, if not provided it returns the training and test thresholds. |
The equal training sensitivity and specificity minimizes the difference between sensitivity and specificity. The one-tailed binomial test checks that test points are predicted no better than by a random prediction with the same fractional predicted area.
data.frame with the thresholds.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Get the cloglog thresholds thresholds(model, type = "cloglog") # Get the logistic thresholds passing the test dataset thresholds(model, type = "logistic", test = test)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Get the cloglog thresholds thresholds(model, type = "cloglog") # Get the logistic thresholds passing the test dataset thresholds(model, type = "logistic", test = test)
Train a model using one of the following methods: Artificial Neural Networks, Boosted Regression Trees, Maxent, Maxnet or Random Forest.
train(method, data, folds = NULL, progress = TRUE, ...)train(method, data, folds = NULL, progress = TRUE, ...)
method |
character or character vector. Method used to train the model, possible values are "ANN", "BRT", "Maxent", "Maxnet" or "RF", see details. |
data |
SWD object with presence and absence/background locations. |
folds |
list. Output of the function randomFolds or folds object created with other packages, see details. |
progress |
logical. If |
... |
Arguments passed to the relative method, see details. |
For the ANN method possible arguments are (for more details see nnet):
size: integer. Number of the units in the hidden layer.
decay numeric. Weight decay, default is 0.
rang numeric. Initial random weights, default is 0.7.
maxit integer. Maximum number of iterations, default is 100.
For the BRT method possible arguments are (for more details see gbm):
distribution: character. Name of the distribution to use, default is "bernoulli".
n.trees: integer. Maximum number of tree to grow, default is 100.
interaction.depth: integer. Maximum depth of each tree, default is 1.
shrinkage: numeric. The shrinkage parameter, default is 0.1.
bag.fraction: numeric. Random fraction of data used in the tree expansion, default is 0.5.
For the RF method the model is trained as classification. Possible arguments are (for more details see randomForest):
mtry: integer. Number of variable randomly sampled at each split,
default is floor(sqrt(number of variables)).
ntree: integer. Number of tree to grow, default is 500.
nodesize: integer. Minimum size of terminal nodes, default is 1.
Maxent models are trained using the arguments
"removeduplicates=false" and "addsamplestobackground=false".
Use the function thinData to remove duplicates and the function
addSamplesToBg to add presence locations to background locations. For
the Maxent method, possible arguments are:
reg: numeric. The value of the regularization multiplier, default is 1.
fc: character. The value of the feature classes, possible values are combinations of "l", "q", "p", "h" and "t", default is "lqph".
iter: numeric. Number of iterations used by the MaxEnt algorithm, default is 500.
Maxnet models are trained using the argument
"addsamplestobackground = FALSE", use the function addSamplesToBg
to add presence locations to background locations. For the Maxnet method,
possible arguments are (for more details see maxnet):
reg: numeric. The value of the regularization intensity, default is 1.
fc: character. The value of the feature classes, possible values are combinations of "l", "q", "p", "h" and "t", default is "lqph".
The folds argument accepts also objects created with other packages: ENMeval or blockCV. In this case the function converts internally the folds into a format valid for SDMtune.
When multiple methods are given as method argument, the function returns a
named list of model object, with the name corresponding to the used method,
see examples.
An SDMmodel or SDMmodelCV or a list of model objects.
Sergio Vignali
Venables, W. N. & Ripley, B. D. (2002) Modern Applied Statistics with S. Fourth Edition. Springer, New York. ISBN 0-387-95457-0.
Brandon Greenwell, Bradley Boehmke, Jay Cunningham and GBM Developers (2019). gbm: Generalized Boosted Regression Models. https://CRAN.R-project.org/package=gbm.
A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18–22.
Hijmans, Robert J., Steven Phillips, John Leathwick, and Jane Elith. 2017. dismo: Species Distribution Modeling. https://cran.r-project.org/package=dismo.
Steven Phillips (2017). maxnet: Fitting 'Maxent' Species Distribution Models with 'glmnet'. https://CRAN.R-project.org/package=maxnet.
Muscarella, R., Galante, P.J., Soley-Guardia, M., Boria, R.A., Kass, J., Uriarte, M. and R.P. Anderson (2014). ENMeval: An R package for conducting spatially independent evaluations and estimating optimal model complexity for ecological niche models. Methods in Ecology and Evolution.
Roozbeh Valavi, Jane Elith, José Lahoz-Monfort and Gurutzeta Guillera-Arroita (2018). blockCV: Spatial and environmental blocking for k-fold cross-validation. https://github.com/rvalavi/blockCV.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") ## Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l", reg = 1.5, iter = 700) # Add samples to background. This should be done preparing the data before # training the model without using data <- addSamplesToBg(data) model <- train("Maxent", data = data) ## Train a Maxnet model model <- train(method = "Maxnet", data = data, fc = "lq", reg = 1.5) ## Cross Validation # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = folds) ## Not run: # Run only if you have the package ENMeval installed ## Block partition using the ENMeval package require(ENMeval) block_folds <- get.block(occ = data@coords[data@pa == 1, ], bg.coords = data@coords[data@pa == 0, ]) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = block_folds) ## Checkerboard1 partition using the ENMeval package cb_folds <- get.checkerboard1(occ = data@coords[data@pa == 1, ], env = predictors, bg.coords = data@coords[data@pa == 0, ], aggregation.factor = 4) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = cb_folds) ## Environmental block using the blockCV package # Run only if you have the package blockCV require(blockCV) # Create sf object sf_df <- sf::st_as_sf(cbind(data@coords, pa = data@pa), coords = c("X", "Y"), crs = terra::crs(predictors, proj = TRUE)) # Spatial blocks spatial_folds <- cv_spatial(x = sf_df, column = "pa", rows_cols = c(8, 10), k = 5, hexagon = FALSE, selection = "systematic") model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = spatial_folds) ## End(Not run) ## Train presence absence models # Prepare presence and absence locations p_coords <- virtualSp$presence a_coords <- virtualSp$absence # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = a_coords, env = predictors[[1:5]]) ## Train an Artificial Neural Network model model <- train("ANN", data = data, size = 10) ## Train a Random Forest model model <- train("RF", data = data, ntree = 300) ## Train a Boosted Regression Tree model model <- train("BRT", data = data, n.trees = 300, shrinkage = 0.001) ## Multiple methods trained together with default arguments output <- train(method = c("ANN", "BRT", "RF"), data = data, size = 10) output$ANN output$BRT output$RF ## Multiple methods trained together passing extra arguments output <- train(method = c("ANN", "BRT", "RF"), data = data, size = 10, ntree = 300, n.trees = 300, shrinkage = 0.001) output# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") ## Train a Maxent model model <- train(method = "Maxent", data = data, fc = "l", reg = 1.5, iter = 700) # Add samples to background. This should be done preparing the data before # training the model without using data <- addSamplesToBg(data) model <- train("Maxent", data = data) ## Train a Maxnet model model <- train(method = "Maxnet", data = data, fc = "lq", reg = 1.5) ## Cross Validation # Create 4 random folds splitting only the presence data folds <- randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = folds) ## Not run: # Run only if you have the package ENMeval installed ## Block partition using the ENMeval package require(ENMeval) block_folds <- get.block(occ = data@coords[data@pa == 1, ], bg.coords = data@coords[data@pa == 0, ]) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = block_folds) ## Checkerboard1 partition using the ENMeval package cb_folds <- get.checkerboard1(occ = data@coords[data@pa == 1, ], env = predictors, bg.coords = data@coords[data@pa == 0, ], aggregation.factor = 4) model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = cb_folds) ## Environmental block using the blockCV package # Run only if you have the package blockCV require(blockCV) # Create sf object sf_df <- sf::st_as_sf(cbind(data@coords, pa = data@pa), coords = c("X", "Y"), crs = terra::crs(predictors, proj = TRUE)) # Spatial blocks spatial_folds <- cv_spatial(x = sf_df, column = "pa", rows_cols = c(8, 10), k = 5, hexagon = FALSE, selection = "systematic") model <- train(method = "Maxnet", data = data, fc = "l", reg = 0.8, folds = spatial_folds) ## End(Not run) ## Train presence absence models # Prepare presence and absence locations p_coords <- virtualSp$presence a_coords <- virtualSp$absence # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = a_coords, env = predictors[[1:5]]) ## Train an Artificial Neural Network model model <- train("ANN", data = data, size = 10) ## Train a Random Forest model model <- train("RF", data = data, ntree = 300) ## Train a Boosted Regression Tree model model <- train("BRT", data = data, n.trees = 300, shrinkage = 0.001) ## Multiple methods trained together with default arguments output <- train(method = c("ANN", "BRT", "RF"), data = data, size = 10) output$ANN output$BRT output$RF ## Multiple methods trained together passing extra arguments output <- train(method = c("ANN", "BRT", "RF"), data = data, size = 10, ntree = 300, n.trees = 300, shrinkage = 0.001) output
Split a dataset randomly in training and testing datasets or training, validation and testing datasets.
trainValTest(x, test, val = 0, only_presence = FALSE, seed = NULL)trainValTest(x, test, val = 0, only_presence = FALSE, seed = NULL)
x |
SWD object containing the data that have to be split in training, validation and testing datasets. |
test |
numeric. The percentage of data withhold for testing. |
val |
numeric. The percentage of data withhold for validation, default
is |
only_presence |
logical. If |
seed |
numeric. The value used to set the seed in order to have
consistent results, default is |
When only_presence = FALSE, the proportion of presence and
absence is preserved.
A list with the training, validation and testing or training and testing SWD objects accordingly.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets # and splitting only the presence locations datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Split presence locations in training (60%), validation (20%) and testing # (20%) datasets and splitting the presence and the absence locations datasets <- trainValTest(data, val = 0.2, test = 0.2) train <- datasets[[1]] val <- datasets[[2]] test <- datasets[[3]]# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets # and splitting only the presence locations datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Split presence locations in training (60%), validation (20%) and testing # (20%) datasets and splitting the presence and the absence locations datasets <- trainValTest(data, val = 0.2, test = 0.2) train <- datasets[[1]] val <- datasets[[2]] test <- datasets[[3]]
Compute the max TSS of a given model.
tss(model, test = NULL)tss(model, test = NULL)
model |
SDMmodel or SDMmodelCV object. |
test |
SWD object when |
For SDMmodelCV objects, the function computes the
mean of the training TSS values of the k-folds. If test = TRUE it computes
the mean of the testing TSS values for the k-folds. If test is an
SWD object, it computes the mean TSS values for the provided
testing dataset.
The value of the TSS of the given model.
Sergio Vignali
Allouche O., Tsoar A., Kadmon R., (2006). Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology, 43(6), 1223–1232.
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute the training TSS tss(model) # Compute the testing TSS tss(model, test = test) # Same example but using cross validation instead of training and # testing datasets. Create 4 random folds splitting only the presence # locations folds = randomFolds(train, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = train, fc = "l", folds = folds) # Compute the training TSS tss(model) # Compute the testing TSS tss(model, test = TRUE) # Compute the TSS for the held apart testing dataset tss(model, test = test)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute the training TSS tss(model) # Compute the testing TSS tss(model, test = test) # Same example but using cross validation instead of training and # testing datasets. Create 4 random folds splitting only the presence # locations folds = randomFolds(train, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = train, fc = "l", folds = folds) # Compute the training TSS tss(model) # Compute the testing TSS tss(model, test = TRUE) # Compute the TSS for the held apart testing dataset tss(model, test = test)
The function randomly permutes one variable at time (using training and absence/background datasets) and computes the decrease in training AUC. The result is normalized to percentages. Same implementation of MaxEnt java software but with the additional possibility of running several permutations to obtain a better estimate of the permutation importance. In case of more than one permutation (default is 10) the average of the decrease in training AUC is computed.
varImp(model, permut = 10, progress = TRUE)varImp(model, permut = 10, progress = TRUE)
model |
SDMmodel or SDMmodelCV object. |
permut |
integer. Number of permutations. |
progress |
logical. If |
Note that it could return values slightly different from MaxEnt Java software due to a different random permutation.
For SDMmodelCV objects the function returns the average and the standard deviation of the permutation importances of the single models.
data.frame with the ordered permutation importance.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute variable importance vi <- varImp(model, permut = 5) vi # Same example but using cross validation instead of training and testing # datasets # Create 4 random folds splitting only the presence locations folds = randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Compute variable importance vi <- varImp(model, permut = 5) vi# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Compute variable importance vi <- varImp(model, permut = 5) vi # Same example but using cross validation instead of training and testing # datasets # Create 4 random folds splitting only the presence locations folds = randomFolds(data, k = 4, only_presence = TRUE) model <- train(method = "Maxnet", data = data, fc = "l", folds = folds) # Compute variable importance vi <- varImp(model, permut = 5) vi
The function performs a data-driven variable selection. Starting from the provided model it iterates through all the variables starting from the one with the highest contribution (permutation importance or maxent percent contribution). If the variable is correlated with other variables (according to the given method and threshold) it performs a Jackknife test and among the correlated variables it removes the one that results in the best performing model when removed (according to the given metric for the training dataset). The process is repeated until the remaining variables are not highly correlated anymore.
varSel( model, metric, bg4cor, test = NULL, env = NULL, method = "spearman", cor_th = 0.7, permut = 10, use_pc = FALSE, interactive = TRUE, progress = TRUE, verbose = TRUE )varSel( model, metric, bg4cor, test = NULL, env = NULL, method = "spearman", cor_th = 0.7, permut = 10, use_pc = FALSE, interactive = TRUE, progress = TRUE, verbose = TRUE )
model |
SDMmodel or SDMmodelCV object. |
metric |
character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc". |
bg4cor |
SWD object. Background locations used to test the correlation between environmental variables. |
test |
SWD. Test dataset used to evaluate the model, not used with aicc and SDMmodelCV objects. |
env |
rast containing the environmental variables, used only with "aicc". |
method |
character. The method used to compute the correlation matrix. |
cor_th |
numeric. The correlation threshold used to select highly correlated variables. |
permut |
integer. Number of permutations. |
use_pc |
logical, use percent contribution. If |
interactive |
logical. If |
progress |
logical. If |
verbose |
logical. If |
An interactive chart showing in real-time the steps performed by the algorithm is displayed in the Viewer pane.
To find highly correlated variables the following formula is used:
The SDMmodel or SDMmodelCV object trained using the selected variables.
Sergio Vignali
# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Prepare background locations to test autocorrelation, this usually gives a # warning message given that less than 10000 points can be randomly sampled bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") ## Not run: # Remove variables with correlation higher than 0.7 accounting for the AUC, # in the following example the variable importance is computed as permutation # importance vs <- varSel(model, metric = "auc", bg4cor = bg, test = test, cor_th = 0.7, permut = 1) vs # Remove variables with correlation higher than 0.7 accounting for the TSS, # in the following example the variable importance is the MaxEnt percent # contribution # Train a model model <- train(method = "Maxent", data = train, fc = "l") vs <- varSel(model, metric = "tss", bg4cor = bg, test = test, cor_th = 0.7, use_pc = TRUE) vs # Remove variables with correlation higher than 0.7 accounting for the aicc, # in the following example the variable importance is the MaxEnt percent # contribution vs <- varSel(model, metric = "aicc", bg4cor = bg, cor_th = 0.7, use_pc = TRUE, env = predictors) vs ## End(Not run)# Acquire environmental variables files <- list.files(path = file.path(system.file(package = "dismo"), "ex"), pattern = "grd", full.names = TRUE) predictors <- terra::rast(files) # Prepare presence and background locations p_coords <- virtualSp$presence bg_coords <- virtualSp$background # Create SWD object data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords, env = predictors, categorical = "biome") # Split presence locations in training (80%) and testing (20%) datasets datasets <- trainValTest(data, test = 0.2, only_presence = TRUE) train <- datasets[[1]] test <- datasets[[2]] # Train a model model <- train(method = "Maxnet", data = train, fc = "l") # Prepare background locations to test autocorrelation, this usually gives a # warning message given that less than 10000 points can be randomly sampled bg_coords <- terra::spatSample(predictors, size = 9000, method = "random", na.rm = TRUE, xy = TRUE, values = FALSE) bg <- prepareSWD(species = "Virtual species", a = bg_coords, env = predictors, categorical = "biome") ## Not run: # Remove variables with correlation higher than 0.7 accounting for the AUC, # in the following example the variable importance is computed as permutation # importance vs <- varSel(model, metric = "auc", bg4cor = bg, test = test, cor_th = 0.7, permut = 1) vs # Remove variables with correlation higher than 0.7 accounting for the TSS, # in the following example the variable importance is the MaxEnt percent # contribution # Train a model model <- train(method = "Maxent", data = train, fc = "l") vs <- varSel(model, metric = "tss", bg4cor = bg, test = test, cor_th = 0.7, use_pc = TRUE) vs # Remove variables with correlation higher than 0.7 accounting for the aicc, # in the following example the variable importance is the MaxEnt percent # contribution vs <- varSel(model, metric = "aicc", bg4cor = bg, cor_th = 0.7, use_pc = TRUE, env = predictors) vs ## End(Not run)
Dataset containing a random generated virtual species. The purpose of this dataset is to demonstrate the use of the functions included in the package.
virtualSpvirtualSp
A list with five elements:
400 random generated coordinates for the presence locations.
300 random generated coordinates for the absence locations.
5000 random generated coordinates for the background locations.
Presence absence map used to extract the presence and absence locations.
Probability map of the random generated virtual species.
The random species has been generated using the package virtualspecies.
Leroy, B. , Meynard, C. N., Bellard, C. and Courchamp, F. (2016), virtualspecies, an R package to generate virtual species distributions. Ecography, 39: 599-607. doi:10.1111/ecog.01388