Package 'caretSDM'

Description

This function includes new predictors to the sdm_area object.

Usage

add_predictors(sa, pred, variables_selected = NULL, gdal = TRUE,
                      lines_as_sdm_area = FALSE)

get_predictors(i)

Arguments

Details

add_predictors returns a sdm_area object with a grid built upon the x parameter. There are two ways to make the grid and resample the variables in sdm_area: with and without gdal. As standard, if gdal is available in you machine it will be used (gdal = TRUE), otherwise sf/stars will be used. lines_as_sdm_area and gdal parameters are passed to sdm_area function, so they will be used in the grid creation and resampling of predictors. They will be retrieved automatically from the sdm_area object.

Value

For add_predictors the same input sdm_area object is returned including the pred data binded to the previous grid. get_predictors retrieves the grid from the i object.

Author(s)

Luíz Fernando Esser ([email protected]) and Reginaldo Ré. https://luizfesser.wordpress.com

See Also

sdm_area get_predictor_names bioc

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc)

# Retrieve predictors data:
get_predictors(sa)

Description

This function includes scenarios in the sdm_area object.

Usage

add_scenarios(sa, scen = NULL, scenarios_names = NULL, pred_as_scen = TRUE,
                     variables_selected = NULL, stationary = NULL, crop_area = NULL)

set_scenarios_names(i, scenarios_names = NULL)

scenarios_names(i)

get_scenarios_data(i)

select_scenarios(i, scenarios_names = NULL)

Arguments

Details

The function add_scenarios adds scenarios to the sdm_area or input_sdm object. If scen has variables that are not present as predictors the function will use only variables present in both objects. stationary variables are those that don't change through the scenarios. It is useful for hidrological variables in fish habitat modeling, for example (see examples below). When adding multiple scenarios in multiple runs, the function will always add a new "current" scenario. To avoid that, set pred_as_scen = FALSE.

Value

add_scenarios returns the input sdm_area or input_sdm object with a new slot called scenarios with scen data as a list, where each slot of the list holds a scenario and each scenario is a sf object. set_scenarios_names sets new names for scenarios in sdm_area/input_sdm object. scenarios_names returns scenarios' names. get_scenarios_data retrieves scenarios data as a list of sf objects. select_scenarios selects scenarios from sdm_area/input_sdm object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

sdm_area input_sdm

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc)

# Include scenarios:
sa <- add_scenarios(sa, scen[1:2]) |> select_predictors(c("bio1", "bio12"))

# Set scenarios names:
sa <- set_scenarios_names(sa, scenarios_names = c(
  "future_1", "future_2",
  "current"
))
scenarios_names(sa)

# Get scenarios data:
scenarios_grid <- get_scenarios_data(sa)
scenarios_grid

# Select scenarios:
sa <- select_scenarios(sa, scenarios_names = c("future_1"))

# Setting stationary variables in scenarios:
sa <- sdm_area(rivs[c(1:200), ], cell_size = 100000, output_crs = 6933, lines_as_sdm_area = TRUE) |>
  add_predictors(bioc) |>
  add_scenarios(scen, stationary = c("LENGTH_KM", "DIST_DN_KM"))

Description

A data.frame with characteristics of each algorithm available in caretSDM. Each column is a different characteristics. This can be helpful for more experienced modelers select algorithms. See the source for a selection method using this data.

Usage

algorithms

Format

## 'algorithms' A data.frame with 230 rows and 60 columns:

X

Algorithms names

Further columns

Algorithms attributes

Source

<https://topepo.github.io/caret/models-clustered-by-tag-similarity.html>

Description

This function obtains background data given a set of predictors.

Usage

background(occ,
           pred = NULL,
           method = "random",
           n_set = 1,
           n_bg = 10000,
           proportion = NULL)

n_background(i)

background_method(i)

background_data(i)

Arguments

Details

background is used in the SDM workflow to obtain background data, a step necessary for MaxEnt algorithm to run. This function helps avoid the use of pseudoabsence data in background algorithms and the use of background data in pseudoabsence algorithms, a very common mistake. If user provides a custom function, it must have the arguments env_sf and occ_sf, which will consist of two "sf"s. The first has the predictor values for the whole study area, while the second has the presence records for the species. The function must return a vector with cell_ids of the background data.

n_background returns the number of background records obtained per species.

background_data returns a list of species names. Each species name will have a lists with background data from class sf.

Value

A occurrences_sdm or input_sdm object with background data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

link{input_sdm} pseudoabsences occurrences_sdm get_occurrences get_predictors

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Background generation:
i <- background(i, proportion = 1) # All available data is obtained as background data.

Description

A stars object with bioclimatic variables (bio1, bio4 and bio12) for the Parana state in Brazil. Data obtained from WorldClim 2.1 at 10 arc-min resolution.

Usage

bioc

Format

## 'bioc' A stars with 1 attribute and 3 bands:

bio1

Annual Mean Temperature

bio4

Temperature Seasonality

bio12

Annual Precipitation

Source

<https://www.worldclim.org/>

Description

Create buffer around records in occ_data to be used as study area

Usage

buffer_sdm(occ_data, size = NULL, occ_crs = NULL, mcp = FALSE)

Arguments

Value

A sf buffer around occ_data records.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

GBIF_data

Examples

# Create sdm_area object:
study_area <- buffer_sdm(occ, size = 50000, occ_crs = 6933)
plot(study_area)

Description

This function aims to unveil the correlation of different algorithms outputs. For that, it uses the predictions on current scenario, but other scenarios can be tested.

Usage

correlate_sdm(i, scenario = "current")

Arguments

Value

A data.frame with pearson correlation between projections.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

if (interactive()) {
  # Create sdm_area object:
  set.seed(1)
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

  # Include scenarios:
  sa <- add_scenarios(sa)

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    repeats = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)

  # Check correlations:
  correlate_sdm(i)
}

Description

Data cleaning wrapper using CoordinateCleaner package.

Usage

data_clean(occ, pred = NULL,
           species = NA, lon = NA, lat = NA,
           capitals = TRUE,
           centroids = TRUE,
           duplicated = TRUE,
           identical = TRUE,
           institutions = TRUE,
           invalid = TRUE,
           terrestrial = TRUE,
           independent_test = TRUE,
           fun = NULL)

Arguments

Details

If the user does not used GBIF_data function to obtain species records, the function may have problems to find which column from the presences table has species, longitude and latitude information. In this regard, we implemented the parameters species, lon and lat so the use can explicitly inform which columns should be used. If they remain as NA (standard) the function will try to guess which columns are the correct one.

Value

A occurrences_sdm object or input_sdm with cleaned presence data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

GBIF_data occurrences_sdm sdm_area input_sdm get_predictor_names

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Clean coordinates (terrestrial is set to false to make the run quicker):
i <- data_clean(i, terrestrial = FALSE)

Description

Calculates ensemble predictions for species distribution models using custom or implemented methods.

Usage

ensemble_sdm(m,
            scen = NULL,
            method = "average",
            metric = NULL,
            fun = NULL
            )

get_ensembles(
  i,
  type = "matrix",
  spp_name = NULL,
  scenario = NULL,
  ensemble_type = NULL
)

add_ensembles(e1, e2)

Arguments

Details

ensembles could be set to three different strategies OR a custom function. The three implemented strategies are: average is the mean occurrence probability, which is a simple mean of predictions; weighted_average is the same average, but weighted by a metric, which needs to be set using argument metric (see mean_validation_metrics for the metrics available). committee_average is the committee average, as known as majority rule, where predictions are binarized and then a mean is obtained. To binarize predictions, user can set a custom function in the fun argument to calculate a threshold for each model. Standardly, the committee average uses the caret::thresholder function to find the threshold that maximizes the sum of sensitivity and specificity (through caretSDM:::.MaxSeSp). Custom function (fun) must use the argument mod, which is the model output from caret package (see get_models) and must return a numeric value (see example). method can also be set to a custom function, which must receive the argument pred_mat, which is a matrix of predictions (columns are models and rows are cells) and return a vector of predictions (one value per cell). See the median example below for a custom function.

get_predictions returns the list of all predictions to all scenarios, all species, all algorithms and all repetitions. Useful for those who wish to implement their own ensemble methods.

get_ensembles returns a matrix of data.frames, where each column is a scenario and each row is a species.

scenarios_names returns the scenarios names in a sdm_area or input_sdm object.

get_scenarios_data returns the data from scenarios in a sdm_area or input_sdm object.

Value

A input_sdm or a predictions object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

sdm_area input_sdm mean_validation_metrics

Examples

if (interactive()) {
  # Create sdm_area object:
  set.seed(1)
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

  # Include scenarios:
  sa <- add_scenarios(sa)

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    repeats = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)

  # Ensemble:
  i <- ensemble_sdm(i, method = "average")
  i
}

# Example from a custom function to obtain the threshold that maximizes
# the sensitivity plus specificity:
MaxSeSp <- function(mod) {
  th <- caret::thresholder(mod,
    threshold = seq(0, 1, by = 0.001),
    final = TRUE,
    statistics = c("Sensitivity", "Specificity")
  )
  th <- th$prob_threshold[which.max(th$Sensitivity + th$Specificity)]
  if (length(th) > 1) mean(th) else th
}

# Example from a custom function to obtain ensembles using the median instead of the mean:
median_ensemble <- function(pred_mat) {
  apply(pred_mat, 1, median, na.rm = TRUE)
}

Description

This function is a wrapper to get records from GBIF using rgbif and return a data.frame ready to be used in caretSDM.

Usage

GBIF_data(s, file = NULL, as_df = FALSE, ...)

Arguments

Value

A data.frame with species occurrences data, or an occurrences object if as_df = FALSE.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

References

https://www.gbif.org

Examples

## Select species names:
# s <- c("Araucaria angustifolia", "Salminus brasiliensis")

## Run function:
# oc <- GBIF_data(s)

Description

An ensembling method to group different GCMs into one SSP scenario

Usage

gcms_ensembles(i, gcms = NULL)

Arguments

Value

A input_sdm object with grouped GCMs.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

GBIF_data occurrences_sdm sdm_area input_sdm get_predictor_names

Examples

if (interactive()) {
  # Create sdm_area object:
  set.seed(1)
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc)

  # Include scenarios:
  sa <- add_scenarios(sa, scen) |> select_predictors(c("bio1", "bio12"))

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i,
    algo = c("naive_bayes"),
    ctrl = ctrl_sdm,
    variables_selected = c("bio1", "bio12")
  ) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)

  # Ensemble:
  i <- ensemble_sdm(i, method = "average")
  i

  # Ensemble GCMs:
  i <- gcms_ensembles(i, gcms = c("ca", "mi"))
  i
}

Description

This function creates a new input_sdm object.

Usage

input_sdm(...)

add_input_sdm(i1, i2)

Arguments

Details

If sdm_area is used, it can include predictors and scenarios. In this case, input_sdm will detect and include as scenarios and predictors in the input_sdm output. Objects can be included in any order, since the function will work by detecting their classes. The returned object is used throughout the whole workflow to apply functions.

Value

A input_sdm object containing:

Author(s)

Luiz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

occurrences_sdm sdm_area

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa, scen)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

Description

This functions returns a boolean to check caretSDM object classes.

Usage

is_input_sdm(x)

is_sdm_area(x)

is_occurrences(x)

is_models(x)

is_predictions(x)

Arguments

Value

Boolean.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

is_sdm_area(sa)

is_input_sdm(sa)

Description

Join cell_id data from sdm_area to a occurrences

Usage

join_area(occ, pred)

Arguments

Details

This function is key in this SDM workflow. It attaches cell_id values to occ, deletes records outside pred and allows the use of pseudoabsences. This function also tests if CRS from both occ and pred are equal, otherwise the CRS of pred is used to convert occ.

Value

A occurrences object with cell_id to each record.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

occurrences_sdm sdm_area input_sdm pseudoabsences

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa, scen)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933) |> join_area(sa)

Description

Apply multicollinearity calculation on predictors.

Usage

multicollinearity_sdm(pred,
                             method = NULL,
                             variables_selected = NULL,
                             cumulative_proportion = 0.99,
                             th = 0.5,
                             ...)

selected_variables(i)

Arguments

Details

multicollinearity_sdm is a wrapper function to run usdm::vifcor, usdm::vifstep or a pca in caretSDM, but also provides a way to implement custom functions to reduce multicollinearity. If user provides a custom function, it must have the arguments env_sf and occ_sf, which will consist of two sfs. The first has the predictor values for the whole study area, while the second has the presence records for the species. The function must return a vector with selected variables.

Value

A input_sdm or predictors object with VIF data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

vif_predictors pca_predictors get_predictor_names

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa, scen)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# VIF calculation:
i <- multicollinearity_sdm(i, method = "vifcor", th = 0.5)
i

# Retrieve information about vif:
vif_summary(i)
selected_variables(i)

# Example of custom function:
custom_function <- function(env_sf, occ_sf) {
  env_df <- dplyr::select(sf::st_drop_geometry(env_sf), -"cell_id")
  correlations <- cor(env_df)
  col <- caret::findCorrelation(correlations, cutoff = 0.7)
  selected <- colnames(correlations)[-col]
  return(selected)
}

Description

A data.frame object with Araucaria angustifolia occurrence data obtained from GBIF and filtered with Parana state sf.

Usage

occ

Format

## 'occ' A data.frame with 420 rows and 3 columns (EPSG:6933):

species

Species name

decimalLongitude

Longitude in meters

decimalLatitude

Latitude in meters

Source

<https://www.gbif.org>

Description

This function creates and manage occurrences objects.

Usage

occurrences_sdm(occ,
                independent_test = NULL,
                p = 0.1,
                occ_crs = NULL,
                independent_test_crs = NULL,
                crs = NULL,
                ...)

n_records(i)

species_names(i)

get_coords(i)

get_occurrences(i)

occurrences_as_df(i)

add_occurrences(oc1, oc2)

Arguments

Details

occ must have three columns: species, decimalLongitude and decimalLatitude. When sf it is only necessary a species column. n_records return the number of presence records to each species. species_names return the species names. get_coords return a data.frame with coordinates of species records. get_occurrences return a sf with coordinates of species records, species names and cell_ids. add_occurrences return a occurrences. This function sums two occurrences objects. It can also sum a occurrences object with a data.frame object. occurrences_as_df returns a data.frame with species names and coordinates.

Value

A occurrences object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

input_sdm GBIF_data occ

Examples

# Create occurrences:
oc <- occurrences_sdm(occ, crs = 6933)

Description

A sf object with a polygon for the Paraná state in Brazil. This is a subset of the brazilian map provided by official government agency (IBGE)

Usage

parana

Format

## 'parana' A sf with 1 row and 5 columns:

GID0

State code

CODIGOIB1

State's phone code

NOMEUF2

Name of the state

SIGLAUF3

Abbreviation of the state's name

geom

Geometry column of the sf

Source

<https://www.ibge.gov.br/geociencias/cartas-e-mapas/bases-cartograficas-continuas/15759-brasil.html>

Description

Transform predictors data into PCA-axes.

Usage

pca_predictors(i, cumulative_proportion = 0.99)

pca_summary(i)

get_pca_model(i)

Arguments

Details

pca_predictors Transform predictors data into PCA-axes. If the user wants to use PCA-axes as future scenarios, then scenarios should be added after the PCA transformation (see examples). pca_summary Returns the summary of prcomp function. See ?stats::prcomp. get_pca_model Returns the model built to calculate PCA-axes.

Value

input_sdm object with variables from both predictors and scenarios transformed in PCA-axes.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

vif_predictors sdm_area add_scenarios add_predictors

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# PCA transformation:
i <- pca_predictors(i)

Description

Obtain the Partial Dependence Plots (PDP) to each variable.

Usage

pdp_sdm(i, spp = NULL, algo = NULL, variables_selected = NULL, mean.only = FALSE)

get_pdp_sdm(i, spp = NULL, algo = NULL, variables_selected = NULL)

Arguments

Value

A plot (for pdp_sdm) or a data.frame (for get_pdp_sdm) with PDP values.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

varImp_sdm

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random", n_set = 3)

# Custom trainControl:
ctrl_sdm <- caret::trainControl(
  method = "repeatedcv",
  number = 2,
  repeats = 1,
  classProbs = TRUE,
  returnResamp = "all",
  summaryFunction = summary_sdm,
  savePredictions = "all"
)
# Train models:
i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm)

# PDP plots:
pdp_sdm(i)
get_pdp_sdm(i)

Description

This function creates different plots depending on the input.

Usage

plot_occurrences(i, spp_name = NULL, pa = TRUE, pa_id = 1, ...)

plot_grid(i, ...)

plot_predictors(i, variables_selected = NULL, ...)

plot_scenarios(i, variables_selected = NULL, scenario = NULL, ...)

plot_predictions(i, spp_name = NULL, scenario = NULL, id = NULL, ...)

plot_ensembles(
  i,
  spp_name = NULL,
  scenario = NULL,
  id = NULL,
  ensemble_type = NULL,
  ...
)

mapview_grid(i)

mapview_occurrences(i, spp_name = NULL, pa = TRUE)

mapview_predictors(i, variables_selected = NULL)

mapview_scenarios(i, variables_selected = NULL, scenario = NULL)

mapview_predictions(i, spp_name = NULL, scenario = NULL, id = NULL)

mapview_ensembles(
  i,
  spp_name = NULL,
  scenario = NULL,
  id = NULL,
  ensemble_type = NULL
)

plot_background(i, variables_selected = NULL, ...)

plot_niche(
  i,
  spp_name = NULL,
  variables_selected = NULL,
  scenario = NULL,
  id = NULL,
  ensemble_type = NULL,
  raster = FALSE,
  ...
)

Arguments

Details

We implemented a bestiary of plots to help visualizing the process and results. If you are not familiar with mapview, consider using it to better visualize maps.

Value

The plot or mapview desired.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

WorldClim_data

Description

This function projects SDM models to new scenarios

Usage

predict_sdm(m,
            scen = NULL,
            metric = "ROC",
            th = 0.9,
            tp = "prob",
            file = NULL,
            add.current = TRUE)

get_predictions(i)

add_predictions(p1, p2)

Arguments

Details

tp is a parameter to be passed on caret to retrieve either the probabilities of classes (tp="prob") or the raw output (tp="raw"), which could vary depending on the algorithm used, but usually would be on of the classes (factor vector with presences and pseudoabsences).

get_predictions returns the list of all predictions to all scenarios, all species, all algorithms and all repetitions. Useful for those who wish to implement their own ensemble methods.

scenarios_names returns the scenarios names in a sdm_area or input_sdm object.

get_scenarios_data returns the data from scenarios in a sdm_area or input_sdm object.

Value

A input_sdm or a predictions object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

sdm_area input_sdm mean_validation_metrics

Examples

if (interactive()) {
  # Create sdm_area object:
  set.seed(1)
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

  # Include scenarios:
  sa <- add_scenarios(sa)

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    repeats = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)
  i
}

Description

Provides an automate way for the visualization of projections gain, loss, and stability between different scenarios.

Usage

prediction_change_sdm(i, scenario = NULL, ensemble_type = NULL, species = NULL, th = 0.5)

Arguments

Value

A plot with comparison between current and other scenario.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

species_names scenarios_names

Examples

if (interactive()) {
  # Create sdm_area object:
  set.seed(1)
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc)

  # Include scenarios:
  sa <- add_scenarios(sa, scen) |> select_predictors(c("bio1", "bio12"))

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i,
    algo = c("naive_bayes"),
    ctrl = ctrl_sdm,
    variables_selected = c("bio1", "bio12")
  ) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)

  # Ensemble:
  i <- ensemble_sdm(i, method = "average")

  # Ensemble GCMs:
  i <- gcms_ensembles(i, gcms = c("ca", "mi"))
  i

  # Change Analysis
  prediction_change_sdm(i, scenario = "_ssp585_2090", ensemble_type = "mean_occ_prob")
}

Description

Print method for ensembles

Usage

## S3 method for class 'ensembles'
print(x, ...)

Arguments

Value

Concatenate structured characters to showcase what is stored in the object.

Description

Print method for input_sdm

Usage

## S3 method for class 'input_sdm'
print(x, ...)

Arguments

Value

Concatenate structured characters to showcase what is stored in the object.

Description

Print method for models

Usage

## S3 method for class 'models'
print(x, ...)

Arguments

Value

Concatenate structured characters to showcase what is stored in the object.

Description

Print method for occurrences

Usage

## S3 method for class 'occurrences'
print(x, ...)

Arguments

Value

Concatenate structured characters to showcase what is stored in the object.

Description

Print method for predictions

Usage

## S3 method for class 'predictions'
print(x, ...)

Arguments

Value

Concatenate structured characters to showcase what is stored in the object.

Description

This function obtains pseudoabsences given a set of predictors.

Usage

pseudoabsences(occ,
               pred = NULL,
               method = "random",
               n_set = 10,
               n_pa = NULL,
               variables_selected = NULL,
               th = 0,
               size = 1,
               size_crs = 4326,
               mcp = FALSE)

n_pseudoabsences(i)

pseudoabsence_method(i)

pseudoabsence_data(i)

Arguments

Details

pseudoabsences is used in the SDM workflow to obtain pseudoabsences, a step necessary for most of the algorithms to run. We implemented four methods: "random", which is self-explanatory, "buffer_sdm", "mahal.dist" and "bioclim". The two last are built with the idea that pseudoabsences should be environmentally different from presences. Thus, we implemented two presence-only methods to infer the distribution of the species. "bioclim" uses an envelope approach (bioclimatic envelope), while "mahal.dist" uses a distance approach (mahalanobis distance). th parameter enters here as a threshold to binarize those results. Pseudoabsences are retrieved outside the projected distribution of the species. If user provides a custom function, it must have the arguments env_sf and occ_sf, which will consist of two "sf"s. The first has the predictor values for the whole study area, while the second has the presence records for the species. The function must return a vector with cell_ids of the pseudoabsences, which is the first column of both objects. For buffer_sdm, user needs to specifiy the size of the buffer compatible with buffer CRS.

n_pseudoabsences returns the number of pseudoabsences obtained per species.

pseudoabsence_method returns the method used to obtain pseudoabsences.

pseudoabsence_data returns a list of species names. Each species name will have a lists with pseudoabsences data from class sf.

Value

A occurrences_sdm or input_sdm object with pseudoabsence data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

link{input_sdm} background occurrences_sdm get_occurrences get_predictors

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ[1:50, ], occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i0 <- pseudoabsences(i, method = "random")

# Custom method example:
buffer_pa_custom <- function(env_sf, occ_sf, buffer_dist = 3) {
  # Create buffer around occurrence points
  buffer <- sf::st_buffer(occ_sf, dist = buffer_dist)

  # Union buffers into a single geometry
  buffer_union <- sf::st_union(buffer)

  # Identify cells outside the buffer
  outside_buffer <- sf::st_difference(env_sf, buffer_union)[, 1]

  # Randomly extract cell_ids outside the buffer
  pa_ids_sample <- sample(outside_buffer$cell_id, nrow(occ_sf))

  return(pa_ids_sample)
}

i1 <- pseudoabsences(i, method = buffer_pa_custom)

Description

A sf LINESTRING object with hydrologic variables (LENGTH_KM and DIST_DN_KM) for the Paraná state in Brazil. Data obtained from HydroSHEDS for river flows >= 10m3/s.

Usage

rivs

Format

## 'rivs' A sf with 1031 attributes and 2 fiels:

LENGTH_KM

Length of the river reach segment, in kilometers.

DIST_DN_KM

Distance from the reach outlet, i.e., the most downstream pixel of the reach, to the final downstream location along the river network, in kilometers. This downstream location is either the pour point into the ocean or an endorheic sink.

Source

<https://www.hydrosheds.org/>

Description

A data.frame object with Salminus brasiliensis occurrence data obtained from GBIF and filtered with Parana state sf.

Usage

salm

Format

## 'salm' A data.frame with 46 rows and 3 columns (EPSG:6933):

species

Species name

decimalLongitude

Longitude in meters

decimalLatitude

Latitude in meters

Source

<https://www.gbif.org>

Description

A stars object with bioclimatic variables (bio1, bio4 and bio12) and four future scenarios for the Parana state in Brazil. Data from MIROC6 GCM from WorldClim 2.1 at 10 arc-min resolution.

Usage

scen

Format

## 'scen' A stars with 4 attribute and 3 bands:

ca_ssp245_2090

Intermediate scenario for the year 2090 and GCM CanESM5

ca_ssp585_2090

Extreme scenario for the year 2090 and GCM CanESM5

mi_ssp245_2090

Intermediate scenario for the year 2090 and GCM MIROC6

mi_ssp585_2090

Extreme scenario for the year 2090 and GCM MIROC6

bio1

Annual Mean Temperature

bio4

Temperature Seasonality

bio12

Annual Precipitation

Source

<https://www.worldclim.org/>

Description

A stars object with bioclimatic variables (bio1, bio4 and bio12) and four future scenarios for the Rio Grande do Sul state in Brazil. Data from MIROC6 GCM from WorldClim 2.1 at 10 arc-min resolution.

Usage

scen_rs

Format

## 'scen_rs' A stars with 5 attribute and 3 bands:

current

Current scenario with the average values for the years 1970-2000

ca_ssp245_2090

Intermediate scenario for the year 2090 and GCM CanESM5

ca_ssp585_2090

Extreme scenario for the year 2090 and GCM CanESM5

mi_ssp245_2090

Intermediate scenario for the year 2090 and GCM MIROC6

mi_ssp585_2090

Extreme scenario for the year 2090 and GCM MIROC6

bio1

Annual Mean Temperature

bio4

Temperature Seasonality

bio12

Annual Precipitation

Source

<https://www.worldclim.org/>

Description

This function creates a new sdm_area object.

Usage

sdm_area(x, cell_size = NULL, output_crs = NULL, variables_selected = NULL,
                gdal = TRUE, crop_by = NULL, lines_as_sdm_area = FALSE, crs = NULL)

get_sdm_area(i)

add_sdm_area(sa1, sa2)

Arguments

Details

The function returns a sdm_area object with a grid built upon the x parameter. There are two ways to make the grid and resample the variables in sdm_area: with and without gdal. As standard, if gdal is available in you machine it will be used (gdal = TRUE), otherwise sf/stars will be used. get_sdm_area will return the grid built by sdm_area. add_sdm_area will sum two sdm_area objects. As geoprocessing in caretSDM is performed using sf objects, add_sdm_area simply applies a rbind in the two different areas.

Value

A sdm_area object containing:

Author(s)

Luíz Fernando Esser ([email protected]) and Reginaldo Ré. https://luizfesser.wordpress.com

See Also

WorldClim_data parana input_sdm, add_predictors

Examples

# Create sdm_area object:
sa_area <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Create sdm_area using a subset of rivs (lines):
sa_rivers <- sdm_area(rivs[c(1:100), ],
                      cell_size = 100000,
                      output_crs = 6933,
                      lines_as_sdm_area = TRUE)

Description

This functions transform data from a caretSDM object to be used in other packages.

Usage

sdm_as_stars(x,
             what = NULL,
             spp = NULL,
             scen = NULL,
             id = NULL,
             ens = NULL)

sdm_as_raster(x, what = NULL, spp = NULL, scen = NULL, id = NULL, ens = NULL)

sdm_as_terra(x, what = NULL, spp = NULL, scen = NULL, id = NULL, ens = NULL)

Arguments

Value

The output is the desired class.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

if (interactive()) {
  # Create sdm_area object:
  sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

  # Include predictors:
  sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

  # Include scenarios:
  sa <- add_scenarios(sa)

  # Create occurrences:
  oc <- occurrences_sdm(occ, occ_crs = 6933)

  # Create input_sdm:
  i <- input_sdm(oc, sa)

  # Pseudoabsence generation:
  i <- pseudoabsences(i, method = "random", n_set = 2)

  # Custom trainControl:
  ctrl_sdm <- caret::trainControl(
    method = "boot",
    number = 1,
    classProbs = TRUE,
    returnResamp = "all",
    summaryFunction = summary_sdm,
    savePredictions = "all"
  )

  # Train models:
  i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
    suppressWarnings()

  # Predict models:
  i <- predict_sdm(i, th = 0.8)

  # Transform in stars:
  sdm_as_stars(i)
}

Description

Set of functions to facilitate the use of caretSDM through tidyverse grammatics.

Usage

select_predictors(x, ...)

## S3 method for class 'sdm_area'
select(.data, ...)

## S3 method for class 'input_sdm'
select(.data, ...)

## S3 method for class 'sdm_area'
mutate(.data, ...)

## S3 method for class 'input_sdm'
mutate(.data, ...)

## S3 method for class 'sdm_area'
filter(.data, ..., .by, .preserve)

## S3 method for class 'input_sdm'
filter(.data, ..., .by, .preserve)

## S3 method for class 'occurrences'
filter(.data, ..., .by, .preserve)

filter_species(x, spp = NULL, ...)

Arguments

Value

The transformed sdm_area/input_sdm object.

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

Description

This function manage predictors names in sdm_area objects.

Usage

get_predictor_names(x)

## S3 method for class 'input_sdm'
set_predictor_names(x, new_names)

## S3 method for class 'sdm_area'
set_predictor_names(x, new_names)

get_predictor_names(x)

test_variables_names(sa, scen)

set_variables_names(s1 = NULL, s2 = NULL, new_names = NULL)

Arguments

Details

This functions is available so users can modify predictors names to better represent them. Use carefully to avoid giving wrong names to the predictors. Useful to make sure the predictors names are equal the names in scenarios. test_variables_names Tests if variables in a stars object (scen argument) matches the given sdm_area object (sa argument). set_variables_names will set s1 object variables names as the s2 object variables names OR assign new names to it.

Value

get_predictor_names returns a character vector with predictors names. test_variables_names returns a logical informing if all variables are equal in both objects (TRUE) or not (FALSE). set_variables_names returns the s1 object with new names provided by s2 or new_names.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

parana sdm_area

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 50000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc)

# Check predictors' names:
get_predictor_names(sa)

Description

This function builds a meta-model (Layer 2) using the out-of-fold predictions from models trained in Layer 1.

Usage

stack_sdm(m, meta_algo = "glm", ctrl = NULL, ...)

Arguments

Value

A stacked_models object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

input_sdm sdm_area algorithms train_sdm

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random")

# Custom trainControl:
ctrl_sdm <- caret::trainControl(method = "repeatedcv",
                                number = 2,
                                repeats = 1,
                                classProbs = TRUE,
                                returnResamp = "all",
                                summaryFunction = summary_sdm,
                                savePredictions = "all")

# Train models:
i <- train_sdm(i, algo = c("naive_bayes", "kknn"), ctrl = ctrl_sdm) |>
suppressWarnings()

# Train stacked ensemble:
i <- stack_sdm(i, meta_algo = "nnet", ctrl = ctrl_sdm)

Description

This function is used in caret::trainControl(summaryFunction=summary_sdm) to calculate performance metrics across resamples.

Usage

summary_sdm(data, lev = NULL, model = NULL, custom_fun=NULL)

summary_sdm_presence_only(data, lev, threshold)

validate_on_independent_data(model, data_independent, obs_col_name)

Arguments

Details

See ?caret::defaultSummary for more details and options to pass on caret::trainControl.

Value

A input_sdm or a predictions object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

train_sdm

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random")

# Custom trainControl:
ctrl_sdm <- caret::trainControl(method = "repeatedcv",
                                number = 2,
                                repeats = 1,
                                classProbs = TRUE,
                                returnResamp = "all",
                                summaryFunction = summary_sdm,
                                savePredictions = "all")

# Train models:
i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
suppressWarnings()

Description

This function is a wrapper to fit models in caret using caretSDM data.

Usage

train_sdm(occ,
          pred = NULL,
          algo,
          ctrl = NULL,
          variables_selected = NULL,
          parallel = FALSE,
          ...)

get_tune_length(i)

algorithms_used(i)

get_models(i)

get_validation_metrics(i)

mean_validation_metrics(i)

models_hyperparameters(i)

add_models(m1, m2)

Arguments

Details

The object algorithms has a table comparing algorithms available. If the function detects that the necessary packages are not available it will ask for installation. This will happen just in the first time you use the algorithm. caret::trainControl holds multiple resources for validation and model tuning. Make sure to understand its parameters beforehand. As it is a key function in the modeling process, we also implemented spatial crossvalidation on it. You can set methods to be cv_spatial or cv_cluster and train_sdm will detect that and apply the method according to blockCV package.

get_tune_length return the length used in grid-search for tunning.

algorithms_used return the names of the algorithms used in the modeling process.

get_models returns a list with trained models (class train) to each species.

get_validation_metrics return a list with a data.frame to each species with complete values for ROC, Sensitivity, Specificity, with their respectives Standard Deviations (SD) and TSS to each of the algorithms and pseudoabsence datasets used.

mean_validation_metrics return a list with a tibble to each species summarizing values for ROC, Sensitivity, Specificity and TSS to each of the algorithms used.

models_hyperparameters returns the hyperparameters that returned the best tuning to each model to each species.

Value

A models or a input_sdm object.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

input_sdm sdm_area algorithms

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random")

# Custom trainControl:
ctrl_sdm <- caret::trainControl(
  method = "repeatedcv",
  number = 2,
  repeats = 1,
  classProbs = TRUE,
  returnResamp = "all",
  summaryFunction = summary_sdm,
  savePredictions = "all"
)

# Train models:
i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
  suppressWarnings()

Description

This function calculates tSNE with presences and pseudoabsences data and returns a list of plots.

Usage

tsne_sdm(occ, pred = NULL, variables_selected = NULL)

Arguments

Value

A list of plots, where each plot is a tSNE for a given pseudoabsence dataset.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Description

This function aims to retrieve the tune grid used to build models.

Usage

tuneGrid_sdm(i)

Arguments

Value

A list with data.frames each one representing the table of a given model.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

# Create sdm_area object:
set.seed(1)
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random", n_set = 2)

# Custom trainControl:
ctrl_sdm <- caret::trainControl(
  method = "boot",
  number = 1,
  repeats = 1,
  classProbs = TRUE,
  returnResamp = "all",
  summaryFunction = summary_sdm,
  savePredictions = "all"
)

# Train models:
i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
  suppressWarnings()

# Retrieve tuneGrid from model:
tuneGrid_sdm(i)

Description

This functions set parameters to run a ESM when running train_sdm.

Usage

use_esm(i, spp = NULL, n_records = 20)

Arguments

Details

We supply two different ways to apply the ESM. If species names are provided, then ESM will be applied only in given species. If a number of species records is provided, then ESM will be applied in every species with number of records bellow the given threshold. As standard, use_esm will be apply to every species with less then 20 records.

Value

A input_sdm or occurrences object with ESM parameters.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Use MEM:
i <- use_esm(i, n_records = 999)

Description

This function sums all species records into one. Should be used before the data cleaning routine.

Usage

use_mem(i, add = TRUE, name = "MEM")

Arguments

Value

A input_sdm or occurrences object with MEM data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Use MEM:
i <- use_mem(i)

Description

This function retrieves variable importance as a function of ROC curves to each predictor.

Usage

varImp_sdm(m, id = NULL, ...)

Arguments

Value

A data.frame with variable importance data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 100000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# Pseudoabsence generation:
i <- pseudoabsences(i, method = "random")

# Custom trainControl:
ctrl_sdm <- caret::trainControl(
  method = "repeatedcv",
  number = 2,
  repeats = 1,
  classProbs = TRUE,
  returnResamp = "all",
  summaryFunction = summary_sdm,
  savePredictions = "all"
)

# Train models:
i <- train_sdm(i, algo = c("naive_bayes"), ctrl = ctrl_sdm) |>
  suppressWarnings()

# Variable importance:
varImp_sdm(i)

Description

Apply Variance Inflation Factor (VIF) calculation.

Usage

vif_predictors(pred, area = "all", th = 0.5, maxobservations = 5000, variables_selected =
NULL)

vif_summary(i)

Arguments

Details

vif_predictors is a wrapper function to run usdm::vifcor in caretSDM.

Value

A input_sdm or predictors object with VIF data.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com

See Also

get_predictor_names

Examples

# Create sdm_area object:
sa <- sdm_area(parana, cell_size = 25000, output_crs = 6933)

# Include predictors:
sa <- add_predictors(sa, bioc) |> select_predictors(c("bio1", "bio4", "bio12"))

# Include scenarios:
sa <- add_scenarios(sa, scen)

# Create occurrences:
oc <- occurrences_sdm(occ, occ_crs = 6933)

# Create input_sdm:
i <- input_sdm(oc, sa)

# VIF calculation:
i <- vif_predictors(i)
i

# Retrieve information about vif:
vif_summary(i)
selected_variables(i)

Description

This function allows to download data from WorldClim v.2.1 (https://www.worldclim.org/data/index.html) considering multiple GCMs, time periods and SSPs.

Usage

WorldClim_data(path = NULL,
               period = "current",
               variable = "bioc",
               year = "2090",
               gcm = "mi",
               ssp = "585",
               resolution = 10)

Arguments

Details

This function will create a folder. All the data downloaded will be stored in this folder. Note that, despite being possible to retrieve a lot of data at once, it is not recommended to do so, since the data is very heavy.

Value

If data is not downloaded, the function downloads the data and has no return value.

Author(s)

Luíz Fernando Esser ([email protected]) [https://luizfesser.wordpress.com](https://luizfesser.wordpress.com)

References

[https://www.worldclim.org/data/index.html](https://www.worldclim.org/data/index.html)

Examples

## download data from multiple periods:
# year <- c("2050", "2090")
# WorldClim_data(path = "",
#               period = "future",
#               variable = "bioc",
#               year = year,
#               gcm = "mi",
#               ssp = "126",
#               resolution = 10)

## download data from one specific period
# WorldClim_data(path = "",
#               period = "future",
#               variable = "bioc",
#               year = "2070",
#               gcm = "mi",
#               ssp = "585",
#               resolution = 10)

Description

This function exports caretSDM data.

Usage

write_ensembles(x, path = NULL, ext = ".tif", centroid = FALSE)

write_predictions(x, path = NULL, ext = ".tif", centroid = FALSE)

write_predictors(x, path = NULL, ext = ".tif", centroid = FALSE)

write_models(x, path = NULL)

write_gpkg(x, file_path, file_name)

## S3 method for class 'sdm_area'
write_gpkg(x, file_path, file_name)

write_occurrences(x, path = NULL, grid = FALSE, ...)

write_pseudoabsences(x, path = NULL, ext = ".csv", centroid = FALSE)

write_background(x, path = NULL, ext = ".csv", centroid = FALSE)

write_grid(x, path = NULL, centroid = FALSE)

write_validation_metrics(x, path = NULL)

Arguments

Details

ext can be set accordingly to the desired output. Possible values are .tif and .asc for rasters, .csv for for a spreadsheet, but also one of: c("bna", "csv", "e00", "gdb", "geojson", "gml", "gmt", "gpkg", "gps", "gtm", "gxt", "jml", "map", "mdb", "nc", "ods", "osm", "pbf", "shp", "sqlite", "vdv", "xls", "xlsx"). path ideally should only provide the folder. We recommend using: results/what_are_you_writting. So for writting ensembles users are advised to run: path = "results/ensembles"

Value

No return value, called for side effects.

Author(s)

Luíz Fernando Esser ([email protected]) https://luizfesser.wordpress.com