ensemble_create: Create ensemble of SDM predictions

In smwoodman/eSDM: Ensemble Tool for Predictions from Species Distribution Models

Create ensemble of SDM predictions

Description

Create a weighted or unweighted ensemble of SDM predictions, including associated uncertainty values

Usage

ensemble_create(x, x.idx, w = NULL, x.var.idx = NULL, ...)

## S3 method for class 'sf'
ensemble_create(x, x.idx, w = NULL, x.var.idx = NULL, ...)

## S3 method for class 'data.frame'
ensemble_create(x, x.idx, w = NULL, x.var.idx = NULL, ...)

Arguments

x	object of class sf or class data.frame
x.idx	vector of column names or numerical indices; indicates which columns in x will be used to create the ensemble
w	weights for the ensemble; either a numeric vector the same length as x or a data frame (or tibble) with the same number of rows as x and ncol(w) == length(x.idx). If w is a numeric vector, its values (i.e. the weights) must sum to 1. The default value is 1 / length(x.idx), i.e. an unweighted ensemble
x.var.idx	vector of column names or column indices; indicates columns in x with variance values with which to calculate uncertainty values for the ensemble. If x.var.idx is specified, it must be the same length as x.idx. Use x.var.idx = NULL (the default) if none of the predictions have associated uncertainty values; in this case the uncertainty values for the ensemble will be calculated using the among-model uncertainty. See the 'Details' section for more information
...	Arguments to be passed to methods; specifically designed for passing na.rm argument to sum

Details

ensemble_create is designed to be used after overlaying predictions with overlay_sdm and (if desired) rescaling the overlaid predictions with ensemble_rescale.

This function implements ensemble methods provided in eSDM_GUI. Note that it does not implement regional exclusion, which must be done manually if not using the GUI.

Ensemble uncertainty is calculated using either the within-model uncertainty (if x.var.idx is specified) or the among-model uncertainty (if x.var.idx is NULL). See the eSDM GUI manual for applicable formulas.

Value

An object of the same class as x with two columns appended to the data frame:

• 'Pred_ens' - The ensemble predictions

• 'Var_ens' - The variance of the ensemble predictions, calculated using either the within-model uncertainty (if x.var.idx is specified) or the among-model uncertainty (if x.var.idx is NULL)

Note that all other columns of x will be included in the returned object. Also, if x is of class sf then 1) the geometry list-column will be the last column of the returned object and 2) the agr attribute will be set as 'constant' for 'Pred_ens' and 'Var_ens'

Examples

ensemble_create(preds.1, c("Density", "Density2"), c(0.2, 0.8))
ensemble_create(preds.1, 1:2, c(0.2, 0.8), c("Var1", "Var2"))
ensemble_create(data.frame(a = 1:5, b = 3:7), c(1, 2))

weights.df <- data.frame(runif(325), c(rep(NA, 100), runif(225)))
ensemble_create(preds.1, c("Density", "Density2"), weights.df, na.rm = TRUE)

smwoodman/eSDM documentation built on Aug. 21, 2024, 5:16 p.m.