Description Usage Arguments Details Value Author(s) References See Also Examples

Species distribution modeling and k-fold cross validation for a set of presence/absence data per species, also considering different background extents (optional). Algorithms supported are "glm", "svm", "maxent", "mars", "rf", "cart.rpart" and "cart.tree"

1 2 3 |

`y` |
Object returned by function |

`x` |
RasterStack ot list of RasterStacks of variables for modeling, a.k.a baseline environment/climatology |

`k` |
Integer. Number of folds for cross validation. Default is 10 |

`algorithm` |
Character string of the algorithms for modeling. Options are the following: "glm", "svm", "maxent", "mars", "rf", "cart.rpart" and "cart.tree" (see details) |

`algorithm.args` |
Further arguments to be passed to the selected algorithm for modeling (functions involved are described in details). |

`weighting` |
Logical for model fitting with weighted presence/absences. Applicable for algorithms "glm", "mars",
"rf", cart.tree and "cart.rpart". Default is FALSE.
The processing time is considerably increased if weighting option is selected when the
"mars" algorithm (see |

`threshold` |
Cut value between 0 and 1 to calculate the confusion matrix. Default is NULL (see Details). |

`diagrams` |
Logical. Only applied if |

`tuneRF.args` |
list of arguments from function |

This function calculates the AUC with the function `auc`

from package
PresenceAbsence. **Note:** Package SDMTools must be detached.

If `threshold`

is not specified the value that maximisez the TSS (true skill statistic) is
used to calculate the confusion matrix.

If `y`

contains data for different background extents (see `backgroundRadius`

and
`pseudoAbsences`

), `mopaTrain`

performs the species distribution modeling for
each different background extent, and fits obtained AUCs (corresponding to different background extents)
to three non linear models (Michaelis-Menten, exponential2 and exponential3).
The model that scores the lowest error is automatically selected to extract the Vm coefficient (equation 1 in
Iturbide et al., 2015). Then, the minimum extent at which the AUC surpasses the Vm value is selected
as the threshold extent (see Figure 3 in Iturbide et al., 2015), being the corresponding fitted SDM the
one returned by `mopaFitting`

. If argument `diagrams`

is set to TRUE, A fitted model plot
(as in Fig. 3 in Iturbide et al., 2015) is printed in the plotting environment.

`mopaTrain`

uses the algorithm implementations of the following functions and R packages:

"mars" function

`earth`

from package earth"rf" function

`ranger`

from package ranger"maxent" function

`maxent`

from package dismo"cart.rpart" function

`rpart`

from package rpart"svm" function

`best.svm`

from package e1071"cart.tree" function

`tree`

from package tree"glm" function

`glm`

from package stats

For example, when appying "glm", further arguments from function `glm`

can be
passed to `mopaTrain`

by using `algorithm.args`

.

A list of six components is returned for each species in `x`

:

`$model`

fitted model using all data for training`$auc`

AUC statistic in the cross validation`$kappa`

kappa statistic in the cross validation`$tss`

true skill statistic in the cross validation`$fold.models`

fitted models of each data partition for cross validation`$ObsPred`

cross model prediction (e.g. for further assessment of model accuracy)

M. Iturbide

Iturbide, M., Bedia, J., Herrera, S., del Hierro, O., Pinto, M., Gutierrez, J.M., 2015. A framework for species distribution modelling with improved pseudo-absence generation. Ecological Modelling. DOI:10.1016/j.ecolmodel.2015.05.018.

`mopaPredict`

, `pseudoAbsences`

, `backgroundGrid`

,
`OCSVMprofiling`

, `backgroundRadius`

, `extractFromModel`

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 | ```
## Load presence data
data(Oak_phylo2)
## Load climate data
destfile <- tempfile()
data.url <- "https://raw.githubusercontent.com/SantanderMetGroup/mopa/master/data/biostack.rda"
download.file(data.url, destfile)
load(destfile, verbose = TRUE)
## Spatial reference
r <- biostack$baseline[[1]]
## Create background grid
bg <- backgroundGrid(r)
## Generate pseudo-absences
RS_random <-pseudoAbsences(xy = Oak_phylo2, background = bg$xy,
exclusion.buffer = 0.083*5, prevalence = -0.5, kmeans = FALSE)
## Model training
fittedRS <- mopaTrain(y = RS_random, x = biostack$baseline,
k = 10, algorithm = "glm", weighting = TRUE)
## Extract fitted models
mods <- extractFromModel(models = fittedRS, value = "model")
``` |

mopa documentation built on May 2, 2019, 6:47 a.m.

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.