In andrefaa/ENM_TheMetaLand: Create and process Ecological Niche, including several pre- and post-processing methods

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  warning=F
)

Basic ENMTML Run

We present here a simple model, with the simplest parameters and decisions.
The main objective of this run is to introduce the main arguments and help with some details which are required to run the function.
We will use the files available at the package example, so that everyone can easily reproduce those steps!

Creating the folder structure from the example files

ENMTML example data set consists of five bioclimantic variables for the current period and five different virtual species.

In order to simulate the files and folders needed for an standard ENMTML run we will create different folders containing the example data.

For this we will use the raster package:

if (!"raster"%in%installed.packages()){install.packages("raster")}  
require(raster)  
require(ENMTML)

# First we will create a folder (ENMTML_example) within the working directory
getwd() #' Working directory of R session
d_ex <- file.path(getwd(), 'ENMTML_example')
d_ex
dir.create(d_ex)

# We will now save ENMTML data set within the ENMTML_example folder

# Virtual species occurrences
data("occ")

# See the format of the .TXT file(tab-separated)! 
# We only need three columns:
  # one containing species name
  # one containing longitude information 
  # one containing latitude information

knitr::kable(occ[c(1:4,43:46,90:94,127:130),])
d_occ <- file.path(d_ex, 'occ.txt')
utils::write.table(occ, d_occ, sep = '\t', row.names = FALSE)

# Five bioclimatic variables for current conditions
data("env")
d_env <- file.path(d_ex, 'current_env_var')
dir.create(d_env)
raster::writeRaster(env, file.path(d_env, names(env)), bylayer=TRUE, 
                    overwrite=TRUE, format='GTiff')

# shell.exec(d_ex) # open the directory and folders created
rm(list = c('env','occ'))

# We now we have the minimum data required to create models with ENMTML package!
# A directory with environmental rasters and a .txt file with occurrence.

The following objects contains the path to the folder with predictors and the path to the occurrence file (TXT)

• d_occ : File path to species occurrence file (.txt; tab-separated)
• d_env : Directory path with current environmental conditions (here in .tiff)

Model design

We will now fit models for with the following specifications:

• Five different algorithms: Bioclim, Generalized Linear Model, Random Forests, Support Vector Machine and Maxent
• Random pseudo-absence allocation
• No accessible area restriction
• Random partition between training-test datasets
• No Ensemble
• No Bias correction
• No Collinearity control in the predictors
• Binary maps will be produced using the Maximum Specificity and Sensitivity Threshold

require(ENMTML)

ENMTML(
 pred_dir = d_env,
 proj_dir = NULL,
 result_dir = file.path(d_ex,"Result"),
 occ_file = d_occ,
 sp = 'species',
 x = 'x',
 y = 'y',
 min_occ = 10,
 thin_occ = NULL,
 eval_occ = NULL,
 colin_var = NULL,
 imp_var = FALSE,
 sp_accessible_area = NULL,
 pseudoabs_method = c(method = 'RND'),
 pres_abs_ratio = 1,
 part=c(method= 'KFOLD', folds='2'),
 save_part = FALSE,
 save_final = TRUE,
 algorithm = c('BIO','GLM', 'RDF', 'SVM', 'MXD'),
 thr = c(type='MAX_TSS'),
 msdm = NULL,
 ensemble = NULL,
 extrapolation = FALSE,
 cores = 1
)

Results

ENMTML function will create a folder named Result within the ENMTML_example folder containing the Algorithms sub-folder and several .txt files:

• Evaluation_Table.txt Contains the results for model evaluation, with several metrics
• InfoModeling.txt Information of the chosen modeling parameters
• Number_Unique_Occurrences.txt Number of unique occurrences for each species
• Occurrences_Cleaned.txt Dataset produced after selecting a single occurrence per grid-cell(unique occurrences)
• Thresholds_Algorithm.txt Information about the thresholds used to create the presence-absence maps for each algorithm (Presence-absence maps are created from the Threshold of complete models)
• Moran & Mess Contains information about autocorrelation and environmental similatiry between the datasets used to fit and evaluate the model

What's next?

We will see next several methodological advances for the Standard model and how those are incorpoated in our modelling routine.
We will also cover projection for other periods, extent and include other algorithms and produce and ensemble model!

Hope you understood the logic behind ENMTML and were able to produce your first models!

See you next time!

Feel free to contact us by mail (you can find the paths to André and Santiago e-mails at the end of the GitHub page)

andrefaa/ENM_TheMetaLand documentation built on Nov. 15, 2023, 10:19 a.m.