Ecological Niche Modelling

A tutorial

Pedro Tarroso (FCUP / BIOPOLIS-CIBIO)

21-05-2024

Contents
Preamble 2

Course folder structure and data 2

Chapter 1 - The presence data housekeeping 4

Remove rows with missing coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Remove rows that fail cleaning tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Remove rows based on additional GBIF information . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Remove duplicated coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Check map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Save the new data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Chapter 2 - Processing the raster data 11

General layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Processing climate layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Process EVI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Chapter 3 - Final optimization of presence data 20

Chapter 4 - Variable selection for modelling 23

Define a training area with a buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Calculate pairwise correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Clustering and selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Chapter 5 - Prepare projection variables 28

Prepare current dataset for projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Prepare future climate projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Chapter 6 - Model building 30

Model building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Repeat for the other two species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Final considerations for the modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

Chapter 7 - Model ensemble 46

Ensembling Vipera aspis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Ensembling Vipera latastei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Ensembling Vipera seoanei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Final considerations for the ensembling 54

1
Chapter 8 - Model projections 54
Projecting Vipera aspis models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Projecting Vipera latastei models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Projecting Vipera latastei models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Final considerations for projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Chapter 9 - Finally, answering the question 58

Conclusion 64

Preamble
This guide will help you build ecological niche models using the R package biomod2. It provides a basic guide
for data preparation, model building, and ensemble.
The purpose of this text is to support the practical classes associated with the Ecological Modelling course
and to serve as a guide for the code in the scripts. The guide’s structure follows the structure of the code
provided in the practical classes, with additional information to help students follow along more easily.
The study system is composed of three related species of European vipers: Vipera aspis, Vipera latastei, and
Vipera seoanei, which meet in the Iberian Peninsula in a narrow three-way contact zone. In this guide, we
will model each of these species and predict the contact zone where the three vipers intersect. Reptiles are
often considered good study systems for ecological niche modeling because they rely on external factors such
as temperature to regulate their internal temperature. Most of our predictor variables will be climate-related,
but we will also include a vegetation-related variable. Since climate variables are available for different time
periods, we will use ecological niche models to answer the following question:
• How will climate change impact the contact zone between the three vipers in Iberia?
Although the code is divided into several scripts for logical clarity, it could also be written as a single long
script. In this document, I have opted to divide the chapters according to each script. For example, script 01
will be discussed in chapter 1, and so on.
Disclaimer: Many methods and processes have been simplified for teaching purposes. While the code
provides a basic and correct modeling approach, it may lack certain details and additional efforts needed to
make the models more robust. These points will be noted where applicable in the text.

NOTE: This is a work in progress. If you find any errors, please comment and send them
ò back to me!

Course folder structure and data

To follow this guide you should have a folder with data and scripts. You should have 9 scripts:
• 01_Presence_Data.R
• 02_ProcessRasters.R
• 03_OptimizePresenceData.R
• 04_StaticalVariableSelection.R
• 05_PrepareProjectionVars.R
• 06_ModelBuiding.R
• 07_ensembleModels.R
• 08_projectAll.R
• 09_prepareFinalMaps.R
You should also have two folders:

2
 • data: where all the needed data is saved and where scripts are generating data
• models: where model data from Biomod will be saved

NOTE: there should be some folders named original within the data. These will have the
ò raw data used for the modelling scripts that is obtained along the tutorial. Do not change
these folders.

The data folder should have data organised in three folders

• data
– other : Accessory data need for some scripts
1. World countries shapefile from NaturalEarth
2. Iberian Peninsula shapefile (processed from above file)
– rasters: Here are the predictor variables to be used in the modelling process and where the
processed rasters will be saved.
∗ original: original raw data (obtained in chapter 2 and 5).
· climate: will store 19 Bioclimatic variables for present and future predictions.
· evi: Enhanced Vegetation Index obtained from the OpenGeoHub Foundation though
OpenLandMap.
– species: Species data as downloaded from GBIF and where processed presence data will be saved
∗ original: Should have 3 zip files (one for each species of viper) as downloaded from GBIF
and the extracted CSV text file with data and correctly named with species name (Vaspis,
Vlatastei, Vseoanei).
The Bioclimatic climate variables are processed from monthly climate data (monthly minimum, mean and
maximum temperatures and monthly precipitation) and coded as follows:
• BIO1: Annual Mean Temperature
• BIO2: Mean Diurnal Range (Mean of monthly (max temp - min temp))
• BIO3: Isothermality (BIO2/BIO7) (×100)
• BIO4: Temperature Seasonality (standard deviation ×100)
• BIO5: Max Temperature of Warmest Month
• BIO6: Min Temperature of Coldest Month
• BIO7: Temperature Annual Range (BIO5-BIO6)
• BIO8: Mean Temperature of Wettest Quarter
• BIO9: Mean Temperature of Driest Quarter
• BIO10: Mean Temperature of Warmest Quarter
• BIO11: Mean Temperature of Coldest Quarter
• BIO12: Annual Precipitation
• BIO13: Precipitation of Wettest Month
• BIO14: Precipitation of Driest Month
• BIO15: Precipitation Seasonality (Coefficient of Variation)
• BIO16: Precipitation of Wettest Quarter
• BIO17: Precipitation of Driest Quarter
• BIO18: Precipitation of Warmest Quarter
• BIO19: Precipitation of Coldest Quarter
The EVI is an index derived from satellite imagery that reflects landscape productivity related to vegetation
content by measuring the greenness. The original data is bi-monthly summarized (resulting in 1 variable each
pair of month), and in this case, it has been further summarized into a single file representing the annual
maximum value achieve for 2020.
Species data were collected from GBIF without any filtering. The doi of the data citation are:
• V. aspis: doi:10.15468/dl.wmv6q6
• V. latastei: doi:10.15468/dl.fa23t9

3
 • V. seoanei: doi:10.15468/dl.cvfxu6
All scripts start by setting the working directory. This is important because all paths in the scripts are
relative to this directory. For example, if you placed your folder “Practical_EcoMod” on the desktop, the
paths would be:
• Windows: C:\Users\Peter\Desktop\Practical_EcoMod
• Mac /Users/Peter/Desktop/Practical_EcoMod
• Linux: /home/peter/Desktop/Practical_EcoMod
therefore the full path to the file data/other/iberia.shp would be:
• Windows: C:\Users\Peter\Desktop\Practical_EcoMod\data\other\iberia.shp
• Mac /Users/Peter/Desktop/Practical_EcoMod/data/other/iberia.shp
• Linux: /home/peter/Desktop/Practical_EcoMod/data/other/iberia.shp
by setting the working directory as:
• Windows (notice the use of double \\ ):
setwd("C:\\Users\\Peter\\Desktop\\Practical_EcoMod")

• Mac :
setwd("/Users/Peter/Desktop/Practical_EcoMod")

• Linux :
setwd("/home/peter/Desktop/Practical_EcoMod")

Once the working directory is set, all files can be referenced relative to this directory. For example, instead
of providing the full path to the file, you can simply use the relative pathdata/other/iberia.shp for R to
locate the file.
Each individual script begins by setting this working directory, but in this document, this step is omitted
after being set once here.

Chapter 1 - The presence data housekeeping

In this chapter, we will open and clean the presence data for each of the three species. Cleaning the data
involves removing erroneous entries such as those with missing coordinates, incorrect coordinates (e.g., outside
the species’ known range), imprecise locations (high uncertainty), and other issues. The quality filters applied
to the presence data will be based on both geographical information and attribute information provided by
GBIF for each presence point.
We need a special package called CoordinateCleaner that provides functions for cleaning presence data. If
you don’t have it installed, you can install it either through the RStudio menu or by running the command:
install.packages("CoordinateCleaner")

NOTE: Besides here, only in Chapter 6 will you be asked to install some packages in this
ò document. However, be aware that if a package is not installed, R will generate an error
indicating that the package cannot be found or is not available. In such cases, install the
necessary package using the instruction provided above (changing the name of the package to
the one you need!).

Packages only need to be installed once. However, to use the functionality of an installed package in an R
session, you must load it with the library() command. This command needs to be executed in every new R
session where you want to use the additional package.

4
open the package
library(CoordinateCleaner)

We need to open the raw GBIF downloaded data into R. Since we downloaded it in the simple text CSV
format, we can use native R functions to open it. If you check the data with a simple text editor, you would
see that it has a header (the first line has column names), and each column is separated by a TAB. We
specify these settings in the arguments of the read.table() function to correctly read the data. Additionally,
we set two other arguments: quote="" and comment.char="". These settings prevent the possible use of
characters such as double quotes (“) or hash symbols (#) in the data, which could potentially break the
reading process.”.
vasp <- read.table("data/species/original/Vaspis.csv", sep="\t", header=TRUE, quote="")
vlat <- read.table("data/species/original/Vlatastei.csv", sep="\t", header=TRUE, quote="")
vseo <- read.table("data/species/original/Vseoanei.csv", sep="\t", header=TRUE, quote="")

We can keep track of the dimensions of each dataset to check how many points we remove in the process.1
dim(vasp)

## [1] 35929 50
dim(vlat)

## [1] 4477 50
dim(vseo)

## [1] 2070 50

Remove rows with missing coordinates

The column names that GBIF uses for coordinates are named decimalLongitude and decimalLatitude. We
need to check these columns, as some of the presence data may be provided without locations.
To retrieve the rows with missing coordinate data (NA), we can create a mask that detects NAs. The function
is.na returns TRUE if the value in the row is NA, and FALSE if it contains data. Mathematical operations
involving numbers and NAs typically result in NA. We leverage this functionality to check whether any of
the longitude or latitude (or both) columns contain NAs.
mask_vasp <- is.na(vasp$decimalLatitude + vasp$decimalLongitude)
vasp <- vasp[!mask_vasp,]

The exclamtion point (!) inverts the logical values: TRUE become FALSE, and FALSE become TRUE. This
is useful because TRUEs are the rows preserved and we want to preserved the non-NA in the filtering with [ ].
We repeat for the other species:
mask_vlat <- is.na(vlat$decimalLatitude + vlat$decimalLongitude)
vlat <- vlat[!mask_vlat,]

mask_vseo <- is.na(vseo$decimalLatitude + vseo$decimalLongitude)

vseo <- vseo[!mask_vseo,]

Just the dimensions again to track how many points we still have after this first filter:
dim(vasp)

## [1] 33813 50
1 InRStudio, you can monitor the Environment panel, usually located at the top right corner of the interface, for updates on
the dataset dimensions.

5
dim(vlat)

## [1] 4152 50
dim(vseo)

## [1] 1862 50
# Course folder structure and da

Remove rows that fail cleaning tests

CoordinateCleaner package offer some data filtering. It provide a series of tests that we might select to flag
with rows pass and do not pass the test. The tests we are going to use are: * capitals: if coordinates are in
the country capital coordinate, it usually indicates lack of precision * centroids: if the coordinate is exactly
at the centroid of a country, than it also indicate lack of precision * equal: if longitude and latitude are equal
is often an error * gbif: this tests if the coordinate refers to the GBIF headquarters. * instituions: some
coordinates might be attributes to the institution providing rather than to the observation location. We
remove those. * zeros: if the coordinates are set to zero, (intersection of equation with prime meridian) are
often wrong.
We set the test in a separate object an then run the cleaning function that will flag each row.
tests <- c("capitals", "centroids", "equal", "gbif", "institutions", "zeros")

flags_vasp <- clean_coordinates(x = vasp, lon = "decimalLongitude", lat = "decimalLatitude",

species = "species", countries = "countryCode",
country_refcol = "iso_a2", tests = tests)

## Testing coordinate validity

## Flagged 0 records.
## Testing equal lat/lon
## Flagged 1 records.
## Testing zero coordinates
## Flagged 1 records.
## Testing country capitals
## Flagged 65 records.
## Testing country centroids
## Flagged 121 records.
## Testing GBIF headquarters, flagging records around Copenhagen
## Flagged 0 records.
## Testing biodiversity institutions
## Flagged 11 records.
## Flagged 192 of 33813 records, EQ = 0.01.
If you check the last 8 columns of object “flags_vasp” you see a TRUE/FALSE columns for each test plus a
summary that has a TRUE if the presence passed all tests. We want to keep those recods only.
vasp <- vasp[flags_vasp$.summary,]

Repeat for the other two species:

6
flags_vlat <- clean_coordinates(x = vlat, lon = "decimalLongitude", lat = "decimalLatitude",
species = "species", countries = "countryCode",
country_refcol = "iso_a2", tests = tests)

## Testing coordinate validity

## Flagged 0 records.
## Testing equal lat/lon
## Flagged 0 records.
## Testing zero coordinates
## Flagged 5 records.
## Testing country capitals
## Flagged 4 records.
## Testing country centroids
## Flagged 0 records.
## Testing GBIF headquarters, flagging records around Copenhagen
## Flagged 0 records.
## Testing biodiversity institutions
## Flagged 0 records.
## Flagged 9 of 4152 records, EQ = 0.
vlat <- vlat[flags_vlat$.summary,]

flags_vseo <- clean_coordinates(x = vseo, lon = "decimalLongitude", lat = "decimalLatitude",

species = "species", countries = "countryCode",
country_refcol = "iso_a2", tests = tests)

## Testing coordinate validity

## Flagged 0 records.
## Testing equal lat/lon
## Flagged 0 records.
## Testing zero coordinates
## Flagged 0 records.
## Testing country capitals
## Flagged 2 records.
## Testing country centroids
## Flagged 22 records.
## Testing GBIF headquarters, flagging records around Copenhagen
## Flagged 0 records.
## Testing biodiversity institutions
## Flagged 0 records.

7
## Flagged 24 of 1862 records, EQ = 0.01.
vseo <- vseo[flags_vseo$.summary,]

Remove rows based on additional GBIF information

GBIF provides uncertainty related to the precision of the coordinate. We can use this information to further
filter our dataset by removing coordinates with very high uncertainty. Out models will be made with a spatial
resolution of 10km (will see in later chapters) so if the coordinate is reported to have an higher uncertainty
we remove it. However, not all rows have uncertainty reported. We have to preserved those rows without
information as well because we don’t know the precision.
# Remove high coordinate uncertainty
u_vasp <- vasp$coordinateUncertaintyInMeters
vasp <- vasp[u_vasp <= 10000 | is.na(u_vasp),]

u_vlat <- vlat$coordinateUncertaintyInMeters

vlat <- vlat[u_vlat <= 10000 | is.na(u_vlat),]

u_vseo <- vseo$coordinateUncertaintyInMeters

vseo <- vseo[u_vseo <= 10000 | is.na(u_vseo),]

We can also remove Fossil records if they are reported by GBIF:

# Remove fossil records
vasp <- vasp[vasp$basisOfRecord != "FOSSIL_SPECIMEN",]
vlat <- vlat[vlat$basisOfRecord != "FOSSIL_SPECIMEN",]
vseo <- vseo[vseo$basisOfRecord != "FOSSIL_SPECIMEN",]

For some records, an individual count is reported. This can be higher than one (for instace, how many times
a species is seen in a camera trap) but it can also be zero, indicating absence. We only keep those that have 1
or more count or if this info is not reported.
# Sometimes there are records of absences. Remove them.
i_vasp <- vasp$individualCount
vasp <- vasp[i_vasp > 0 | is.na(i_vasp) , ]

i_vlat <- vlat$individualCount

vlat <- vlat[i_vlat > 0 | is.na(i_vlat) , ]

i_vseo <- vseo$individualCount

vseo <- vseo[i_vseo > 0 | is.na(i_vseo) , ]

To preserve some quality we remove very old records that might not exist anymore or might have very
imprecise location. Here we arbitrarily chose 1970 to define a date of reliable records.
# Remove very old records
vasp <- vasp[vasp$year > 1970 | is.na(vasp$year), ]
vlat <- vlat[vlat$year > 1970 | is.na(vlat$year), ]
vseo <- vseo[vseo$year > 1970 | is.na(vseo$year), ]

Remove duplicated coordinates

Some of the records might be observed at the same location (e.g. same camera trapping site at different field
work seasons). We want only one record per location, so we can filter out repeated location, keeping just one
(preserving the uniques).

8
# Select only coordinates and remove duplicated coordinates
vasp <- data.frame(species="Vaspis", x=vasp$decimalLongitude, y=vasp$decimalLatitude)
vasp <- unique(vasp)

vlat <- data.frame(species="Vlatastei", x=vlat$decimalLongitude, y=vlat$decimalLatitude)

vlat <- unique(vlat)

vseo <- data.frame(species="Vseoanei", x=vseo$decimalLongitude, y=vseo$decimalLatitude)

vseo <- unique(vseo)

We can check the size again:

dim(vasp)

## [1] 14323 3
dim(vlat)

## [1] 2193 3
dim(vseo)

## [1] 1356 3

Check map
We can merge the three species in the same dataframe to have only one file.
species <- rbind(vasp, vlat, vseo)

and we can plot our points to check in the map if we spot some obvious errors:
sp_factor <- as.factor(species$species)
levels(sp_factor)

## [1] "Vaspis" "Vlatastei" "Vseoanei"

levels(sp_factor) <- c("orange", "red", "blue")
color <- as.character(sp_factor)
plot(species$x, species$y, asp=1, cex=0.3, col=color)

abline(v=-0.5, col='blue')
abline(h=41, col='blue')

abline(v=5, col='orange', lty=2)

abline(h=41, col='orange', lty=2)

9
 50
45
species$y

40
35

−10 0 10 20

species$x
As shown by the plot with the lines, V. seoanei in blue have a few suspicious points at the east of main
distribution and also at the south, while V. aspis has some wrong points at south west of the main distribution.
We can remove those points based on the coordinates:
For V. soanei everything that is at the right of -0.5º longitude OR south of 41º latitude is to remove:
mask <- species$species == "Vseoanei" & ( species$x > -0.5 | species$y < 41)
species <- species[!mask,]

For V aspis we remove everythin that is at left of 5ª longitude AND south of 41º latitude.
mask <- species$species == "Vaspis" & ( species$x < 5 & species$y < 41)
species <- species[!mask,]

Save the new data

We can check the final dimensions:
dim(vasp)

## [1] 14323 3
dim(vlat)

## [1] 2193 3
dim(vseo)

## [1] 1356 3
We set a file name and we write our data set:
filename <- "data/species/speciesPresence_v1.csv"
write.table(species, filename, sep="\t", row.names=FALSE, col.names=TRUE)

We saved as version 1 because we will have to further remove more data to match our model spatial resolution
(Chapter 3).

10
Chapter 2 - Processing the raster data
In this chapter, we are going to obtain and process the predictors in the data/raster folder. In the process
of building an ecological niche model, we need to select and retrieve variables related to the niche of our
species of interest and our specific hypothesis or question. Often, these variables can be found in several
public databases (e.g., WorldClim climate, satellite products) or other sources, but they usually require some
level of processing. Our goal here is to obtain these variable and perform some GIS raster processing inside R
to ensure uniform geographical properties among the raster layers. These steps include:
• Setting the same extent for all layers (aligning the rasters)
• Setting the same resolution (in this case, the study will be performed at a 10km resolution)
• Applying the same No Data mask to all layers
Defining the resolution and extent of the study area requires careful analysis. In this example, we set an
extent that covers the combined distribution of the three species. We opt for a 10km resolution because many
presence data from national atlases and other sources are available at this resolution. This allows us to use
these data in our model (in the previous chapter, we removed points that had an uncertainty higher than
10km). We also choose this resolution for practical reasons. A very high resolution would require much more
processing time and storage space, which may not be feasible within the available time to run this example.
However, the resolution should always be carefully balanced between data availability (both predictors and
presence data), processing limitations, and, most importantly, the requirements of the study system. For
example, it might not make sense to study elephant distributions at a very high spatial resolution.
Note that in this chapter we are limiting the training area, i.e., the area where the models will retrieve
presence and absence data to calibrate and find the optimal statistical solution. Later, in chapter 5, we will
process the variables to focus only on the area of model projections, which is the Iberian Peninsula.
For this chapter, we need to obtain the 19 bioclimatic variables and the EVI variable. All files should be
downloaded to data/rasters/original under respestive folders climate and evi*.
The terra library is essential for niche modeling. It provides functionality to handle spatial data such as
rasters and vectors, as well as perform geoprocessing. This transforms R into a very competent GIS tool.
The library geodata provides access to climate data from WorldClim and other spatial data.
library(terra)
library(geodata)

General layers
Throughout the tutorial, we will define a training area and a projection area for the models. We will explore
each in detail later, but here’s a brief overview: the training area is where we gather data to build the models,
usually covering the species’ distribution range. The projection area is where we aim to generate model
predictions. In this case, the training area includes most of Western and Central Europe and part of North
Africa, aligning with our focal species’ distribution. The projection area is the Iberian Peninsula (Portugal
and Spain), where we will analyze potential contact zones.
Obtain a vector of country polygons with geodata and save it in the appropriate folder.
countries <- world(path = "data/other")

Now we use the information of the countries within the layer to get our projection area. Since with want only
the continental landmass, we have to separate (disaggregate) the polygons with disagg command and keep
the two largest polygons.
iberia <- countries[countries$GID_0 %in% c("PRT", "ESP")]
iberia <- disagg(iberia)
areas <- expanse(iberia)
iberia <- iberia[order(areas, decreasing=TRUE)[1:2]]
writeVector(iberia, "data/other/iberia.shp", overwrite=TRUE)

11
Processing climate layers
First, we are going to obtain the layers from Worldclim through geodata library. These are the 19 bioclimatic
variables (see above for details). The function allow us to set a priori the resolution we want (5 arc minutes
= 0.083333(3) degrees ~ 10km) and the path where to write them.
clim <- worldclim_global(var = 'bio', res = 5, path = 'data/rasters/original')
plot(clim)
wc2.1_5m_bio_1 wc2.1_5m_bio_2 wc2.1_5m_bio_3 wc2.1_5m_bio_4
20 100
20 2000
50

50

50

50
15 80
0 1500
60
−20 10 1000
40
−50

−50

−50

−50
−40 5 500
20
0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

wc2.1_5m_bio_5 wc2.1_5m_bio_6 wc2.1_5m_bio_7 wc2.1_5m_bio_8

40 20 70
60 20
50

50

50

50
30 0
20 50 0
10 −20 40
0 30 −20
−40
−50

−50

−50

−50
−10 20 −40
−20 −60 10 −60
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

wc2.1_5m_bio_9 wc2.1_5m_bio_10 wc2.1_5m_bio_11 wc2.1_5m_bio_12

7000
30 20 6000
20
50

50

50

50
20 0 5000
0 10 4000
0 −20 3000
−20 −10
2000
−50

−50

−50

−50
−20 −40
−40 −30 1000
−60 0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

wc2.1_5m_bio_13 wc2.1_5m_bio_14 wc2.1_5m_bio_15 wc2.1_5m_bio_16

1500 400 200 3500
50

50

50

50

3000
300 150 2500
1000 2000
200 100 1500
−50

−50

−50

−50

500 100 50 1000

500
0 0 0 0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

The original layer names are lengthy and complex. Since we’ll frequently reference these variables by name
throughout the tutorial, we’ll simplify them with easily identifiable names. Given that we’re opening all files
sequentially from bio1 to bio19, we can rename the layers accordingly.
names(clim)

## [1] "wc2.1_5m_bio_1" "wc2.1_5m_bio_2" "wc2.1_5m_bio_3" "wc2.1_5m_bio_4"

## [5] "wc2.1_5m_bio_5" "wc2.1_5m_bio_6" "wc2.1_5m_bio_7" "wc2.1_5m_bio_8"
## [9] "wc2.1_5m_bio_9" "wc2.1_5m_bio_10" "wc2.1_5m_bio_11" "wc2.1_5m_bio_12"
## [13] "wc2.1_5m_bio_13" "wc2.1_5m_bio_14" "wc2.1_5m_bio_15" "wc2.1_5m_bio_16"
## [17] "wc2.1_5m_bio_17" "wc2.1_5m_bio_18" "wc2.1_5m_bio_19"
names(clim) <- paste0("BIO_", 1:19)
plot(clim)

12
 BIO_1 BIO_2 BIO_3 BIO_4
20 100
20 2000
50

50

50

50
15 80
0 1500
60
−20 10 1000
40
−50

−50

−50

−50
−40 5 500
20
0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

BIO_5 BIO_6 BIO_7 BIO_8

40 20 70
60 20
50

50

50

50
30 0
20 50 0
10 −20 40
0 30 −20
−40
−50

−50

−50

−50
−10 20 −40
−20 −60 10 −60
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

BIO_9 BIO_10 BIO_11 BIO_12

7000
30 20 6000
20
50

50

50

50
20 0 5000
0 10 4000
0 −20 3000
−20 −10
2000
−50

−50

−50

−50
−20 −40
−40 −30 1000
−60 0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

BIO_13 BIO_14 BIO_15 BIO_16

1500 400 200 3500
50

50

50

50
3000
300 150 2500
1000 2000
200 100 1500
−50

−50

−50

−50
500 100 50 1000
500
0 0 0 0
−150 −50 50 150 −150 −50 50 150 −150 −50 50 150 −150 −50 50 150

The study area (training area for models) needs to encompass the full distribution of the three species. We
could either check the extreme coordinates of the presence data or use coarse distribution polygons (e.g.,
IUCN distributions) to define the study area. Here, we opt for a simpler approach by defining a rectangle
that we know includes the species’ distributions. This rectangle extends from -11.0 to 27.0 degrees longitude
and from 25 to 55 degrees latitude.
e <- ext(-11, 27, 25, 55)
plot(clim[[1]])
plot(e, add=T)

20
50

0
0

−20
−50

−40

−150 −100 −50 0 50 100 150

13
We can now use the rectangle extent to crop the climate layers.
clim <- crop(clim, e)
plot(clim)

BIO_1 BIO_2 BIO_3 BIO_4

55

55

55

55
25 16 55 900
20 14 50 800
45

45

45

45
12 45 700
15 40 600
10 10
8 35 500
35

35

35

35
5 30 400
0 6 25
4 20 300
25

25

25

25
−5
−10 10 −10 10 −10 10 −10 10

BIO_5 BIO_6 BIO_7 BIO_8

55

55

55

55
10 40
40 35 30
5
45

45

45

45
30 30 20
0
25 10
20 −5
35

35

35

35
20
−10 0
10 15
−10
25

25

25

25
−10 10 −10 10 −10 10 −10 10

BIO_9 BIO_10 BIO_11 BIO_12

55

55

55

55
30 35 15
30 2000
10
45

45

45

45
20 25 1500
20 5
10 0 1000
15
35

35

35

35
0 10 −5 500
5 −10
−10
25

25

25

25
0
−10 10 −10 10 −10 10 −10 10

BIO_13 BIO_14 BIO_15 BIO_16

55

55

55

55
120 120 700
250
100 100 600
200
45

45

45

45
80 80 500
150 60 400
60 300
100
35

35

35

35
40 40 200
50 20 20 100
25

25

25

25
0 0 0 0
−10 10 −10 10 −10 10 −10 10

Process EVI
The Enhanced Vegetation Index (EVI) indicates land productivity and serves here as a continuous variable,
acting as a proxy for habitat. We will use 2020 data available from the OpenGeoHub website, which hosts
a variety of free spatial data useful for ecological niche modeling. The EVI data is global with a 250m
resolution, resulting in large file sizes. We need to manually download the six files for 2020 and save them in
the data/rasters/evi folder. Once processed, these files can be deleted to save space. The links to each file
are provided in the table.

Name Month
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200101_20200228_go_epsg.4326_v20230608
Jan-Feb
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200301_20200430_go_epsg.4326_v20230608
Mar-Apr
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200501_20200630_go_epsg.4326_v20230608
May-Jun
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200701_20200831_go_epsg.4326_v20230608
Jul-Aug
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200901_20201031_go_epsg.4326_v20230608
Sep-Oct
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20201101_20201231_go_epsg.4326_v20230608
Nov-Dec

After downloading to the correct folder, we can open the rasters in R.

evifiles <- list.files("data/rasters/original/evi/", "*tif", full.names=TRUE)
evi <- rast(evifiles)

As they are huge files, to avoid processing unneeded areas, we crop to the study area defined above.

14
 evi <- crop(evi, e)

## |---------|---------|---------|---------|=========================================
plot(evi)

mwm.inpaint_p.90_250m_s_20200101_20200228_go_epsg.4326_v20230608
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200301_20200430_go_epsg.4326_v20230608
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200501_20200630_go_epsg.4326_
1.00
50

50

50
0.80 0.80 0.80
45

45

45
0.60 0.60 0.60
40

40

40
0.40 0.40 0.40
35

35

35
0.20 0.20 0.20
30

30

30
−10 −5 0 5 10 15 20 25 −10 −5 0 5 10 15 20 25 −10 −5 0 5 10 15 20 25

mwm.inpaint_p.90_250m_s_20200701_20200831_go_epsg.4326_v20230608
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20200901_20201031_go_epsg.4326_v20230608
evi_mod13q1.tmwm.inpaint_p.90_250m_s_20201101_20201231_go_epsg.4326_
1.00

0.80
50

50

50
0.80 0.80
45

45

45
0.60
0.60 0.60
40

40

40
0.40 0.40
0.40
35

35

35

0.20 0.20 0.20

30

30

30

−10 −5 0 5 10 15 20 25 −10 −5 0 5 10 15 20 25 −10 −5 0 5 10 15 20 25

We need to further summarize the data into a single file by obtaining the maximum EVI value per pixel for
the year 2020. We can use the app function from terra, which applies a function to summarize all layers on a
pixel-by-pixel basis.
evi <- app(evi, max, na.rm=T)

## |---------|---------|---------|---------|=========================================
plot(evi)

15
50 1.00

0.80
45

0.60
40

0.40
35

0.20
30

−10 0 10 20 We still need to change the resolu-

tion of the EVI to match the climate data. However, the original resolution of the EVI (0.002245788 degrees)
does not allow aggregation by an integer factor. Therefore, we need to use another function to resample the
data to a new resolution. By providing the climate raster to this function, it resamples the EVI to match the
exact spatial properties of the climate dataset. We use the “average” method to the resampling process, so
each group of pixels is summarised by average to the new, larger one.
evi <- resample(evi, clim, "average")
plot(evi)

16
55
0.80
50

0.70

0.60
45

0.50
40

0.40

0.30
35

0.20
30

0.10
25

−10 0 10 20

We can also update the layer name:

names(evi)

## [1] "max"
names(evi) <- "evi"

Although it is not mandatory, it is good practice to have the same No Data mask for all variables used in
modeling. We need to check the No Data areas of each layer and set them to a common area. The climate
layers already share the same No Data mask since they were created using the same process. However, it is
not guaranteed that the EVI has the same mask. Here, any pixel set as No Data in a single layer will be set
as No Data in all layers.
Let’s extract and check the combination of all No Data masks. The app command applies a function (in this
case, the sum function) to all corresponding pixels in all layers. By using is.na(), we get TRUE or FALSE
if the pixel is NA or not in each layer. By summing those values, we will know in how many layers that pixel
is set to NA.
na.clim <- app(is.na(clim), sum)
plot(na.clim)

17
55
0
19
50
45
40
35
30
25

−10 0 10 20

As expected, the pixel is either set to NA in all 19 layers or it has data in all.
Now we sum the EVI No Data to check if there are other pixels with No Data. If so, we will have to set a
common mask for all layers.
na.evi <- is.na(evi)
plot(na.evi)

18
55
False
True
50
45
40
35
30
25

−10 0 10 20

Plot the sum of both masks:

plot(na.clim + na.evi)
55

0
1
19
50

20
45
40
35
30
25

−10 0 10 20

Notice that a few pixels have the value of 1. This means that only the EVI had those pixels set as No Data.

19
We need to set a final mask where we detect pixels set to NA in, at least, one layer and apply the mask to all
rasters processed.
na.mask <- (na.clim + na.evi) > 0
plot(na.mask)
55

False
True
50
45
40
35
30
25

−10 0 10 20

clim[na.mask] <- NA
evi[na.mask] <- NA

We have finished the processing of the raster variables. The climate and EVI data sets are now fully aligned
and with the same resolution. We just need to save them into different TIF files2 .
# Write Rasters to file
# Chelsa is a single file with 19 bands/layers
writeRaster(clim, "data/rasters/climate.tif", overwrite=TRUE)
# EVI is a single file with single band/layer
writeRaster(evi, "data/rasters/evi.tif", overwrite=TRUE)

Note that these files a full raster georeferenced images. The images can be visualised as well in any GIS
(e.g. QGIS).

Chapter 3 - Final optimization of presence data

In this chapter, we will load the presence data we worked on in Chapter 1 and the raster variables we
processed in Chapter 2. We will optimize the presence data by considering the spatial resolution of the
variables.
This ensures that no presence is repeated in the same pixel (also referred to as a cell). Having multiple
presences in the same pixel would mean that we would provide the same set of environmental conditions
repeated as many times as the presences. This can create an artificial bias in the models towards these
2We use the “overwrite=TRUE” just in the case some version of the files were already present and to avoid the error

20
values. By setting a resolution and removing duplicate presences, we also reduce bias towards areas with
more sampling effort (an area with more effort tend to have more duplicates at wider spatial resolutions).
We will only need terra library for this chapter.
library(terra)

We can now open the relevant data for this chapter. Since in the last chapter we fully aligned the raster
variables, we only need to open a single one now as it will provide enough information for this process. The
presence data set is that produced at the end of chapter 1.
evi <- rast("data/rasters/evi.tif")
pres <- read.table("data/species/speciesPresence_v1.csv", sep="\t", header=TRUE)

Now we will extract raster values at each presence location point. For that, we will use the extract function
provided by the terra package. This function will detect in which pixel/cell the presence is located and
extract that information. If a presence is located in a No Data pixel, we will be able to identify it and remove
it. However, here we are also interested in identifying if the pixel is the same or not. We cannot rely on EVI
values for that, as different pixels might have the same value.
We can set the cells=TRUE parameter in the extract function. For each presence, this will provide a unique
pixel/cell identifier. Thus, if two presences are in the same pixel, they will have the same identifier, and we
can keep only one, removing the duplicates.
dt <- extract(evi, pres[,c("x", "y")], cells=TRUE)
head(dt)

## ID evi cell
## 1 1 0.5606292 70845
## 2 2 0.5706229 69470
## 3 3 0.5267091 71757
## 4 4 0.6034736 68556
## 5 5 0.5350373 70382
## 6 6 0.4996728 68569
Now we remove those presences that fall in No Data by checking the evi column of the extracted data table:
mask <- is.na(dt$evi)

# how many presences fall in missing data?

sum(mask)

## [1] 10
dt <- dt[!mask,]
pres <- pres[!mask,]

Now that both the presence and extrated datasets are free of missing data, we need to check for pixel
duplicates. However, there is a detail that adds a bit of complexity to the process. Since we are working with
three species, we have to check for duplicates independently for each species. This is because the three species
can coexist in the same pixel (sympatry), and we do not want to confound this situation as pixel duplicates.
The easiest way to do this is to create a loop over the species so that we can detect duplicates for each species
independently. The code is organized as follows: 1. Check the names of the species to loop over (three
different species in this case) 2. Create a column in the presence data set that will store a value of TRUE is
the presence is duplicated in the pixel (i.e., if there was already another presence before at the same pixel) or
FALSE if the presence is unique or the first record for a given pixel. 3. Loop over species 1. Identify the
rows of presence table referring to the current species in the loop 2. Detect of duplicates only for the current
species 3. Update the duplicated column created in step 2 with the relevant information for the species. 4.
Print in the console how many duplicates were found for each species

21
sps <- unique(pres$species)

pres$duplicated <- NA

for (sp in sps) {

rows <- which(pres$species == sp)
sp.dup <- duplicated(dt$cell[rows])
pres$duplicated[rows] <- sp.dup
print(paste("Species", sp, "has", sum(sp.dup), "duplicates"))
}

## [1] "Species Vaspis has 9349 duplicates"

## [1] "Species Vlatastei has 959 duplicates"
## [1] "Species Vseoanei has 750 duplicates"
We can check the first rows of data to check the table and how many presences were identified in total (each
TRUE is equivalent to 1, and a FALSE to 0)
head(pres)

## species x y duplicated
## 1 Vaspis 2.70 42.05 FALSE
## 2 Vaspis 2.09 42.31 FALSE
## 3 Vaspis 2.70 41.87 FALSE
## 4 Vaspis 1.97 42.49 FALSE
## 5 Vaspis 2.09 42.13 FALSE
## 6 Vaspis 3.06 42.50 FALSE
sum(pres$duplicated)

## [1] 11058
Everything seems coherent, so we proceed to remove the duplicated rows. We have to invert the logical value
as we want to keep the rows that are not duplicated (were set as FALSE). The sub setting with [ ] only keeps
the rows set as TRUE, thus, by inverting with the exclamation point (!) before, we invert and set TRUE to
FALSE, and FALSE to TRUE.
final_pres <- pres[!pres$duplicated, 1:3]

We only kept the first 3 columns (species, name, longitude and latitude) as they are the information we need
to model.
We can get a summary of how may presence data points we kept at the end of presence processing.
dim(final_pres)

## [1] 6796 3
table(final_pres$species)

##
## Vaspis Vlatastei Vseoanei
## 4969 1227 600
Finally, we end this chapter by saving a new file with the optimized data set that will be used for modelling.
# write to file
filename <- "data/species/speciesPresence_v2.csv"
write.table(final_pres, filename, sep="\t", row.names=FALSE, col.names=TRUE)

22
Chapter 4 - Variable selection for modelling
It is common practice to collect all predictor variables available that are deemed relevant for our study system
and hypotheses. These include direct variables (like temperature) and processed variables (like distances to
habitat categories). However, it is rarely a good idea to include all variables into the modeling process.
To address this, we can reduce the number of variables by checking the pairwise correlation among them.
Correlated variables will bring mostly the same information to the model and might add a confounding effect
when analyzing the variable importance in the models. There is no universal maximum correlation value for
modeling, but avoiding correlations higher than 0.7 is generally considered sufficient. We will set this target
for this example.
We will only need terra library for this.
library(terra)

We need the raster variables created in Chapter 2 and the optimised presence data from Chapter 3.
evi <- rast("data/rasters/evi.tif")
clim <- rast("data/rasters/climate.tif")
pres <- read.table("data/species/speciesPresence_v2.csv", sep="\t", header=TRUE)

The rasters are fully aligned (Chapter 2) and we can stack them together. Having the 20 variables (19
bioclimatic + 1 EVI) in the same object will simplify the process of selection.
rst <- c(clim, evi)
names(rst)

## [1] "BIO_1" "BIO_2" "BIO_3" "BIO_4" "BIO_5" "BIO_6" "BIO_7" "BIO_8"

## [9] "BIO_9" "BIO_10" "BIO_11" "BIO_12" "BIO_13" "BIO_14" "BIO_15" "BIO_16"
## [17] "BIO_17" "BIO_18" "BIO_19" "evi"

Define a training area with a buffer

We need to define a training area for the modeling. This restricts the available area from where we retrieve
data for model calibration, including both presence data and pseudo-absence data. There are many ways
to select this area, but we will follow a strategy that defines a spatial buffer around the presence points.
This buffer limits the area from which we can retrieve environmental data from the rasters. A buffer should
provide enough environmental variation for the modeling. However, a very wide buffer can pose problems, as
we might include areas that are far from the species’ optimal conditions. For instance, if we model a species
only present in the Iberian Peninsula, but we choose a buffer large enough to include Iceland, we will be
providing useless environmental variation to the model.
In this chapter, the buffer will define the training area that will provide environmental variation to the model
and where we should check the correlations. Although we have more environmental data outside the buffer,
that data will not be relevant for the model and hence should not be used for correlation analysis.
Here, we define the buffer size as 1 degree, which is roughly equivalent to ~110,000 meters.

NOTE: The buffer defined here will be important for Chapter 6 when we build the models.
ò We will have to define the same buffer size.”

23
 NOTE on spatial data in R: When we open the presence data in R with the read.table
ò command, we just create a data frame (like a spreadsheet in Excel). Although we might have
columns for longitude and latitude, they are just numbers organized in two columns. Since
creating buffers is a spatial operation, we have to inform R that the presences are spatial
points. In other words, we have to formalize that the data frame is, in fact, a list of spatial
points. This can be done with the vect command from the terra package, which is also used
to open spatial data files such as shapefiles. We must define which columns store the longitude
and latitude (in our case, the ‘x’ and ‘y’ columns).

v <- vect(pres, geom= c("x", "y"))

Now we can define our buffers with a 1-degree radius. The buffer function creates an individual buffer
around each presence, and we then aggregate them into a single polygon.
bsize <- 1
buf <- buffer(v, bsize)
buf <- aggregate(buf)

To check how everything is looking, we can plot presence and buffers over the EVI variable.
plot(evi)
plot(buf, cex=0.25, add=T)
plot(v, add=T)
55

0.80
50

0.70

0.60
45

0.50
40

0.40

0.30
35

0.20
30

0.10
25

−10 0 10 20 As we can see in the figure, the

buffer is created around presences. This includes ocean areas which will be ignored later as they provides
only NAs.
We can use the buffer to extract the environmental data for all 20 variables.
dt <- extract(rst, buf, ID=FALSE)
head(dt)

24
## BIO_1 BIO_2 BIO_3 BIO_4 BIO_5 BIO_6 BIO_7 BIO_8 BIO_9 BIO_10 BIO_11 BIO_12
## 1 NA NA NA NA NA NA NA NA NA NA NA NA
## 2 NA NA NA NA NA NA NA NA NA NA NA NA
## 3 NA NA NA NA NA NA NA NA NA NA NA NA
## 4 NA NA NA NA NA NA NA NA NA NA NA NA
## 5 NA NA NA NA NA NA NA NA NA NA NA NA
## 6 NA NA NA NA NA NA NA NA NA NA NA NA
## BIO_13 BIO_14 BIO_15 BIO_16 BIO_17 BIO_18 BIO_19 evi
## 1 NA NA NA NA NA NA NA NA
## 2 NA NA NA NA NA NA NA NA
## 3 NA NA NA NA NA NA NA NA
## 4 NA NA NA NA NA NA NA NA
## 5 NA NA NA NA NA NA NA NA
## 6 NA NA NA NA NA NA NA NA
There are many NAs that come from ocean pixels. We should remove those rows and keep only the valid
data before proceeding with correlations. Since the rasters are fully aligned (Chapter 2), including the No
Data mask, we can use the NAs from a single raster to filter all other variables.
dt <- dt[!is.na(dt[,1]),]
head(dt)

## BIO_1 BIO_2 BIO_3 BIO_4 BIO_5 BIO_6 BIO_7 BIO_8

## 8 9.446667 5.786667 29.61447 509.8863 20.14000 0.6000000 19.54000 10.74000
## 18 9.408896 5.943919 30.25103 508.3113 20.18378 0.5351351 19.64865 10.68108
## 19 9.499875 6.132250 30.78748 512.5685 20.44500 0.5270000 19.91800 10.69300
## 20 9.612041 6.353250 31.41441 520.0265 20.71900 0.4950000 20.22400 10.71300
## 31 9.459243 6.206061 31.10382 510.2022 20.35273 0.4000000 19.95273 10.69364
## 32 9.627584 6.378666 31.63864 514.4630 20.68200 0.5210000 20.16100 10.72333
## BIO_9 BIO_10 BIO_11 BIO_12 BIO_13 BIO_14 BIO_15 BIO_16 BIO_17
## 8 5.166667 15.74667 3.450000 798 99 40 30.06747 287 143
## 18 5.153153 15.66441 3.411712 799 99 41 30.21513 288 143
## 19 5.267500 15.77517 3.431667 803 98 41 29.25621 286 144
## 20 5.363167 15.93750 3.436833 825 98 43 27.84145 288 148
## 31 11.040606 15.70848 3.421818 818 101 42 30.15000 293 146
## 32 5.438000 15.90683 3.519333 813 101 41 29.37465 289 147
## BIO_18 BIO_19 evi
## 8 212 192 0.5183802
## 18 211 192 0.4421933
## 19 214 192 0.6296464
## 20 190 195 0.5911894
## 31 213 198 0.4945980
## 32 182 195 0.6150222
Now we have only data! We can proceed with Pearson correlations.

Calculate pairwise correlations

Getting the Pearson’s correlation scores for each pair of variables is very easy in R. We have all raster data
for the training area (buffer) organized in a data frame where each column is a different variable. We just
calculate correlations based on that table.
corr <- cor(dt)
# Rounding to 3 decimal places to facilitate visualization.
round(corr, 3)

## BIO_1 BIO_2 BIO_3 BIO_4 BIO_5 BIO_6 BIO_7 BIO_8 BIO_9 BIO_10

25
## BIO_1 1.000 0.429 0.516 -0.087 0.870 0.896 0.217 0.410 0.737 0.956
## BIO_2 0.429 1.000 0.675 0.433 0.767 0.073 0.855 0.129 0.422 0.566
## BIO_3 0.516 0.675 1.000 -0.345 0.520 0.458 0.202 0.020 0.538 0.420
## BIO_4 -0.087 0.433 -0.345 1.000 0.325 -0.476 0.831 0.201 -0.181 0.207
## BIO_5 0.870 0.767 0.520 0.325 1.000 0.585 0.661 0.360 0.676 0.957
## BIO_6 0.896 0.073 0.458 -0.476 0.585 1.000 -0.222 0.275 0.690 0.737
## BIO_7 0.217 0.855 0.202 0.831 0.661 -0.222 1.000 0.178 0.174 0.468
## BIO_8 0.410 0.129 0.020 0.201 0.360 0.275 0.178 1.000 -0.046 0.448
## BIO_9 0.737 0.422 0.538 -0.181 0.676 0.690 0.174 -0.046 1.000 0.685
## BIO_10 0.956 0.566 0.420 0.207 0.957 0.737 0.468 0.448 0.685 1.000
## BIO_11 0.962 0.304 0.587 -0.349 0.739 0.967 -0.006 0.314 0.764 0.844
## BIO_12 -0.514 -0.489 -0.263 -0.240 -0.630 -0.340 -0.443 -0.319 -0.382 -0.583
## BIO_13 -0.242 -0.348 -0.096 -0.292 -0.387 -0.100 -0.373 -0.219 -0.131 -0.326
## BIO_14 -0.734 -0.544 -0.472 -0.034 -0.766 -0.585 -0.380 -0.256 -0.678 -0.746
## BIO_15 0.627 0.385 0.431 -0.118 0.583 0.539 0.201 0.134 0.609 0.595
## BIO_16 -0.272 -0.375 -0.101 -0.322 -0.425 -0.112 -0.407 -0.251 -0.155 -0.365
## BIO_17 -0.719 -0.530 -0.468 -0.019 -0.747 -0.582 -0.359 -0.270 -0.636 -0.725
## BIO_18 -0.751 -0.511 -0.512 0.059 -0.760 -0.658 -0.304 -0.069 -0.777 -0.736
## BIO_19 -0.100 -0.301 0.103 -0.497 -0.293 0.114 -0.458 -0.426 0.097 -0.242
## evi -0.230 -0.487 -0.269 -0.193 -0.392 -0.039 -0.435 -0.055 -0.344 -0.302
## BIO_11 BIO_12 BIO_13 BIO_14 BIO_15 BIO_16 BIO_17 BIO_18 BIO_19 evi
## BIO_1 0.962 -0.514 -0.242 -0.734 0.627 -0.272 -0.719 -0.751 -0.100 -0.230
## BIO_2 0.304 -0.489 -0.348 -0.544 0.385 -0.375 -0.530 -0.511 -0.301 -0.487
## BIO_3 0.587 -0.263 -0.096 -0.472 0.431 -0.101 -0.468 -0.512 0.103 -0.269
## BIO_4 -0.349 -0.240 -0.292 -0.034 -0.118 -0.322 -0.019 0.059 -0.497 -0.193
## BIO_5 0.739 -0.630 -0.387 -0.766 0.583 -0.425 -0.747 -0.760 -0.293 -0.392
## BIO_6 0.967 -0.340 -0.100 -0.585 0.539 -0.112 -0.582 -0.658 0.114 -0.039
## BIO_7 -0.006 -0.443 -0.373 -0.380 0.201 -0.407 -0.359 -0.304 -0.458 -0.435
## BIO_8 0.314 -0.319 -0.219 -0.256 0.134 -0.251 -0.270 -0.069 -0.426 -0.055
## BIO_9 0.764 -0.382 -0.131 -0.678 0.609 -0.155 -0.636 -0.777 0.097 -0.344
## BIO_10 0.844 -0.583 -0.326 -0.746 0.595 -0.365 -0.725 -0.736 -0.242 -0.302
## BIO_11 1.000 -0.430 -0.154 -0.701 0.640 -0.174 -0.689 -0.742 0.039 -0.192
## BIO_12 -0.430 1.000 0.899 0.714 -0.251 0.915 0.773 0.741 0.784 0.369
## BIO_13 -0.154 0.899 1.000 0.371 0.151 0.990 0.453 0.480 0.864 0.182
## BIO_14 -0.701 0.714 0.371 1.000 -0.800 0.396 0.984 0.891 0.255 0.541
## BIO_15 0.640 -0.251 0.151 -0.800 1.000 0.133 -0.766 -0.617 0.186 -0.501
## BIO_16 -0.174 0.915 0.990 0.396 0.133 1.000 0.468 0.502 0.885 0.196
## BIO_17 -0.689 0.773 0.453 0.984 -0.766 0.468 1.000 0.891 0.318 0.515
## BIO_18 -0.742 0.741 0.480 0.891 -0.617 0.502 0.891 1.000 0.204 0.423
## BIO_19 0.039 0.784 0.864 0.255 0.186 0.885 0.318 0.204 1.000 0.177
## evi -0.192 0.369 0.182 0.541 -0.501 0.196 0.515 0.423 0.177 1.000
The correlation matrix provides the correlation score for each pair of variables. It is symmetric (the correlation
of A to B is the same as B to A) and the diagonal gives the self-correlation, which is obviously 1.
This matrix is not easy to use for the variable elimination process. Remember that we set a goal of not
having an absolute correlation higher than 0.7, so we should find a way to select important variables that will
provide a new set where the maximum pairwise correlation won’t exceed that value.
One of the best and simplest ways is to organize the variables in a dendrogram by using a hierarchical
clustering method. We should group in the same cluster variables that are similar to each other (highly
correlated). To do this, we need to transform the correlation score into a dissimilarity or distance matrix,
where increasing values will reflect that variables are less similar (low correlation). For that we need to:
1. Get the absolute value of the correlation because we just care about the magnitude of the correlation
and not the direction (two variables that are correlated at -0.87 provide the same information as those

26
 that are correlated at +0.87). At this point, the matrix will have values from 0 (low correlation, thus
highly distant) to 1 (high correlation, showing high similarity and low distance).
2. Ensure that lower values reflect similarity (high correlation) and higher values describe dissimilarity
(low correlation). To achieve this, we need to invert the values by subtracting the absolute value from 1.
# Convert correlation to distance
dist <- 1 - abs(corr)

We need to discard one of the symmetrical sides and the diagonal. An easy way to do this is to convert the
matrix to a dist object:
dist <- as.dist(dist)

Clustering and selection

With the new distance matrix we can start the clustering process. We choose the method ‘single’ so every
cluster is grouped by the minimum distance/correlation.
# Hierarchical cluster of distance
hc <- hclust(dist, method="single")
plot(hc, hang=-1)

Cluster Dendrogram
0.4
Height

0.2
0.0

BIO_8
evi
BIO_3
BIO_19
BIO_12
BIO_13
BIO_16
BIO_9
BIO_15
BIO_18
BIO_14
BIO_17
BIO_1
BIO_6
BIO_11
BIO_5
BIO_10
BIO_4
BIO_2
BIO_7

dist
hclust (*, "single")
We can select 5 variables, one from each major group in the tree. Variables can be selected based on how
easily they can be interpreted in terms of the niche of the species. In this case, we select:
• evi
• BIO_8: Mean daily Temperature Wettest Quarter
• BIO_3: Isothermality
• BIO_12: Annual Precipitation Amount
• BIO_1: Annual Temperature

27
We select the vars using the variables names we defined previously and create a new raster with only those
variables.
sel_vars <- c("evi", "BIO_8", "BIO_3", "BIO_12", "BIO_1")
sel_rst <- rst[[sel_vars]]

We select the variables using the names we defined previously and create a new raster with only those
variables.
# Check correlation of final dataset
sel_dt <- dt[,sel_vars]
cor(sel_dt)

## evi BIO_8 BIO_3 BIO_12 BIO_1

## evi 1.00000000 -0.05473548 -0.26896220 0.3687846 -0.2295176
## BIO_8 -0.05473548 1.00000000 0.01993914 -0.3192758 0.4101052
## BIO_3 -0.26896220 0.01993914 1.00000000 -0.2634080 0.5157017
## BIO_12 0.36878463 -0.31927583 -0.26340797 1.0000000 -0.5142887
## BIO_1 -0.22951757 0.41010518 0.51570173 -0.5142887 1.0000000
The maximum correlation does not pass the value of 0.51, which is generally acceptable.
We save the selected variables to a file named final_vars.tif. You can check the variables in any GIS.
# Write Final raster to model
writeRaster(sel_rst, "data/rasters/final_vars.tif", overwrite=TRUE)

Chapter 5 - Prepare projection variables

Our initial question is about the contact zone between three species of vipers in the Iberian Peninsula.
Although the original distributions of two of the species, Vipera aspis and Vipera latastei extend beyond
the Iberian Peninsula, we don’t need the final projections to cover the full extent of their distributions. To
answer our question, we only need predictions for the Iberian Peninsula.
In this chapter, we will prepare the variables for model projection. This includes current climate variables and
NDVI, as well as future climate predictions for the same selected variables used in model training (Chapter
4). The projection area is defined as the Iberian Peninsula. We will crop and mask the variables to show
data only for the Iberian Peninsula.
A mandatory detail is that all groups of projection variables must have the same layer names as those used in
model training so that the models can find the respective variables to produce predictions.
The names we have are:
• evi
• BIO_8
• BIO_3
• BIO_12
• BIO_1
Keep in mind that NDVI is a static variable in this scheme, meaning we don’t have future projections of
change for NDVI, so we use the same variable for all datasets. This approach is not usually done, as NDVI
is very susceptible to changes due to climate change and human activity. Therefore, we are assuming that
the habitat, for which we use NDVI as a surrogate, is stable over time. Variables like elevation, or derived
variables like slope, are more easily assumed to be static.
Again, we only need terra library for this chapter, and geodata to download future climate layers at the
end of the chapter.

28
library(terra)
library(geodata)

To crop and cut the rasters, we use the Iberian Peninsula shapefile
# Open Iberia shape file
v <- vect("data/other/iberia.shp")

Prepare current dataset for projection

We open the current data set that we prepare in chapter 4.
# Open current data rasters
vars <- rast("data/rasters/final_vars.tif")

We use the Iberian Peninsula shapefile to crop and mask the rasters. Cropping means reducing the extent,
while masking is used to define No Data for all pixels outside the Iberian Peninsula.
# Cut Current data and save
vars <- crop(vars, v, mask=TRUE)

The current dataset for projection is ready, and we can write it to a file. All projection files will be saved in
the data/rasters folder, and filenames will start with a “proj_” prefix.
writeRaster(vars, "data/rasters/proj_current.tif", overwrite=TRUE)

Prepare future climate projections

Climate projections add extra complexity because there are many variants. As it is difficult to predict future
climate, Global Circulation Models (GCMs) are given different scenarios for possible paths of societal behavior
towards greenhouse gas production. These scenarios are called the Shared Socioeconomic Pathways (SSPs)
and range from maintaining the current trend of emissions (worst-case scenario) to maintaining and even
reducing emissions (best-case scenario). This allows us to predict climate over a range of conditions that we
should consider in our niche model projections. Therefore, we can project niche models to different GCMs,
different SSPs, and different time periods.
In this tutoral we will project the models to * One single GCM (mpi-esm1-2-hr) * Two SSPS (ssp126 and
ssp585) * Three time periods (2011-2040, 2041-2070, 2071-2100)
This result in a combination of 1x2x3=6 model projections.
All raw future climate rasters will be saved in the folder data/rasters/original/climate.
The easiest way to process all data is to loop over AGES and SSPs to produce all combinations of projections.
This will allow to crop and mask all needed projections variables but it requires a nested for loop.
First we can set some objects with needed information:
ages <- c("2021-2040", "2041-2060", "2061-2080")
scenarios <- c("126", "585")

# bioclimatic in the same order as in vars to simplify

bios <- c("BIO_8", "BIO_3", "BIO_12", "BIO_1")

We can now loop over ages and ssp scenarios. The process inside the nested loop is:
1. DowNload the dataset for the combination of GCM model + age + ssp (~10km spatial resolution) using
geodata and writing to data/rasters/original.
2. Crop the rasters to the study area (same extent as vars).
3. Rename layers to follow our name convention.

29
 4. Select only the variables of interest from the 19 bioclimate variables.
5. Join the evi (static) to the set of selected bioclimate variables
6. Write the produced raster with the 4 layers in the data/rasters/ folder and following a filename
convention that starts by proj_age_ssp in TIF format.
for (age in ages) {
for (ssp in scenarios) {
# download and rename
clim_fut <- cmip6_world(model='MPI-ESM1-2-HR', ssp=ssp, time=age, var='bioc', res=5, path='data/
clim_fut <- crop(clim_fut, vars[[1]], mask=TRUE)
names(clim_fut) <- paste0("BIO_", 1:19)
#filter the selected vars
clim_fut <- clim_fut[[bios]]
# merge with evi
fut <- c(vars[["evi"]], clim_fut)
# Write raster to file
writeRaster(fut, paste0("data/rasters/proj_", age, "_", ssp, ".tif"), overwrite=TRUE)
}
}

At the end of this loop, we should have 6 projection rasters for future time periods plus 1 projection for
current period.

Chapter 6 - Model building

Until now, we have been preparing data for Ecological Niche Modelling (ENM). These included_
1. Clean presence locations
2. Aligned and selected environmental variables for training
3. Cropped projection rasters for present and future climates
At this point, we are ready to produce the first models. We will be using a package named biomod2 that
provides functionality for modeling, ensembling multiple models, and retrieving model projections.
Building a single model is straightforward. However, if we want to vary some initial conditions (resampling,
pseudo-absence selection, etc.) and use multiple modeling algorithms (GLM, GAM, CART, Maxent, etc.), we
need to produce a large number of models and predictions. The biomod2 library facilitates this process by
internally tracking all model variants and providing us with a simpler interface to control crucial parameters
of the modeling process that require our intervention.
The purpose of this chapter is to build an ENM with:
• 3 sets of pseudo-absences drawn randomly from the buffer area around presences
• 4 modeling algorithms including Generalised Linear Models (GLM), General Additive Models (GAM),
Maximum Entropy (Maxnet), and eXtreme Gradient Boosting Training (XGBOOST)
• 5 replicates using K-Fold resampling strategy.
The combination results in 3x4x5 = 60 build models. Besides those, we will have to project models to present
conditions and 6 future climate layers from a combination of ages and SSPs (covered in chapter 8).
“We need the terra and biomod2 packages. While terra provides GIS capabilities, biomod2 offers the tools to
build models. If biomod2 is not available, it requires installation. Additionally, biomod2 depends on other
libraries, including dismo, xgboost, maxnet, gam, and tidyterra. We don’t need to interact directly with these
packages, but biomod2 requires them. We can install these packages either via the RStudio menu or by
executing the following code:
install.packages("biomod2")
install.packages("dismo")

30
install.packages("xgboost")
install.packages("maxnet")
install.packages("gam")
install.packages("tidyterra")

And we can open the need libraries:

library(terra)
library(biomod2)

Open the relevant data for modelling.

pres <- read.table("data/species/speciesPresence_v2.csv", sep="\t", header=TRUE)
vars <- rast("data/rasters/final_vars.tif")

Recall that the presence table has data for the 3 species and that the selected variables are in a single raster
file.

Model building
In this chapter, we will build models for three species. We’ll start with a detailed walkthrough for Vipera
aspis, and then apply the same process to the other two species. All models will be saved in the models
directory, which should be empty at the start. A folder for each species, organized by biomod2, will be created
automatically.
The first step is to provide data into a formatting functions that will prepare the data in a way that biomod2
can understand and track parameters.
We need a response variable which is a vector of 1s with the same size as presence locations for the species.
This is because we don’t have absences (zeros) and we are using a presence-only model strategy by creating
pseudo-absences.
We also need a table of coordinates for each presence which is present in our presence data file.
Finally we need our explanatory variables or predictors that we prepared before.
# How many presences for Vaspis?
n_va <- sum(pres$species == "Vaspis")

# repeat 1 to the same length of presences available

resp_va <- rep(1, n_va)

# get the coordinates for Vaspis

coords_va <- pres[pres$species == "Vaspis", 2:3]

In the formatting data, we also provide some other modeling parameters. We need to define the strategy for
selecting pseudo-absences, which is based on a buffer around presences. We use the same buffer as before
(Chapter 2), which is 110,000 km. We ask for three different sets of 10,000 pseudo-absences followinf the disk
strategy (same as buffer), which will allow us to capture some variation in the initial conditions for each
model.
The dir.name sets the folder where models are to be saved and the resp.name sets a name for the project and
folder to be created inside the models folder.
vaData <- BIOMOD_FormatingData(resp.var = resp_va,
expl.var = vars,
resp.xy = coords_va,
resp.name = "Vaspis",
dir.name = "models",

31
 PA.nb.rep = 3,
PA.nb.absences = 10000,
PA.strategy = "disk",
PA.dist.max = 110000)

##
## -=-=-=-=-=-=-=-=-=-=-=-=-=-= Vaspis Data Formating -=-=-=-=-=-=-=-=-=-=-=-=-=-=
##
## ! No data has been set aside for modeling evaluation
## ! No data has been set aside for modeling evaluation
##
## Checking Pseudo-absence selection arguments...
##
## ! No data has been set aside for modeling evaluation
## > Disk pseudo absences selection
## > Pseudo absences are selected in explanatory variables|---------|---------|---------|---------|==
## > random pseudo absences selection
## > Pseudo absences are selected in explanatory variables
##
## ! No data has been set aside for modeling evaluation
## ! No data has been set aside for modeling evaluation
## -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Done -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
We have the data formatted, so we can proceed to the modeling step. Here, we need to define the model
algorithms to run and the train/test strategy. This allows us to control overfitting by providing examples
to train the model and a set of independent examples to test its predictive ability. We choose K-fold
cross-validation for this purpose. In this case, we divide the dataset into five equal-sized folds and test four
against one for each fold.
As models should be evaluated for performance, we choose two common techniques for this: TSS (True Skill
Statistic) and ROC (Receiver Operating Characteristic)
There are many algorithms available for modelling. We are limiting to 4 different algorithms which are
generally fast to train and still provide a good example of the ensemble modelling strategy. We choose:
• Generalized Linear Model (GLM): A general regression model using a binomial family of distributions
(logistic regression)
• Generalized Additive Models (GAM): A regression-based approach where smoothing functions are
applied to each predictor.
• Maximum Entropy (Maxnet): A machine learning method sharing some similarities with GAM and is
an R version of the common MaxEnt approach.
• Extreme Gradient Boosting (XGBOOST): A machine learning method related to decision trees, known
to be fast and perform well.
Each of these algorithms has specific parameters to tune. The biomod2 library offers different strategies
for tuning them. We are choosing the ‘bigboss’ strategy, which consists of a list of the best parameters as
defined by the package authors. This means that the authors tested several algorithms under different data
and provided a list of options for each algorithm that generally performs well in most common situations.
Variable importance is a measure of how much each predictor influences each model. In the context of
Ecological Niche Modeling (ENM), it detects the variable most determining the distribution of the species.
This is done via permutations, set in var.import. We set this to 3, which is a low number of permutations,
but since they take some time to run, we need to keep this number low for this example.
vaModel <- BIOMOD_Modeling(bm.format = vaData,
modeling.id = "EcoMod",
models = c("GAM", "GLM", "MAXNET", "XGBOOST"),

32
 CV.strategy = "kfold",
CV.k = 5,
CV.do.full.models = FALSE,
OPT.strategy = "bigboss",
var.import = 3,
metric.eval = c("TSS", "ROC"))
vaModel

##
## -=-=-=-=-=-=-=-=-=-=-=-=-=-=-= BIOMOD.models.out -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
##
## Modeling folder : models
##
## Species modeled : Vaspis
##
## Modeling id : EcoMod
##
## Considered variables : evi BIO_8 BIO_3 BIO_12 BIO_1
##
##
## Computed Models : Vaspis_PA1_RUN1_GAM Vaspis_PA1_RUN1_GLM
## Vaspis_PA1_RUN1_MAXNET Vaspis_PA1_RUN1_XGBOOST Vaspis_PA1_RUN2_GAM
## Vaspis_PA1_RUN2_GLM Vaspis_PA1_RUN2_MAXNET Vaspis_PA1_RUN2_XGBOOST
## Vaspis_PA1_RUN3_GAM Vaspis_PA1_RUN3_GLM Vaspis_PA1_RUN3_MAXNET
## Vaspis_PA1_RUN3_XGBOOST Vaspis_PA1_RUN4_GAM Vaspis_PA1_RUN4_GLM
## Vaspis_PA1_RUN4_MAXNET Vaspis_PA1_RUN4_XGBOOST Vaspis_PA1_RUN5_GAM
## Vaspis_PA1_RUN5_GLM Vaspis_PA1_RUN5_MAXNET Vaspis_PA1_RUN5_XGBOOST
## Vaspis_PA2_RUN1_GAM Vaspis_PA2_RUN1_GLM Vaspis_PA2_RUN1_MAXNET
## Vaspis_PA2_RUN1_XGBOOST Vaspis_PA2_RUN2_GAM Vaspis_PA2_RUN2_GLM
## Vaspis_PA2_RUN2_MAXNET Vaspis_PA2_RUN2_XGBOOST Vaspis_PA2_RUN3_GAM
## Vaspis_PA2_RUN3_GLM Vaspis_PA2_RUN3_MAXNET Vaspis_PA2_RUN3_XGBOOST
## Vaspis_PA2_RUN4_GAM Vaspis_PA2_RUN4_GLM Vaspis_PA2_RUN4_MAXNET
## Vaspis_PA2_RUN4_XGBOOST Vaspis_PA2_RUN5_GAM Vaspis_PA2_RUN5_GLM
## Vaspis_PA2_RUN5_MAXNET Vaspis_PA2_RUN5_XGBOOST Vaspis_PA3_RUN1_GAM
## Vaspis_PA3_RUN1_GLM Vaspis_PA3_RUN1_MAXNET Vaspis_PA3_RUN1_XGBOOST
## Vaspis_PA3_RUN2_GAM Vaspis_PA3_RUN2_GLM Vaspis_PA3_RUN2_MAXNET
## Vaspis_PA3_RUN2_XGBOOST Vaspis_PA3_RUN3_GAM Vaspis_PA3_RUN3_GLM
## Vaspis_PA3_RUN3_MAXNET Vaspis_PA3_RUN3_XGBOOST Vaspis_PA3_RUN4_GAM
## Vaspis_PA3_RUN4_GLM Vaspis_PA3_RUN4_MAXNET Vaspis_PA3_RUN4_XGBOOST
## Vaspis_PA3_RUN5_GAM Vaspis_PA3_RUN5_GLM Vaspis_PA3_RUN5_MAXNET
## Vaspis_PA3_RUN5_XGBOOST
##
##
## Failed Models : none
##
## -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
This command will take some time to run as it needs to build all the combinations of models we specified. It
displays general information on the process, allowing us to check if any models are failing. If models do fail,
we may need to investigate the reasons and adjust some parameters. If everything runs successfully, we can
plot the results to evaluate the models’ performance.

33
plt <- bm_PlotEvalMean(vaModel, dataset="calibration")

0.45

0.40

GAM
GLM
TSS

0.35 MAXNET
XGBOOST

0.30

0.25

0.69 0.72 0.75 0.78

ROC
plt <- bm_PlotEvalMean(vaModel, dataset="validation")

34
 0.4

GAM
GLM
TSS

MAXNET
XGBOOST

0.3

0.69 0.72 0.75 0.78

ROC

NOTE: We are assigning the results of the functions to a plt object that gets replaced with
ò each plot. Since the functions used to plot also return the data used to build the plots, assigning
the results to an object prevents displaying the entire dataset in the console. Although we are
not using this data here, you can check the contents of plt at any time!

These plots show the model performance for the calibration and validation sets. Remember that we used the
K-fold strategy with K=5K=5 folds. This means that all five combinations of four folds against one fold were
used for calibration (training) and validation, respectively.
The plots show the TSS (True Skill Statistic) and the ROC/AUC (Receiver Operating Characteristic/Area
Under Curve) values. The closer these values are to 1, the better the model performance. Of course, different
algorithms, different sets of pseudo-absences, and different sets of training data (folds) will provide slight
variations in performance. Generally, the models we built are acceptable, as indicated by ROC/AUC values
higher than 0.65 and TSS greater than 0.25. In general, XGBoost and Maxnet, both machine learning
methods, outperform the simpler regression-based methods.
Both calibration and validation show acceptable performance, indicating good model fitting, predictive ability,
and a low level of overfitting.
We can compare together the different runs and algorithms:
plt <- bm_PlotEvalBoxplot(vaModel, group.by = c('algo', 'run'))

35
 ROC TSS
0.8

0.6 RUN1
calibration

RUN2
RUN3
RUN4
RUN5
0.4
AM

LM

AM
ET

ST

LM

ET

ST
N

O
G

G
G

G
AX

AX
BO

BO
M

M
XG

XG
An important step in the modeling results analysis is to understand how much each predictor variable is
contributing to each model.
plt <- bm_PlotVarImpBoxplot(vaModel, group.by = c('expl.var', 'run', 'algo'))

36
 GAM GLM

50%
40%
30%
20%
10%
0% RUN1
RUN2
_1

_3

_8

_1

i
2

_3

_8
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI
RUN3
BI

BI
MAXNET XGBOOST RUN4
RUN5
50%
40%
30%
20%
10%
0%
_1

i
2

_3

_8

_1

_3

_8

i
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI
BI

BI

In general, it is shown that EVI is a low contributor for this species, while climate variables are generally
more associated with the distribution of V. aspis. Although precipitation (BIO_12) is associated with high
presence for Maxnet and GLM, a temperature variable (BIO_3) is generally important to all models.
We can plot the response curves for each variable, which show how the predictions of the model vary with
the range of values for each variable. This involves using the models to predict across the entire range of
values for one variable while keeping the other variables fixed at a constant value. In this example, the other
variables are fixed at their median value.
plt <- bm_PlotResponseCurves(vaModel, fixed.var = 'median')

37
 Response curves for Vaspis's models
evi BIO_8 BIO_3
1.00
0.75
0.50
0.25
0.00

10

20

30

40
0.

0.

0.

0.

0.
−1
BIO_12 BIO_1
1.00
0.75
0.50
0.25
0.00

0
00

00

00

00

10

20
50

10

15

20

25
PA1_RUN1_GAM Vaspis_PA1_RUN4_GAM Vaspis_PA2_RUN2_GAM Vaspis_PA2_RUN5_GAM Vaspis_P

PA1_RUN1_GLM Vaspis_PA1_RUN4_GLM Vaspis_PA2_RUN2_GLM Vaspis_PA2_RUN5_GLM Vaspis_P

PA1_RUN1_MAXNET Vaspis_PA1_RUN4_MAXNET Vaspis_PA2_RUN2_MAXNET Vaspis_PA2_RUN5_MAXNET Vaspis_P

PA1_RUN1_XGBOOST Vaspis_PA1_RUN4_XGBOOST Vaspis_PA2_RUN2_XGBOOST Vaspis_PA2_RUN5_XGBOOST Vaspis_P

PA1_RUN2_GAM Vaspis_PA1_RUN5_GAM Vaspis_PA2_RUN3_GAM Vaspis_PA3_RUN1_GAM Vaspis_P

PA1_RUN2_GLM Vaspis_PA1_RUN5_GLM Vaspis_PA2_RUN3_GLM Vaspis_PA3_RUN1_GLM Vaspis_P

PA1_RUN2_MAXNET Vaspis_PA1_RUN5_MAXNET Vaspis_PA2_RUN3_MAXNET Vaspis_PA3_RUN1_MAXNET Vaspis_P

PA1_RUN2_XGBOOST Vaspis_PA1_RUN5_XGBOOST Vaspis_PA2_RUN3_XGBOOST Vaspis_PA3_RUN1_XGBOOST Vaspis_P

PA1_RUN3_GAM Vaspis_PA2_RUN1_GAM Vaspis_PA2_RUN4_GAM Vaspis_PA3_RUN2_GAM Vaspis_P

PA1_RUN3_GLM Vaspis_PA2_RUN1_GLM Vaspis_PA2_RUN4_GLM Vaspis_PA3_RUN2_GLM Vaspis_P

There are Vaspis_PA2_RUN1_MAXNET

PA1_RUN3_MAXNET many combinations of models, runs, and pseudo-absence sets, making
Vaspis_PA2_RUN4_MAXNET these plots complex to
Vaspis_PA3_RUN2_MAXNET Vaspis_P
interpret. After ensembling (chapter 7), we will construct simpler response curves. However, it is important to
note that, in
PA1_RUN3_XGBOOST general, the responses show similar
Vaspis_PA2_RUN1_XGBOOST patterns. We should give more importance
Vaspis_PA2_RUN4_XGBOOST to the responses
Vaspis_PA3_RUN2_XGBOOST Vaspis_P
of the variables that are more significant for each algorithm. For instance, it is not useful tssp in sspso focus
on EVI for Maxnet and GLM algorithms.
The models were built in the environmental space. The environmental values for the four predictors were
extracted from the locations (presence and pseudo-absences) by biomod2. The models were then built, tested,
and evaluated.
At this point, we have a large combination of models that will be ensembled into a single one in the next
chapter. However, we can still project the model to the entire study area. This will create a raster file that
we can open in any GIS, which is useful for checking the spatial prediction patterns for each individual model
and perform visual checking. The projections we will use to answer our initial question will be made in
chapter 8, only after ensembling the models.
vaProj <- BIOMOD_Projection(bm.mod = vaModel,
proj.name = 'Present',
new.env = vars,
models.chosen = 'all')

We can directly plot all models, but it might take a long time to finish since there are many models (PA x
Runs x Algorithm = 3 x 5 x 4 = 60 models) to show. However, all models are built and saved into a single
file, models/Vaspis/proj_Present/proj_Present_Vaspis.tif, which can be inspected in any GIS.

Repeat for the other two species

Vipera latastei
The code is basically the same, just changing the input data and directory names for Vipera latastei.

38
n_vl <- sum(pres$species == "Vlatastei")
coords_vl <- pres[pres$species == "Vlatastei", 2:3]

vlData <- BIOMOD_FormatingData(resp.var = rep(1, n_vl),

expl.var = vars,
resp.xy = coords_vl,
resp.name = "Vlatastei",
dir.name = "models",
PA.nb.rep = 3,
PA.nb.absences = 10000,
PA.strategy = "disk",
PA.dist.max = 110000)

vlModel <- BIOMOD_Modeling(bm.format = vlData,

modeling.id = "EcoMod",
models = c("GAM", "GLM", "MAXNET", "XGBOOST"),
CV.strategy = "kfold",
CV.k = 5,
CV.do.full.models = FALSE,
OPT.strategy = "bigboss",
var.import = 3,
metric.eval = c("TSS", "ROC"))
# Plotting examples
plt <- bm_PlotEvalMean(vlModel, dataset="calibration")

0.35

GAM
GLM
TSS

MAXNET
XGBOOST

0.30

0.675 0.700 0.725 0.750

ROC

39
plt <- bm_PlotEvalMean(vlModel, dataset="validation")

0.35

GAM
GLM
TSS

MAXNET
0.30
XGBOOST

0.25

0.650 0.675 0.700 0.725 0.750 0.775

ROC
plt <- bm_PlotEvalBoxplot(vlModel, group.by = c('algo', 'run'))

40
 ROC TSS

0.7

0.6 RUN1
calibration

RUN2
RUN3
0.5
RUN4
RUN5
0.4

0.3
AM

LM

AM
ET

ST

LM

ET

ST
N

O
G

G
G

G
AX

AX
BO

BO
M

M
XG

XG
plt <- bm_PlotVarImpBoxplot(vlModel, group.by = c('expl.var', 'run', 'algo'))

GAM GLM
70%
60%
50%
40%
30%
20%
10%
0% RUN1
RUN2
_1

_3

_8

_1

_3

_8

i
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI

RUN3
BI

BI

MAXNET XGBOOST RUN4

70% RUN5
60%
50%
40%
30%
20%
10%
0%
_1

_3

_8

_1

_3

_8

i
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI
BI

BI

41
 plt <- bm_PlotResponseCurves(vlModel, fixed.var = 'median')

Response curves for Vlatastei's models

evi BIO_8 BIO_3

0.75
0.50
0.25
0.00
0

−5

10

15

20

30

35

40

45
0.

0.

0.

BIO_12 0. BIO_1

0.75
0.50
0.25
0.00
0

5
00

00

00

10

15

20
50

10

15

20

_RUN1_GAM Vlatastei_PA1_RUN2_GAM Vlatastei_PA1_RUN3_GAM Vlatastei_PA1_RUN4_GAM Vlata

_RUN1_GLM Vlatastei_PA1_RUN2_GLM Vlatastei_PA1_RUN3_GLM Vlatastei_PA1_RUN4_GLM Vlata

_RUN1_MAXNET Vlatastei_PA1_RUN2_MAXNET Vlatastei_PA1_RUN3_MAXNET Vlatastei_PA1_RUN4_MAXNET Vlata

_RUN1_XGBOOST Vlatastei_PA1_RUN2_XGBOOST Vlatastei_PA1_RUN3_XGBOOST Vlatastei_PA1_RUN4_XGBOOST Vlata

vlProj <- BIOMOD_Projection(bm.mod = vlModel,

proj.name = 'Present',
new.env = vars,
models.chosen = 'all')

Vipera seoanei
The input data is now for Vipera seoanei.
n_vs <- sum(pres$species == "Vseoanei")
coords_vs <- pres[pres$species == "Vseoanei", 2:3]

vsData <- BIOMOD_FormatingData(resp.var = rep(1, n_vs),

expl.var = vars,
resp.xy = coords_vs,
resp.name = "Vseoanei",
dir.name = "models",
PA.nb.rep = 3,
PA.nb.absences = 10000,
PA.strategy = "disk",
PA.dist.max = 110000)

vsModel <- BIOMOD_Modeling(bm.format = vsData,

modeling.id = "EcoMod",
models = c("GAM", "GLM", "MAXNET", "XGBOOST"),
CV.strategy = "kfold",

42
 CV.k = 5,
CV.do.full.models = FALSE,
OPT.strategy = "bigboss",
var.import = 3,
metric.eval = c("TSS", "ROC"))
# Plotting examples
plt <- bm_PlotEvalMean(vsModel, dataset="calibration")

0.55

GAM
GLM
TSS

0.50
MAXNET
XGBOOST

0.45

0.40
0.74 0.76 0.78 0.80 0.82 0.84 0.86
ROC
plt <- bm_PlotEvalMean(vsModel, dataset="validation")

43
 0.6

GAM
0.5
GLM
TSS

MAXNET
XGBOOST

0.4

0.70 0.75 0.80 0.85

ROC
plt <- bm_PlotEvalBoxplot(vsModel, group.by = c('algo', 'run'))

ROC TSS

0.8

0.7 RUN1
calibration

RUN2
RUN3

0.6 RUN4
RUN5

0.5

0.4
AM

LM

AM
ET

ST

LM

ET

ST
N

O
G

G
G

G
AX

AX
BO

BO
M

M
XG

XG

44
plt <- bm_PlotVarImpBoxplot(vsModel, group.by = c('expl.var', 'run', 'algo'))

GAM GLM

70%
60%
50%
40%
30%
20%
10%
0% RUN1
RUN2
_1

_3

_8

_1

_3

_8

i
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI
RUN3
BI

BI
MAXNET XGBOOST RUN4

70% RUN5
60%
50%
40%
30%
20%
10%
0%
_1

_3

_8

_1

_3

_8

i
ev

ev
_1

_1
O

O
O

O
BI

BI

BI

BI

BI

BI
BI

BI

plt <- bm_PlotResponseCurves(vsModel, fixed.var = 'median')

45
 Response curves for Vseoanei's models
evi BIO_8 BIO_3
1.00
0.75
0.50
0.25
0.00
2

−5

10

15

35

40

45
0.

0.

0.

0.

0.

0.
BIO_12 BIO_1
1.00
0.75
0.50
0.25
0.00
0

5
00

00

00

10

15
50

10

15

20

RUN1_GAM Vseoanei_PA1_RUN2_GAM Vseoanei_PA1_RUN3_GAM Vseoanei_PA1_RUN4_GAM Vse

RUN1_GLM Vseoanei_PA1_RUN2_GLM Vseoanei_PA1_RUN3_GLM Vseoanei_PA1_RUN4_GLM Vse

RUN1_MAXNET Vseoanei_PA1_RUN2_MAXNET Vseoanei_PA1_RUN3_MAXNET Vseoanei_PA1_RUN4_MAXNET Vse

RUN1_XGBOOST Vseoanei_PA1_RUN2_XGBOOST Vseoanei_PA1_RUN3_XGBOOST Vseoanei_PA1_RUN4_XGBOOST Vse

vlProj <- BIOMOD_Projection(bm.mod = vsModel,

proj.name = 'Present',
new.env = vars,
models.chosen = 'all')

Final considerations for the modelling

Now, all models are available in the models directory. We will access them in the next chapters for ensembling,
projecting, and plotting.
By examining the different plots for each species, one can observe that different variables have varying
importance for predicting each species. This is a significant finding, as it allows us to describe the niche of
each species in terms of environmental preferences. Some species’ models will be more dependent on climate
variables, making their distributions more susceptible to climate change. Species that rely more on static
variables (NDVI, in this example) will have distributions in projections for other time periods that are less
prone to change.

Chapter 7 - Model ensemble

In this chapter, we require the 60 models resulting from the combinations of pseudo-absence sets, modeling
algorithms, and resampling strategies for each species. Ensembling models involves summarizing all this
information into a single prediction. This allows us to capture a wide range of possible responses arising from
the different combinations. Some algorithms have a stochastic nature, like machine learning methods, and
even without changing any other conditions, running the model with the same data might generate slightly
different predictions. Other algorithms are more static, such as GLM: running the model with exactly the
same data will generate the same prediction. That’s one reason we vary the initial conditions of the model
(e.g., pseudo-absences dataset and resampling). This wider range of predictions is supposed to capture slight
variations in information about the species’ niche and thus create a more robust final model.

46
For this chapter we will need again the biomod2 package.
library(biomod2)

Ensembling Vipera aspis

Assuming that we have a clean workspace, we have to load the models from the folder where they were saved
models/vaspis. As biomod2 deals with very different algorithms and data, it saves many information and data
that might be cumbersome to understand what is in the folder. However, by importing into R one of the files,
we can access all information through biomod2 functions. It requires a small coding trick to fully import the
model data with a common object name. We have to load and retrive that data in two lines of code:
vaName <- load("models/Vaspis/Vaspis.EcoMod.models.out")
vaModel <- eval(str2lang(vaName))
vaModel

The object vaModel contains all the models built in the previous chapter. Now, we can proceed with
ensembling. The package provides a straightforward function with customizable arguments for ensembling.
We specify the models we want to ensemble and choose the type of ensembling. In this example, we are
ensembling all models together, meaning that all combinations of pseudo-absences, algorithms, and runs are
merged. However, we could choose to ensemble only specific subsets, such as by algorithm, by setting em.by
= "PA_dataset+repet", which would result in separate ensembles for each algorithm. This approach could
be useful for evaluating the performance of each algorithm individually. Here, we are ensembling all models
together using em.by = "all" while models.chosen = 'all' aensures that all built models are available
for ensembling.
We select the median as the ensembling statistic. Other options available include mean and weighted mean,
where weights are typically based on model performance metrics or confidence intervals.
Since the resulting ensemble model provides predictions that have not yet been tested, we evaluate its
performance using the same metrics as before (TSS and ROC). Variable importance is also assessed using 3
permutations, as in previous steps.
vaEnsbl <- BIOMOD_EnsembleModeling(bm.mod = vaModel,
models.chosen = 'all',
em.by = 'all',
em.algo = 'EMmedian',
metric.eval = c('TSS', 'ROC'),
var.import = 3)

vaEnsbl

##
## -=-=-=-=-=-=-=-=-=-=-=-=-= BIOMOD.ensemble.models.out -=-=-=-=-=-=-=-=-=-=-=-=-=
##
## sp.name : Vaspis
##
## expl.var.names : evi BIO_8 BIO_3 BIO_12 BIO_1
##
##
## models computed:
## Vaspis_EMmedianByTSS_mergedData_mergedRun_mergedAlgo, Vaspis_EMmedianByROC_mergedData_mergedRun_merge
##
## models failed: none
##
## -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
At this stage, we should have a new file in the models/Vaspis folder containing the ensembling results, which

47
we can access later as needed.
We can evaluate the ensembled model by plotting its performance metrics. Now we don’t have calibration
and validation datasets and we test with full data.
plt <- bm_PlotEvalBoxplot(vaEnsbl, group.by = c('algo', 'algo'))

ROC TSS

0.7
calibration

0.6

EMmedian

0.5

0.4
n

n
ia

ia
ed

ed
m

m
EM

EM

Both metrics show reasonable performance. However, at this stage, we should consider adding more predictors
or potentially removing some to improve model performance. If an important predictor defining the species
niche is missing, it could significantly affect performance. Another strategy could involve adjusting the width
of the buffer for pseudo-absences. Increasing environmental variability might facilitate model fitting, but this
approach should be carefully balanced and justified (as it might be a contentious subject!).
We can also check the variable importance.
plt <- bm_PlotVarImpBoxplot(vaEnsbl, group.by = c('expl.var', 'algo', 'algo'))

48
 EMmedian

40%

30%

EMmedian
20%

10%

0%
_1

_3

_8

i
ev
_1
O

O
O
BI

BI

BI
BI

In this model, BIO_12 and BIO_7 are clearly the most important variables, while NDVI contributes the
least to the models.”
And finally we can check variable response curves.
plt <- bm_PlotResponseCurves(vaEnsbl, fixed.var = 'median')

49
 Response curves for Vaspis's models
evi BIO_8 BIO_3

0.6

0.4

0.2
0

10

20

30

40
0.

0.

0.

0.

0.

−1
BIO_12 BIO_1

0.6

0.4

0.2
0

00

00

00

00

10

20
50

10

15

20

25

Vaspis_EMmedianByTSS_mergedData_mergedRun_mergedAlgo Vaspis_EMmedianByROC_mergedData_mergedRun_mergedAlgo

The ensembled models provide a single response curve for each variable, which is easier to interpret than
before. We should give more importance to the BIO_12 and BIO_7 curves, as those are the variables that
contribute more to the models. The model seems to increase the predictive probability with the increase of
the value of these variables. Notice how the NDVI, which is the least important variable, appears to have a
trend, but the prediction range associated with it is low.

Ensembling Vipera latastei

Repeat the code, changing the model.
vlName <- load("models/Vlatastei/Vlatastei.EcoMod.models.out")
vlModel <- eval(str2lang(vlName))

# Ensemble
vlEnsbl <- BIOMOD_EnsembleModeling(bm.mod = vlModel,
models.chosen = 'all',
em.by = 'all',
em.algo = 'EMmedian',
metric.eval = c('TSS', 'ROC'),
var.import = 3)

# Plotting examples
plt <- bm_PlotEvalBoxplot(vlEnsbl, group.by = c('algo', 'algo'))

50
 ROC TSS

0.7

0.6
calibration

EMmedian

0.5

0.4
n

n
ia

ia
ed

ed
m

m
EM

EM

plt <- bm_PlotVarImpBoxplot(vlEnsbl, group.by = c('expl.var', 'algo', 'algo'))

EMmedian

60%

50%

40%

EMmedian
30%

20%

10%

0%
_1

_3

_8

i
ev
_1
O

O
O
BI

BI

BI
BI

51
 plt <- bm_PlotResponseCurves(vlEnsbl, fixed.var = 'median')

Response curves for Vlatastei's models

evi BIO_8 BIO_3
0.8

0.6

0.4

0.2
0

−5

10

15

20

30

35

40

45
0.

0.

0.

0.

BIO_12 BIO_1
0.8

0.6

0.4

0.2
0

00

00

00

10

15

20
50

10

15

20

Vlatastei_EMmedianByTSS_mergedData_mergedRun_mergedAlgo Vlatastei_EMmedianByROC_mergedData_mergedRun_mergedAlg

Ensembling Vipera seoanei

And for the third species:
vsName <- load("models/Vseoanei/Vseoanei.EcoMod.models.out")
vsModel <- eval(str2lang(vsName))

# Ensemble
vsEnsbl <- BIOMOD_EnsembleModeling(bm.mod = vsModel,
models.chosen = 'all',
em.by = 'all',
em.algo = 'EMmedian',
metric.eval = c('TSS', 'ROC'),
var.import = 3)

# Plotting examples
plt <- bm_PlotEvalBoxplot(vsEnsbl, group.by = c('algo', 'algo'))

52
 ROC TSS

0.8

0.7
calibration

EMmedian

0.6
n

n
ia

ia
ed

ed
m

m
EM

EM

plt <- bm_PlotVarImpBoxplot(vsEnsbl, group.by = c('expl.var', 'algo', 'algo'))

EMmedian

60%

50%

40%

EMmedian
30%

20%

10%
_1

_3

_8

i
ev
_1
O

O
O
BI

BI

BI
BI

53
 plt <- bm_PlotResponseCurves(vsEnsbl, fixed.var = 'median')

Response curves for Vseoanei's models

evi BIO_8 BIO_3

0.75

0.50

0.25

0.00
2

−5

10

15

35

40

45
0.

0.

0.

0.

0.

0.

BIO_12 BIO_1

0.75

0.50

0.25

0.00
0

00

00

00

10

15
50

10

15

20

Vseoanei_EMmedianByTSS_mergedData_mergedRun_mergedAlgo Vseoanei_EMmedianByROC_mergedData_mergedRun_mergedA

Final considerations for the ensembling

The ensembling process reduces the complexity of having multiple models to analyze while ensuring the
robustness of predictions. With the model ensembles now constructed, we are ready to project these models
onto the Iberian Peninsula.

Chapter 8 - Model projections

In this chapter, we are going to use the ensemble models to project onto our region of interest, which is the
Iberian Peninsula. Additionally, we have multiple ages and SSPs to project. By the end of this chapter, we
will have projections for
• Current conditions
• Period 2011-2040
– SSP 126
– SSP 585
• Period 2041-2060
– SSP 126
– SSP 485
• Period 2061-2080
– SSP 126
– SSP 586
This results in 7 model projections for the Iberian Peninsula.

54
We need both terra library for opening the varaibles, and biomod2 for interpreting the models.
library(terra)
library(biomod2)

We can define here some variables that will be useful for looping over ages and SSPs (similar to Chapter 5)
to read and project models to different time periods.”
ages <- c("2021-2040", "2041-2060", "2061-2080")
ssps <- c("126", "585")

Again, we will proceed with the first species and then run the same code for the other two to obtain all
projections.
We can open the raster for the current period variables as this will be constant for all current projections.
curp <- rast("data/rasters/proj_current.tif")

Projecting Vipera aspis models

WE use the same trick as in the previous chapter to open the ensembled model available in the models/Vaspis
folder:
vaName <- load("models/Vaspis/Vaspis.EcoMod.ensemble.models.out")
vaEnsbl <- eval(str2lang(vaName))

Until now, we have been focusing in the continuous probability predictions. However, models can be converted
to binary (presence vs. absence) by setting a threshold that results in best performance. We can check which
threshold biomod2 is using for these models.
get_evaluations(vaEnsbl)

## full.name merged.by.PA
## 1 Vaspis_EMmedianByTSS_mergedData_mergedRun_mergedAlgo mergedData
## 2 Vaspis_EMmedianByTSS_mergedData_mergedRun_mergedAlgo mergedData
## 3 Vaspis_EMmedianByROC_mergedData_mergedRun_mergedAlgo mergedData
## 4 Vaspis_EMmedianByROC_mergedData_mergedRun_mergedAlgo mergedData
## merged.by.run merged.by.algo filtered.by algo metric.eval cutoff
## 1 mergedRun mergedAlgo TSS EMmedian TSS 498.0
## 2 mergedRun mergedAlgo TSS EMmedian ROC 500.5
## 3 mergedRun mergedAlgo ROC EMmedian TSS 498.0
## 4 mergedRun mergedAlgo ROC EMmedian ROC 500.5
## sensitivity specificity calibration validation evaluation
## 1 72.650 66.797 0.394 NA NA
## 2 71.926 67.609 0.738 NA NA
## 3 72.650 66.797 0.394 NA NA
## 4 71.926 67.609 0.738 NA NA
Depending on the metric, the threshold varies as it has to maximize the performance of each metric separately.
For TSS, the threshold (cutoff) is set to 0.474, and for ROC, it is set to 0.468. Note that biomod2 multiplies
the probabilities by 1000, so they range between 0 and 1000.
We can project the model to current conditions in the Iberian Peninsula. The function will create both
continuous predictions and binary predictions using the thresholds mentioned above. It is important to name
the projections appropriately for easy identification of the rasters. We set the proj.name to Current and
a new folder named proj_Current will be created inside models/Vaspis. In this folder, two rasters will be
saved for continuous and binary predictions. Also, the argument new.env refers to the new environmental
conditions that are the raster variables we want to project to.

55
vaProj <- BIOMOD_EnsembleForecasting(bm.em = vaEnsbl,
proj.name = 'Current',
new.env = curp,
models.chosen = 'all',
metric.binary = 'TSS')

We will have to do the same for each combination of age and SSP projections. For that, we use a nested for
loop as before (chapter 5). The logic is as follows:
• For each combination of age and ssp
1. Construct the respective filename using age and SSP and open the raster.
2. Project the ensembled model to the respective time period, correctly naming the projection with
age_ssp.
for (age in ages) {
for (ssp in ssps) {
projVars <- rast(paste0("data/rasters/proj_", age, "_", ssp, ".tif"))
vaProj <- BIOMOD_EnsembleForecasting(bm.em = vaEnsbl,
proj.name = paste0(age, "_", ssp),
new.env = projVars,
models.chosen = 'all',
metric.binary = 'TSS')
}
}

At this stage, we should have a folder inside models/Vaspis for each projection that we built. We now need
to run the code for the other two species to complete all projections.

Projecting Vipera latastei models

We load the Vipera latastei ensembled model and run the code.
vlName <- load("models/Vlatastei/Vlatastei.EcoMod.ensemble.models.out")
vlEnsbl <- eval(str2lang(vlName))

get_evaluations(vlEnsbl)

# Project to current time

vlProj <- BIOMOD_EnsembleForecasting(bm.em = vlEnsbl,
proj.name = 'Current',
new.env = curp,
models.chosen = 'all',
metric.binary = 'TSS')
# Project to future periods
for (age in ages) {
for (ssp in ssps) {
projVars <- rast(paste0("data/rasters/proj_", age, "_", ssp, ".tif"))
vlProj <- BIOMOD_EnsembleForecasting(bm.em = vlEnsbl,
proj.name = paste0(age, "_", ssp),
new.env = projVars,
models.chosen = 'all',
metric.binary = 'TSS')
}
}

56
Projecting Vipera latastei models
Nowe for Vipera seoanei.
vsName <- load("models/Vseoanei/Vseoanei.EcoMod.ensemble.models.out")
vsEnsbl <- eval(str2lang(vsName))

get_evaluations(vsEnsbl)

# Project to current time

vsProj <- BIOMOD_EnsembleForecasting(bm.em = vsEnsbl,
proj.name = 'Current',
new.env = curp,
models.chosen = 'all',
metric.binary = 'TSS')
# Project for future
for (age in ages) {
for (ssp in ssps) {
projVars <- rast(paste0("data/rasters/proj_", age, "_", ssp, ".tif"))
vsProj <- BIOMOD_EnsembleForecasting(bm.em = vsEnsbl,
proj.name = paste0(age, "_", ssp),
new.env = projVars,
models.chosen = 'all',
metric.binary = 'TSS')
}
}

Final considerations for projections

At this point, we have all the projections needed to answer our initial question. By defining a smaller
projection zone (the Iberian Peninsula) relevant to our research question, we have optimized both computer
storage and processing time, as we avoid unnecessary projections over a larger area.
All files are saved in respective folders and we can access them using any GIS. As an example, here are the
model projections for Vipera aspis for the current and future periods, only for scenario ssp 585.
p1 <- rast("models/Vaspis/proj_Current/proj_Current_Vaspis_ensemble.tif")
p2 <- rast("models/Vaspis/proj_2021-2040_585/proj_2021-2040_585_Vaspis_ensemble.tif")
p3 <- rast("models/Vaspis/proj_2041-2060_585/proj_2041-2060_585_Vaspis_ensemble.tif")
p4 <- rast("models/Vaspis/proj_2061-2080_585/proj_2061-2080_585_Vaspis_ensemble.tif")

layout(matrix(1:4, 1))
plot(p1[[1]], main="Present", range=c(0,1000))
plot(p2[[1]], main="2021-2040", range=c(0,1000))
plot(p3[[1]], main="2041-2060", range=c(0,1000))
plot(p4[[1]], main="2061-2080", range=c(0,1000))
Present 2021−2040 2041−2060 2061−2080
1000 1000 1000 1000
800 800 800 800
600 600 600 600
40

40

40

40

400 400 400 400

200 200 200 200
0 0 0 0
−8 −4 0 2 −8 −4 0 2 −8 −4 0 2 −8 −4 0 2

57
Chapter 9 - Finally, answering the question
In this chapter, we will open all projection rasters created previously and analyze how these predictions
can help identify sympatric areas for the three species. As the results showed, the models reveal different
relationships with the predictors, resulting in distinct distributions. How do these potential distributions
intersect with each other?
We will use the terra package since most operations in this chapter involve geoprocessing with rasters. While
any GIS software could be used, using terra allows us to automate some processes.
library(terra)

We can also prepare some variables to loop over future periods, as we did before.
ages <- c("2021-2040", "2041-2060", "2061-2080")
ssps <- c("126", "585")

We can begin by importing the rasters of the current projections for the three species. We divide by 1000 to
get the projections in the range of 0 to 1.
va <- rast("models/Vaspis/proj_Current/proj_Current_Vaspis_ensemble.tif")[[1]]
va <- va / 1000
vl <- rast("models/Vlatastei/proj_Current/proj_Current_Vlatastei_ensemble.tif")[[1]]
vl <- vl / 1000
vs <- rast("models/Vseoanei/proj_Current/proj_Current_Vseoanei_ensemble.tif")[[1]]
vs <- vs / 1000

We can plot the maps:

layout(matrix(1:3, 1))
plot(va, main="V. aspis", range=c(0,1))
plot(vl, main="V. latastei", range=c(0,1))
plot(vs, main="V. seoanei", range=c(0,1))
V. aspis V. latastei V. seoanei
1.0 1.0 1.0

0.8 0.8 0.8

0.6 0.6 0.6

40

40

40

0.4 0.4 0.4

0.2 0.2 0.2

0.0 0.0 0.0

−8 −6 −4 −2 0 2 −8 −6 −4 −2 0 2 −8 −6 −4 −2 0 2

We need some processing to determine the sympatric zone between the vipers. Since we are using continuous
probability maps, an easy way to estimate the maximum probability of finding all three species together is
by calculating the product of the three maps. The potential maximum score would be 1, where all species
have the highest probability, while any location with a probability of 0 for any species will have a sympatry
probability of zero.
Additionally, there are other methods to calculate sympatry. For instance, using the binary presence/absence
maps produced earlier, the sum of all maps would provide a range of values from 0 to 3, indicating how many
species are present at each location.
Here, we will calculate the product of the probabilities:
CZ <- va*vl*vs
names(CZ) <- "Present"
plot(CZ, main= "Current contact zone probability")

58
 Current contact zone probability
0.30
43

0.25
42

0.20
41
40

0.15
39

0.10
38

0.05
37

0.00
36

−8 −6 −4 −2 0 2

We have to calculate the same for all combinations of age and ssp projections. For that we use the same style
of nested loop we have been using. We will be adding layers to CZ raster with correct names, following the
age_ssp order.
The logic of the loop is the following: * for each combination of age and ssp: 1. Reconstruct the general
name of the file 2. Reconstruct the path for V aspis projection file 3. Import the raster and divide by 1000 4.
Reconstruct the path for V latastei projection file 5. Import the raster and divide by 1000 6. Reconstruct
the path for V seoanei projection file 7. Import the raster and divide by 1000 8. Create the future prediction
of contact zone as futCZ 9. name the layer correctly with age_ssp 10. Stack this layer to the CZ raster.
for (ssp in ssps) {
for (age in ages) {
nm <- paste0("proj_", age, "_", ssp)
vaFile <- paste0("models/Vaspis/", nm, "/", nm, "_Vaspis_ensemble.tif")
va <- rast(vaFile)[[1]]/1000
vlFile <- paste0("models/Vlatastei/", nm, "/", nm, "_Vlatastei_ensemble.tif")
vl <- rast(vlFile)[[1]]/1000
vsFile <- paste0("models/Vseoanei/", nm, "/", nm, "_Vseoanei_ensemble.tif")
vs <- rast(vsFile)[[1]]/1000

# Produce Future contact zone and rename accordingly to age and ssp
futCZ <- va*vl*vs
names(futCZ) <- paste0(age, "_", ssp)

# Grow CZ raster with each futCZ to store all in same file

CZ <- c(CZ, futCZ)
}
}

59
We can now plot the raster or contact zone predictions.
layout(matrix(c(1,1,1,1:7), 5, byrow=TRUE))
plot(CZ[[1]], main=names(CZ)[1], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[2]], main=names(CZ)[2], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[3]], main=names(CZ)[3], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[4]], main=names(CZ)[4], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[5]], main=names(CZ)[5], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[6]], main=names(CZ)[6], col=hcl.colors(25), range=c(0, 0.3))
plot(CZ[[7]], main=names(CZ)[7], col=hcl.colors(25), range=c(0, 0.3))

60
 Present

43
0.25
42

0.20
41
40

0.15
39

0.10
38

0.05
37
36

0.00
−5 0

2021−2040_126 2041−2060_126
36 37 38 39 40 41 42 43

36 37 38 39 40 41 42 43

0.25 0.25

0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0.00 0.00
−5 0 −5 0

2061−2080_126 2021−2040_585
36 37 38 39 40 41 42 43

36 37 38 39 40 41 42 43

0.25 0.25

0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0.00 0.00
−5 0 −5 0

2041−2060_585 2061−2080_585
36 37 38 39 40 41 42 43

36 37 38 39 40 41 42 43

0.25 0.25

0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0.00 0.00
−5 0 61 −5 0
We can opt for other visualization strategies to show the change in the contact zone probabilities. For instance,
we can show histograms of the probabilities to highlight how they will decrease for certain scenarios. The
easiest way to do this is to extract the raster data into a data frame:
CZdata <- data.frame(CZ)
head(CZdata)

## Present X2021.2040_126 X2041.2060_126 X2061.2080_126 X2021.2040_585

## 18 0.09168883 NA NA NA NA
## 19 0.12138378 0.11272800 0.09560678 0.08763062 0.09560786
## 20 0.06488832 0.04411405 0.04377840 0.04236144 0.04340827
## 21 0.06788331 0.04453695 0.04372839 0.04458352 0.04625597
## 22 0.10601677 0.06910594 0.06706800 0.06771072 0.06597292
## 23 0.09536476 0.05224032 0.05335960 0.05381501 0.05408720
## X2041.2060_585 X2061.2080_585
## 18 NA NA
## 19 0.10096125 0.05645640
## 20 0.03452803 0.02412800
## 21 0.03658774 0.02379909
## 22 0.05417736 0.03859200
## 23 0.03964464 0.02956976
As the whole projected area is dominated by near-zero values (low probability of contact), we will remove
these values based on a very small threshold. This will prevent the histograms from being dominated by
the large number of pixels with near-zero values, providing a clearer view of the changes in contact zone
probabilities.
CZdata[CZdata < 0.01] <- NA

For the histograms, we need to define the width of the bars and the range of values. In this example, the
values vary between 0 and 0.28, so we choose a width of 0.01 and create a sequence of cutting points (breaks).
brk <- seq(0, 0.35, 0.01)

We can retrieve the names of the layers in the raster to use as titles for plotting. The order of the layers
corresponds to the order of columns in the data frame.
titles <- names(CZ)

Instead of plotting each column individually, as we did above for plotting the rasters, we can use a for loop
that iterates over the columns. We set some nice colors for the bars and plot!
layout(matrix(c(1,1,1,1:7), 5, byrow=TRUE))
for (i in 1:ncol(CZdata)) {
hist(CZdata[,i], breaks=brk, main=titles[i], col="steelblue1", border="white")
}

62
 Present
800
600
Frequency

400
200
0

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

CZdata[, i]

2021−2040_126 2041−2060_126
700

800
Frequency

Frequency

400
300
0

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

CZdata[, i] CZdata[, i]

2061−2080_126 2021−2040_585
700
Frequency

Frequency
400

300
0

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

CZdata[, i] CZdata[, i]

2041−2060_585 2061−2080_585
800

400 800
Frequency

Frequency
400
0

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

CZdata[, i] 63 CZdata[, i]
We can see that the worst-case scenario predicts a contraction of the contact zone to a narrow area by the end
of the century. Currently, this area is where all three vipers are found together, while other regions represent
high potential for all species, although not all species reach those areas. One reason could be that they are at
the limit of their ecological niche, and one species may be excluded by another that is more dominant in the
region.
We can write the raster to a file that can be opened in any GIS for further exploration of the results.
{r] writeRaster(CZ, "models/sympatry.tif")

Conclusion
We have seen the process of ecological niche modeling using R, focusing on viper species in the Iberian
Peninsula as a case study. We covered topics such as data collection, preprocessing predictors and presence
data, model building, ensembling, and projection to future climate scenarios. The general steps we covered
were:
• Data Collection: Gather presence data for the species of interest along with environmental variables
like bioclimate varibles and NDVI.
• Data Preprocessing: Process the data by checking for duplicates in presence record and other
erroneous coordinates; making sure the predictors are fully geographically align and that correlations
among variables are adequate; prepare a training and projection area.
• Model Building: Use the biomod2 package to build ecological niche models, selecting appropriate
pseudo-absences, algorithms, and resampling strategies.
• Model Evaluation: Evaluate model performance using metrics like True Skill Statistic (TSS) and
Receiver Operating Characteristic (ROC) curves.
• Model Ensembling: Ensemble multiple models to create a more robust prediction by combining
different pseudo-absences sets, algorithms, and resampling strategies.
• Projection: Project the ensembled models to future climate scenarios to assess potential distribution
changes under different climate conditions.
• Spatial Analysis: Analyzs spatial patterns in the model predictions, for instance, identify sympatric
zones, and assess changes in contact zones over time.
At this point, you should understand that:
• Ecological niche modeling provides valuable insights into species distributions and responses to environ-
mental changes.
• Model performance depends on various factors such as data quality, variable selection, algorithm choice,
and vary between model evaluation metrics.
• Ensembling multiple models improves prediction accuracy and robustness by capturing a wider range
of situtations and simplify the analyses of complex model results.
• Future climate projections highlight potential shifts in species distributions, and might provide a good
insight for conservation planning and management strategies.
• Spatial analysis is an important process in the ecological modelling an, in this example, was essential to
identify areas of overlap and potential interactions between species.

64