IberianFishHSMs: Compilation of models and tools to evaluate the habitat suitability at the microhabitat scale for Iberian freshwater fish species. The modelling approaches include sets of Habitat Suitability Curves (HSCs), Artificial Neural Networks (Multilayer Perceptrons, ANN-MLPs), Random Forests (RFs), Generalised Additive Models (GAMs), and Support Vector Machines (SVMs), all based on empirical data collected across several river basins in the Iberian Peninsula. The HSCs have been reviewed by an expert panel, which co-authored the package. The package also compiles sets of HSCs and Fuzzy Rule-Based Systems (FRBSs) based on expert knowledge, which was gathered through on-line forms. Functions for evaluating the habitat suitability from hydraulic simulations and for plotting the modelled suitability are also included. A thorough example of the capabilities of the packages appears below. The package includes 797 microhabitat suitability models for 43 species and up to 4 size classes (Large, Medium, Small, Very small) and spawning suitability (only for Salmo trutta and Salmo salar).
This package was developed by:
- Rafael Muñoz-Mas1 (Author, Creator and Maintainer) - [email protected]
- Carlos Alonso2 (Contributor)
- Enric Aparicio Manau3 (Contributor)
- Fernando Cobo Gradín4 (Contributor)
- Gustavo González Fernández5 (Contributor)
- Javier Gortázar6 (Contributor)
- Francisco Martínez-Capel7 (Contributor)
- Francisco J. Oliva-Paterna8 (Contributor)
- Jose Prenda9 (Contributor)
- Jose Maria Santos10 (Contributor)
- Rafael Miranda Ferreiro11 (Contributor)
1 Tragsatec, Carrer Ramon Gordillo 7 - 46010 València, Spain. 2 3 GRECO, Institute of Aquatic Ecology, University of Girona, E-17071 Girona. 4 Laboratorio de Hidrobiologia, Departamento de Zoología, Genetica y Antropología Física. Facultad de Biología. Universidad de Santiago de Compostela, 15782 Santiago de Compostela, Spain. 5 Icthios Gestión Ambiental SL. [email protected]. 6 ECOHIDRÁULICA, S.L. C/ Villamanín 16, Madrid 28011. 7 Institut d’Investigació per a la Gestió Integrada de Zones Costaneres (IGIC), Universitat Politècnica de València, C/Paranimf, 1, 46730 Gandia, València, Spain. 8 Departamento de Zoología y Antropología Física. 30100. Universidad de Murcia. Murcia. 9 Dpto. de Ciencias Integradas. Universidad de Huelva. 10 Forest Research Centre, Associate Laboratory TERRA, School of Agriculture, University of Lisbon. 11 Universidad de Navarra, Instituto de Biodiversidad y Medioambiente (BIOMA), 31008 Pamplona, Navarra, Spain.
IberianFishHSMs can be installed from GitHub with:
# install.packages("devtools")
devtools::install_github("RafaMMas/IberianFishHSMs")The following example illustrates the capabilities of IberianFishHSMs and the general workflow for predicting microhabitat suitability using outputs from hydraulic simulations and field data. Field data includes the substrate and cover (shelter) distribution across the study site, by patch, pixel or cell. The data used in this example is available from the package.
library(IberianFishHSMs)ListModels allows inspecting the available models. The queries can be
categorized by Species, Size, River, Model.type, Sampled.season,
Valid.season, and/or Data.origin. This function also provides de Codes
to internally call the microhabitat suitability models during the
habitat evaluation. The output of this function also renders a summary
of their main characteristics, including sample sizes, and several
performance criteria for those models optimised through
cross-validation.
Selected.models <- ListModels(Species = "Salmo trutta", verbose = F)
## General characteristics
Selected.models$Current.summary.table[,1:15]
#> Code Model River River.short
#> 603 ABELZ Salmo.trutta.GAM.Large.Curueno.All Curueño (Douro) Curueno
#> 604 ABEMQ Salmo.trutta.HSC.Large.Curueno.All Curueño (Douro) Curueno
#> 605 ABEMV Salmo.trutta.NNET.Large.Curueno.All Curueño (Douro) Curueno
#> 606 ABEMZ Salmo.trutta.RF.Large.Curueno.All Curueño (Douro) Curueno
#> 607 ABENR Salmo.trutta.SVM.Large.Curueno.All Curueño (Douro) Curueno
#> 608 ABEMN Salmo.trutta.GAM.Medium.Curueno.All Curueño (Douro) Curueno
#> 609 ABEMR Salmo.trutta.HSC.Medium.Curueno.All Curueño (Douro) Curueno
#> 610 ABEMW Salmo.trutta.NNET.Medium.Curueno.All Curueño (Douro) Curueno
#> 611 ABENO Salmo.trutta.RF.Medium.Curueno.All Curueño (Douro) Curueno
#> 612 ABENS Salmo.trutta.SVM.Medium.Curueno.All Curueño (Douro) Curueno
#> 613 ABEMO Salmo.trutta.GAM.Small.Curueno.All Curueño (Douro) Curueno
#> 614 ABEMS Salmo.trutta.HSC.Small.Curueno.All Curueño (Douro) Curueno
#> 615 ABEMX Salmo.trutta.NNET.Small.Curueno.All Curueño (Douro) Curueno
#> 616 ABENP Salmo.trutta.RF.Small.Curueno.All Curueño (Douro) Curueno
#> 617 ABENT Salmo.trutta.SVM.Small.Curueno.All Curueño (Douro) Curueno
#> Model.type Default Species Size Sampled.season
#> 603 GAM TRUE Salmo trutta Large Summer
#> 604 HSC TRUE Salmo trutta Large Summer
#> 605 NNET TRUE Salmo trutta Large Summer
#> 606 RF TRUE Salmo trutta Large Summer
#> 607 SVM TRUE Salmo trutta Large Summer
#> 608 GAM TRUE Salmo trutta Medium Summer
#> 609 HSC TRUE Salmo trutta Medium Summer
#> 610 NNET TRUE Salmo trutta Medium Summer
#> 611 RF TRUE Salmo trutta Medium Summer
#> 612 SVM TRUE Salmo trutta Medium Summer
#> 613 GAM TRUE Salmo trutta Small Summer
#> 614 HSC TRUE Salmo trutta Small Summer
#> 615 NNET TRUE Salmo trutta Small Summer
#> 616 RF TRUE Salmo trutta Small Summer
#> 617 SVM TRUE Salmo trutta Small Summer
#> Valid.season Valid.seasons.short Data.origin N.presences
#> 603 Winter; Spring; Summer; Autumn All Field survey 24
#> 604 Winter; Spring; Summer; Autumn All Field survey 24
#> 605 Winter; Spring; Summer; Autumn All Field survey 24
#> 606 Winter; Spring; Summer; Autumn All Field survey 24
#> 607 Winter; Spring; Summer; Autumn All Field survey 24
#> 608 Winter; Spring; Summer; Autumn All Field survey 81
#> 609 Winter; Spring; Summer; Autumn All Field survey 81
#> 610 Winter; Spring; Summer; Autumn All Field survey 81
#> 611 Winter; Spring; Summer; Autumn All Field survey 81
#> 612 Winter; Spring; Summer; Autumn All Field survey 81
#> 613 Winter; Spring; Summer; Autumn All Field survey 385
#> 614 Winter; Spring; Summer; Autumn All Field survey 385
#> 615 Winter; Spring; Summer; Autumn All Field survey 385
#> 616 Winter; Spring; Summer; Autumn All Field survey 385
#> 617 Winter; Spring; Summer; Autumn All Field survey 385
#> N.absences N.interviews.or.studies
#> 603 896 NA
#> 604 896 NA
#> 605 896 NA
#> 606 896 NA
#> 607 896 NA
#> 608 839 NA
#> 609 839 NA
#> 610 839 NA
#> 611 839 NA
#> 612 839 NA
#> 613 535 NA
#> 614 535 NA
#> 615 535 NA
#> 616 535 NA
#> 617 535 NA
## Performance
Selected.models$Current.summary.table[,16:23]
#> True.positive False.positive True.negative False.negative Sensitivity
#> 603 24 215 681 0 1.000
#> 604 NA NA NA NA NA
#> 605 24 106 790 0 1.000
#> 606 24 106 790 0 1.000
#> 607 22 157 739 2 0.917
#> 608 70 218 621 11 0.864
#> 609 NA NA NA NA NA
#> 610 75 197 642 6 0.926
#> 611 75 197 642 6 0.926
#> 612 68 200 639 13 0.840
#> 613 275 210 325 110 0.714
#> 614 NA NA NA NA NA
#> 615 286 211 324 99 0.743
#> 616 286 211 324 99 0.743
#> 617 262 179 356 123 0.681
#> Specificity TSS Balanced.accuracy
#> 603 0.760 0.760 0.880
#> 604 NA NA NA
#> 605 0.882 0.882 0.941
#> 606 0.882 0.882 0.941
#> 607 0.825 0.741 0.871
#> 608 0.740 0.604 0.802
#> 609 NA NA NA
#> 610 0.765 0.691 0.846
#> 611 0.765 0.691 0.846
#> 612 0.762 0.601 0.801
#> 613 0.607 0.322 0.661
#> 614 NA NA NA
#> 615 0.606 0.348 0.674
#> 616 0.606 0.348 0.674
#> 617 0.665 0.346 0.673
## Models and Codes
Selected.models[2:3]
#> $Models
#> [1] "Salmo.trutta.GAM.Large.Curueno.All"
#> [2] "Salmo.trutta.HSC.Large.Curueno.All"
#> [3] "Salmo.trutta.NNET.Large.Curueno.All"
#> [4] "Salmo.trutta.RF.Large.Curueno.All"
#> [5] "Salmo.trutta.SVM.Large.Curueno.All"
#> [6] "Salmo.trutta.GAM.Medium.Curueno.All"
#> [7] "Salmo.trutta.HSC.Medium.Curueno.All"
#> [8] "Salmo.trutta.NNET.Medium.Curueno.All"
#> [9] "Salmo.trutta.RF.Medium.Curueno.All"
#> [10] "Salmo.trutta.SVM.Medium.Curueno.All"
#> [11] "Salmo.trutta.GAM.Small.Curueno.All"
#> [12] "Salmo.trutta.HSC.Small.Curueno.All"
#> [13] "Salmo.trutta.NNET.Small.Curueno.All"
#> [14] "Salmo.trutta.RF.Small.Curueno.All"
#> [15] "Salmo.trutta.SVM.Small.Curueno.All"
#>
#> $Codes
#> [1] "ABELZ" "ABEMQ" "ABEMV" "ABEMZ" "ABENR" "ABEMN" "ABEMR" "ABEMW" "ABENO"
#> [10] "ABENS" "ABEMO" "ABEMS" "ABEMX" "ABENP" "ABENT"ListModels output values can be use directly within
PredictHabitatSuitability to carry out habitat evaluations. In
particula, $Codes are the names used to store internally the available
microhabitat suitability models.
There exist an interaction between training data and the selected
modelling technique, which can vary variable effects and importance
(Eugster, Leisch, and Strobl 2014). By means of cross-validation, during
the development of the microhabitat suitability models we carried out a
variable selection approach discarding those cover types that not
improved model performance for Machine Learning (i.e., ANN-MLPs, RFs,
SVMs) and statistical approaches (i.e., GAMs) and an analogous approach
during the development of the Habitat Suitability Curves/Criteria (HSCs)
and Fuzzy Rule-based Systems (FRBSs). This approach lead to different
sets of relevant cover types in each model, although the core of
relevant cover types is expected to coincide. The function
ListSelectedCoverTypes allows inspecting the selected cover types for
each model. The queries can be categorized by Species, Size, River,
Model.type, Sampled.season, Valid.season, and/or Data.origin. When
evaluating the microhabitat suitability, PredictHabitatSuitability
selects and aggregates the appropriate cover types internally.
Therefore, this function does not need to be called explicitly.
ListSelectedCoverTypes(Species = "Salmo trutta")
#> Code Model Leaves Algae Root
#> 603 ABELZ Salmo.trutta.GAM.Large.Curueno.All TRUE FALSE TRUE
#> 604 ABEMQ Salmo.trutta.HSC.Large.Curueno.All FALSE FALSE TRUE
#> 605 ABEMV Salmo.trutta.NNET.Large.Curueno.All TRUE FALSE TRUE
#> 606 ABEMZ Salmo.trutta.RF.Large.Curueno.All FALSE FALSE TRUE
#> 607 ABENR Salmo.trutta.SVM.Large.Curueno.All TRUE FALSE TRUE
#> 608 ABEMN Salmo.trutta.GAM.Medium.Curueno.All TRUE FALSE FALSE
#> 609 ABEMR Salmo.trutta.HSC.Medium.Curueno.All FALSE FALSE TRUE
#> 610 ABEMW Salmo.trutta.NNET.Medium.Curueno.All FALSE FALSE FALSE
#> 611 ABENO Salmo.trutta.RF.Medium.Curueno.All FALSE FALSE FALSE
#> 612 ABENS Salmo.trutta.SVM.Medium.Curueno.All FALSE TRUE FALSE
#> 613 ABEMO Salmo.trutta.GAM.Small.Curueno.All FALSE FALSE FALSE
#> 614 ABEMS Salmo.trutta.HSC.Small.Curueno.All FALSE FALSE TRUE
#> 615 ABEMX Salmo.trutta.NNET.Small.Curueno.All FALSE FALSE TRUE
#> 616 ABENP Salmo.trutta.RF.Small.Curueno.All TRUE FALSE TRUE
#> 617 ABENT Salmo.trutta.SVM.Small.Curueno.All TRUE TRUE FALSE
#> Aquatic.vegetation Reed Wood Sand Rock Cave Shade
#> 603 TRUE FALSE FALSE FALSE TRUE TRUE TRUE
#> 604 TRUE TRUE TRUE FALSE TRUE TRUE TRUE
#> 605 TRUE TRUE FALSE FALSE FALSE TRUE TRUE
#> 606 TRUE FALSE FALSE FALSE FALSE FALSE TRUE
#> 607 FALSE TRUE FALSE FALSE FALSE FALSE TRUE
#> 608 TRUE FALSE TRUE TRUE FALSE TRUE TRUE
#> 609 TRUE TRUE TRUE FALSE TRUE TRUE TRUE
#> 610 TRUE FALSE TRUE FALSE FALSE TRUE TRUE
#> 611 FALSE FALSE TRUE FALSE FALSE TRUE TRUE
#> 612 FALSE FALSE TRUE TRUE FALSE TRUE TRUE
#> 613 TRUE TRUE FALSE FALSE TRUE TRUE TRUE
#> 614 TRUE TRUE TRUE FALSE TRUE FALSE TRUE
#> 615 TRUE TRUE TRUE FALSE TRUE FALSE TRUE
#> 616 TRUE FALSE FALSE TRUE TRUE TRUE TRUE
#> 617 FALSE TRUE TRUE TRUE TRUE TRUE FALSEThe package includes a number of models (i.e., 797) and species (i.e.,
43) of several size classes and activities. Among other things, the
available models vary by modelling technique, sampling site and the
aggregation of data from different sites. The variable Default
obtained with ListModels indicates which models are recommended
because it was considered they performed and generalised better and/or
were developed with data covering a large range of the microhabitat
variables. Nevertheless, users can select other models when they
consider it preferable. For example, when they are going to evaluate the
microhabitat suitability in the basin where data for an alternative
model was collected. In addition to ListSelectedCoverTypes, users can
carry out sensitivity analyses to compare the predictions of the
different available models. This allows inspecting their performance
before selecting any alternative model or set of models. The sensitivity
analysis can be carried out employing the function
PlotHabitatSuitabilityModels as follows:
## Pseudochondrostoma.polylepis, large, FRBS for Spring, Summer, and Autumn
(Selected.model <- ListModels(Species = "Pseudochondrostoma polylepis", verbose = FALSE)$Codes[1])
#> [1] "ABEIO"
PlotHabitatSuitabilityModels(Selected.model = Selected.model, Quantiles = TRUE)
Figure 1 - Sensitivity analysis (i.e., partial dependence plots) for Pseudochondrostoma polylepis large. The figure shows the results for the Fuzzy Rule-Based System (FRBS) to evaluate the microhabitat suitability during Spring, Summer, and Autumn.
This function generates partial dependence-like plots for microhabitat suitability models based on the provided data and selected model (Friedman 2001). Partial dependence plots are created by varying one variable over its range of values while averaging the model’s predictions over all the other features. This method isolates the effect of that single variable on the outcome and thus determining the overall effect. The function allows employing a user defined dataset to test especific combinations of variable values.
(Selected.model <- ListModels(Species = "Salmo trutta", Model.type = "GAM", verbose = FALSE))
#> $Current.summary.table
#> Code Model River River.short
#> 603 ABELZ Salmo.trutta.GAM.Large.Curueno.All Curueño (Douro) Curueno
#> 608 ABEMN Salmo.trutta.GAM.Medium.Curueno.All Curueño (Douro) Curueno
#> 613 ABEMO Salmo.trutta.GAM.Small.Curueno.All Curueño (Douro) Curueno
#> Model.type Default Species Size Sampled.season
#> 603 GAM TRUE Salmo trutta Large Summer
#> 608 GAM TRUE Salmo trutta Medium Summer
#> 613 GAM TRUE Salmo trutta Small Summer
#> Valid.season Valid.seasons.short Data.origin N.presences
#> 603 Winter; Spring; Summer; Autumn All Field survey 24
#> 608 Winter; Spring; Summer; Autumn All Field survey 81
#> 613 Winter; Spring; Summer; Autumn All Field survey 385
#> N.absences N.interviews.or.studies True.positive False.positive
#> 603 896 NA 24 215
#> 608 839 NA 70 218
#> 613 535 NA 275 210
#> True.negative False.negative Sensitivity Specificity TSS
#> 603 681 0 1.000 0.760 0.760
#> 608 621 11 0.864 0.740 0.604
#> 613 325 110 0.714 0.607 0.322
#> Balanced.accuracy
#> 603 0.880
#> 608 0.802
#> 613 0.661
#>
#> $Models
#> [1] "Salmo.trutta.GAM.Large.Curueno.All" "Salmo.trutta.GAM.Medium.Curueno.All"
#> [3] "Salmo.trutta.GAM.Small.Curueno.All"
#>
#> $Codes
#> [1] "ABELZ" "ABEMN" "ABEMO"
(Selected.model <- Load.and.print.model(Selected.model = Selected.model$Codes[1]))
#> $Selected.cover.types
#> Leaves Algae Root Aquatic.vegetation
#> TRUE FALSE TRUE TRUE
#> Reed Wood Sand Rock
#> FALSE FALSE FALSE TRUE
#> Cave Shade
#> TRUE TRUE
#>
#> $Model.type
#> [1] "GAM"
#>
#> $Model
#>
#> Call: mgcv:::bam(formula = as.formula(paste(Selected.species, "~",
#> C.terms)), family = binomial(), data = C.sample.dataset,
#> weights = ifelse(C.sample.dataset[, Selected.species] ==
#> 1, Chromosome[-c(1:length(Cover.types.complete.list))][1],
#> 100 - Chromosome[-c(1:length(Cover.types.complete.list))][1]),
#> method = "fREML", gamma = 1.2, discrete = TRUE)
#>
#> Coefficients:
#> (Intercept) s(Depth).1 s(Depth).2
#> 1.28199 5.31493 12.54161
#> s(Depth).3 s(Velocity).1 s(Velocity).2
#> 12.44123 0.08129 1.14827
#> s(Velocity).3 s(Substrate.index).1 s(Substrate.index).2
#> 2.00335 -1.70746 -2.38911
#> s(Substrate.index).3 s(Cover).1 s(Cover).2
#> -9.81859 2.96345 5.10960
#> s(Cover).3
#> 4.24585
#>
#> Degrees of Freedom: 939 Total (i.e. Null); 928.0187 Residual
#> Null Deviance: 3891
#> Residual Deviance: 1617 AIC: 1639
#> $Selected.cover.types
#> Leaves Algae Root Aquatic.vegetation
#> TRUE FALSE TRUE TRUE
#> Reed Wood Sand Rock
#> FALSE FALSE FALSE TRUE
#> Cave Shade
#> TRUE TRUE
#>
#> $Model.type
#> [1] "GAM"
#>
#> $Model
#>
#> Call: mgcv:::bam(formula = as.formula(paste(Selected.species, "~",
#> C.terms)), family = binomial(), data = C.sample.dataset,
#> weights = ifelse(C.sample.dataset[, Selected.species] ==
#> 1, Chromosome[-c(1:length(Cover.types.complete.list))][1],
#> 100 - Chromosome[-c(1:length(Cover.types.complete.list))][1]),
#> method = "fREML", gamma = 1.2, discrete = TRUE)
#>
#> Coefficients:
#> (Intercept) s(Depth).1 s(Depth).2
#> 1.28199 5.31493 12.54161
#> s(Depth).3 s(Velocity).1 s(Velocity).2
#> 12.44123 0.08129 1.14827
#> s(Velocity).3 s(Substrate.index).1 s(Substrate.index).2
#> 2.00335 -1.70746 -2.38911
#> s(Substrate.index).3 s(Cover).1 s(Cover).2
#> -9.81859 2.96345 5.10960
#> s(Cover).3
#> 4.24585
#>
#> Degrees of Freedom: 939 Total (i.e. Null); 928.0187 Residual
#> Null Deviance: 3891
#> Residual Deviance: 1617 AIC: 1639Each group of models was developed employing specific R packages:
- Artificial Neural Networks (Multilayer Perceptrons, ANN-MLPs) were
developed with the R package
nnet - Random Forests (RFs) with
ranger - Generalised Additive Models (GAMs) with
mgcv - Support Vector Machines (SVMs) with
e1071 - Habitat Suitability Curves (HSCs) with
FuzzyFishHS(https://github.com/RafaMMas/FuzzyFishHS) - Fuzzy Rule-Based Systems (FRBSs) with
FuzzyFishHS(https://github.com/RafaMMas/FuzzyFishHS)
Therefore, once a specific model has been loaded it is possible to harness the capabilites of each specific package to, for example, plot directly the selected model.
library(mgcv)
#> Warning: package 'mgcv' was built under R version 4.1.3
#> Loading required package: nlme
#> This is mgcv 1.8-42. For overview type 'help("mgcv-package")'.
par(mar = c(4,4,1,1), bty = "n")
plot.gam(Selected.model$Model, pages = 1, shade=TRUE, shade.col = "dodgerblue")
Figure
2 - Plot of the Generalised Additive Model for Salmo trutta Large
obtained with data collected in the Curueño River (Douro) and valid all
year round.
The different option to, for example, plot each models depend on each
specific package and in some cases suchs the SVMs obtained with e1071
require indirect approaches for plotting. For instance, using the R
package pdp (Greenwell 2017).
Habitat suitability predictions are carried out based on mean flow velocity (m/s), water depth (m) substrate index (-) and the sum of the number of relevant types of cover in the evaluated microhabitats (i.e. pixels or cells from the hydraulic simulation). The package includes functions to prepare the input data. Additionally, the package includes data on a typical output from a hydraulic simulation and field survey. In particular, it includes data on velocity and depth for 40 simulated flows and the substrate composition and cover availability in the site of the Serpis River Basin (Eastern Iberian Peninsula) studied by Muñoz-Mas et all (2024).
data("Velocity.example.df")
data("Depth.example.df")
data("Substrate.index.example.df")
data("Cover.example.df")
summary(Velocity.example.df[,1:6])
#> x y Velocity.0.05 Velocity.0.1
#> Min. :732536 Min. :4302306 Min. :0.00000 Min. :0.000000
#> 1st Qu.:732596 1st Qu.:4302440 1st Qu.:0.00000 1st Qu.:0.000000
#> Median :732614 Median :4302560 Median :0.00000 Median :0.000000
#> Mean :732615 Mean :4302545 Mean :0.00439 Mean :0.009547
#> 3rd Qu.:732638 3rd Qu.:4302660 3rd Qu.:0.00480 3rd Qu.:0.010900
#> Max. :732686 Max. :4302710 Max. :0.15973 Max. :0.235300
#> Velocity.0.2 Velocity.0.3
#> Min. :0.00000 Min. :0.00000
#> 1st Qu.:0.00000 1st Qu.:0.00000
#> Median :0.00000 Median :0.00020
#> Mean :0.02132 Mean :0.03182
#> 3rd Qu.:0.02330 3rd Qu.:0.03495
#> Max. :0.46400 Max. :1.10900
summary(Depth.example.df[,1:6])
#> x y Depth.0.05 Depth.0.1
#> Min. :732536 Min. :4302306 Min. :0.0001 Min. :0.0001
#> 1st Qu.:732596 1st Qu.:4302440 1st Qu.:0.1022 1st Qu.:0.0986
#> Median :732614 Median :4302560 Median :0.2598 Median :0.2629
#> Mean :732615 Mean :4302545 Mean :0.3001 Mean :0.3008
#> 3rd Qu.:732638 3rd Qu.:4302660 3rd Qu.:0.4631 3rd Qu.:0.4587
#> Max. :732686 Max. :4302710 Max. :1.2471 Max. :1.2837
#> NA's :1917 NA's :1712
#> Depth.0.2 Depth.0.3
#> Min. :0.0002 Min. :0.0002
#> 1st Qu.:0.1005 1st Qu.:0.1087
#> Median :0.2808 Median :0.2960
#> Mean :0.3147 Mean :0.3321
#> 3rd Qu.:0.4729 3rd Qu.:0.4987
#> Max. :1.3332 Max. :1.3676
#> NA's :1548 NA's :1482
summary(Substrate.index.example.df[,-c(1:2)])
#> Vegetation Silt Sand Fine.gravel Gravel
#> Min. :0 Min. : 0.00 Min. :0 Min. : 0.00 Min. : 0.00
#> 1st Qu.:0 1st Qu.: 0.00 1st Qu.:0 1st Qu.: 0.00 1st Qu.: 0.00
#> Median :0 Median : 15.00 Median :0 Median : 0.00 Median : 0.00
#> Mean :0 Mean : 20.59 Mean :0 Mean : 10.49 Mean : 15.61
#> 3rd Qu.:0 3rd Qu.: 50.00 3rd Qu.:0 3rd Qu.: 0.00 3rd Qu.: 20.00
#> Max. :0 Max. :100.00 Max. :0 Max. :100.00 Max. :100.00
#> Cobbles Boulders Bed.rock
#> Min. : 0.00 Min. : 0.00 Min. : 0.000
#> 1st Qu.: 0.00 1st Qu.: 0.00 1st Qu.: 0.000
#> Median : 50.00 Median : 0.00 Median : 0.000
#> Mean : 38.49 Mean : 10.25 Mean : 4.571
#> 3rd Qu.: 50.00 3rd Qu.: 0.00 3rd Qu.: 0.000
#> Max. :100.00 Max. :100.00 Max. :100.000
summary(Cover.example.df[,-c(1:2)])
#> Leaves Algae Root Aquatic.vegetation Reed
#> Min. :0 Min. :0 Min. :0 Min. :0.00000 Min. :0.00000
#> 1st Qu.:0 1st Qu.:0 1st Qu.:0 1st Qu.:0.00000 1st Qu.:0.00000
#> Median :0 Median :0 Median :0 Median :0.00000 Median :0.00000
#> Mean :0 Mean :0 Mean :0 Mean :0.01703 Mean :0.01136
#> 3rd Qu.:0 3rd Qu.:0 3rd Qu.:0 3rd Qu.:0.00000 3rd Qu.:0.00000
#> Max. :0 Max. :0 Max. :0 Max. :1.00000 Max. :1.00000
#> Wood Sand Rock Cave Shade
#> Min. :0.0000 Min. :0 Min. :0.0000 Min. :0.00000 Min. :0.00
#> 1st Qu.:0.0000 1st Qu.:0 1st Qu.:0.0000 1st Qu.:0.00000 1st Qu.:0.00
#> Median :0.0000 Median :0 Median :0.0000 Median :0.00000 Median :1.00
#> Mean :0.4767 Mean :0 Mean :0.1612 Mean :0.02177 Mean :0.53
#> 3rd Qu.:1.0000 3rd Qu.:0 3rd Qu.:0.0000 3rd Qu.:0.00000 3rd Qu.:1.00
#> Max. :1.0000 Max. :0 Max. :1.0000 Max. :1.00000 Max. :1.00The function SubstrateIndex computes the substrate index as a weighted
aggregation of the percentage of different granulometry classes (%). The
substrate classes originally used corresponded to a simplification of
the the American Geophysical Union size scale, namely silt (Ø ≤ 62 µm),
sand (62 µm > Ø ≤ 2 mm), fine gravel (2 > Ø ≤ 8 mm), gravel (8 > Ø ≤
64 mm), cobbles (64 > Ø ≤ 256 mm), boulders (Ø > 256 mm) and bedrock
Muñoz-Mas et al. (2017). The package includes some examples to
illustrate the structure of the input files.
data(Substrate.index.example.df)
summary(Substrate.index.example.df)
#> x y Vegetation Silt Sand
#> Min. :732536 Min. :4302306 Min. :0 Min. : 0.00 Min. :0
#> 1st Qu.:732596 1st Qu.:4302440 1st Qu.:0 1st Qu.: 0.00 1st Qu.:0
#> Median :732614 Median :4302560 Median :0 Median : 15.00 Median :0
#> Mean :732615 Mean :4302545 Mean :0 Mean : 20.59 Mean :0
#> 3rd Qu.:732638 3rd Qu.:4302660 3rd Qu.:0 3rd Qu.: 50.00 3rd Qu.:0
#> Max. :732686 Max. :4302710 Max. :0 Max. :100.00 Max. :0
#> Fine.gravel Gravel Cobbles Boulders
#> Min. : 0.00 Min. : 0.00 Min. : 0.00 Min. : 0.00
#> 1st Qu.: 0.00 1st Qu.: 0.00 1st Qu.: 0.00 1st Qu.: 0.00
#> Median : 0.00 Median : 0.00 Median : 50.00 Median : 0.00
#> Mean : 10.49 Mean : 15.61 Mean : 38.49 Mean : 10.25
#> 3rd Qu.: 0.00 3rd Qu.: 20.00 3rd Qu.: 50.00 3rd Qu.: 0.00
#> Max. :100.00 Max. :100.00 Max. :100.00 Max. :100.00
#> Bed.rock
#> Min. : 0.000
#> 1st Qu.: 0.000
#> Median : 0.000
#> Mean : 4.571
#> 3rd Qu.: 0.000
#> Max. :100.000
head(SubstrateIndex(Substrate.index.example.df, check.completeness = FALSE)) # check.completeness allows inspecting whether names and percentages are correct.
#> Substrate.index
#> 1 3
#> 2 3
#> 3 3
#> 4 3
#> 5 3
#> 6 3
Substrate.index <- SubstrateIndex(Substrate.index.example.df, check.completeness = FALSE)Selected.species <- "Cobitis paludica"
Selected.size <- "Large"
(Selected.models <- ListModels(Species = Selected.species, Size = Selected.size)) ## example determining the selected models
#> Code Model
#> 138 ABCKQ Cobitis.paludica.FRBS.Large.EKorL.Spring.Summer.Autumn
#> 139 ABCKP Cobitis.paludica.FRBS.Large.EKorL.Autumn.Winter.Spring
#> 161 ABCKU Cobitis.paludica.GAM.Large.Estena.and.Yeguas.All
#> 162 ABCLQ Cobitis.paludica.HSC.Large.Estena.and.Yeguas.All
#> 163 ABCMN Cobitis.paludica.NNET.Large.Estena.and.Yeguas.All
#> 164 ABCMV Cobitis.paludica.RF.Large.Estena.and.Yeguas.All
#> 165 ABCNR Cobitis.paludica.SVM.Large.Estena.and.Yeguas.All
#> River River.short Model.type
#> 138 EKorL FRBS
#> 139 EKorL FRBS
#> 161 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas GAM
#> 162 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas HSC
#> 163 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas NNET
#> 164 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas RF
#> 165 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas SVM
#> Default Species Size Sampled.season
#> 138 TRUE Cobitis paludica Large
#> 139 TRUE Cobitis paludica Large
#> 161 TRUE Cobitis paludica Large Summer
#> 162 TRUE Cobitis paludica Large Summer
#> 163 TRUE Cobitis paludica Large Summer
#> 164 TRUE Cobitis paludica Large Summer
#> 165 TRUE Cobitis paludica Large Summer
#> Valid.season Valid.seasons.short Data.origin
#> 138 Spring; Summer; Autumn Spring.Summer.Autumn Expert knowledge
#> 139 Autumn; Winter; Spring Autumn.Winter.Spring Expert knowledge
#> 161 Winter; Spring; Summer; Autumn All Field survey
#> 162 Winter; Spring; Summer; Autumn All Field survey
#> 163 Winter; Spring; Summer; Autumn All Field survey
#> 164 Winter; Spring; Summer; Autumn All Field survey
#> 165 Winter; Spring; Summer; Autumn All Field survey
#> N.presences N.absences N.interviews.or.studies True.positive False.positive
#> 138 NA NA 7 NA NA
#> 139 NA NA 7 NA NA
#> 161 43 1950 NA 39 1200
#> 162 43 1953 NA NA NA
#> 163 43 1950 NA 38 959
#> 164 43 1950 NA 38 959
#> 165 43 1950 NA 36 678
#> True.negative False.negative Sensitivity Specificity TSS
#> 138 NA NA NA NA NA
#> 139 NA NA NA NA NA
#> 161 750 4 0.907 0.385 0.292
#> 162 NA NA NA NA NA
#> 163 991 5 0.884 0.508 0.392
#> 164 991 5 0.884 0.508 0.392
#> 165 1272 7 0.837 0.652 0.490
#> Balanced.accuracy
#> 138 NA
#> 139 NA
#> 161 0.646
#> 162 NA
#> 163 0.696
#> 164 0.696
#> 165 0.745
#> $Current.summary.table
#> Code Model
#> 138 ABCKQ Cobitis.paludica.FRBS.Large.EKorL.Spring.Summer.Autumn
#> 139 ABCKP Cobitis.paludica.FRBS.Large.EKorL.Autumn.Winter.Spring
#> 161 ABCKU Cobitis.paludica.GAM.Large.Estena.and.Yeguas.All
#> 162 ABCLQ Cobitis.paludica.HSC.Large.Estena.and.Yeguas.All
#> 163 ABCMN Cobitis.paludica.NNET.Large.Estena.and.Yeguas.All
#> 164 ABCMV Cobitis.paludica.RF.Large.Estena.and.Yeguas.All
#> 165 ABCNR Cobitis.paludica.SVM.Large.Estena.and.Yeguas.All
#> River River.short Model.type
#> 138 EKorL FRBS
#> 139 EKorL FRBS
#> 161 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas GAM
#> 162 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas HSC
#> 163 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas NNET
#> 164 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas RF
#> 165 Estena (Guadiana) + Yeguas (Guadalquivir) Estena.and.Yeguas SVM
#> Default Species Size Sampled.season
#> 138 TRUE Cobitis paludica Large
#> 139 TRUE Cobitis paludica Large
#> 161 TRUE Cobitis paludica Large Summer
#> 162 TRUE Cobitis paludica Large Summer
#> 163 TRUE Cobitis paludica Large Summer
#> 164 TRUE Cobitis paludica Large Summer
#> 165 TRUE Cobitis paludica Large Summer
#> Valid.season Valid.seasons.short Data.origin
#> 138 Spring; Summer; Autumn Spring.Summer.Autumn Expert knowledge
#> 139 Autumn; Winter; Spring Autumn.Winter.Spring Expert knowledge
#> 161 Winter; Spring; Summer; Autumn All Field survey
#> 162 Winter; Spring; Summer; Autumn All Field survey
#> 163 Winter; Spring; Summer; Autumn All Field survey
#> 164 Winter; Spring; Summer; Autumn All Field survey
#> 165 Winter; Spring; Summer; Autumn All Field survey
#> N.presences N.absences N.interviews.or.studies True.positive False.positive
#> 138 NA NA 7 NA NA
#> 139 NA NA 7 NA NA
#> 161 43 1950 NA 39 1200
#> 162 43 1953 NA NA NA
#> 163 43 1950 NA 38 959
#> 164 43 1950 NA 38 959
#> 165 43 1950 NA 36 678
#> True.negative False.negative Sensitivity Specificity TSS
#> 138 NA NA NA NA NA
#> 139 NA NA NA NA NA
#> 161 750 4 0.907 0.385 0.292
#> 162 NA NA NA NA NA
#> 163 991 5 0.884 0.508 0.392
#> 164 991 5 0.884 0.508 0.392
#> 165 1272 7 0.837 0.652 0.490
#> Balanced.accuracy
#> 138 NA
#> 139 NA
#> 161 0.646
#> 162 NA
#> 163 0.696
#> 164 0.696
#> 165 0.745
#>
#> $Models
#> [1] "Cobitis.paludica.FRBS.Large.EKorL.Spring.Summer.Autumn"
#> [2] "Cobitis.paludica.FRBS.Large.EKorL.Autumn.Winter.Spring"
#> [3] "Cobitis.paludica.GAM.Large.Estena.and.Yeguas.All"
#> [4] "Cobitis.paludica.HSC.Large.Estena.and.Yeguas.All"
#> [5] "Cobitis.paludica.NNET.Large.Estena.and.Yeguas.All"
#> [6] "Cobitis.paludica.RF.Large.Estena.and.Yeguas.All"
#> [7] "Cobitis.paludica.SVM.Large.Estena.and.Yeguas.All"
#>
#> $Codes
#> [1] "ABCKQ" "ABCKP" "ABCKU" "ABCLQ" "ABCMN" "ABCMV" "ABCNR"Hydraulics <- data.frame(Velocity = Velocity.example.df$Velocity.0.05,
Depth = Depth.example.df$Depth.0.05,
Substrate.index = Substrate.index,
Cover.example.df[,-c(1:2)])
Predictions <- PredictHabitatSuitability(Selected.models = Selected.models, data = Hydraulics, HSC.aggregation = "geometric")summary(Predictions)
#> ABCKQ ABCKP ABCKU ABCLQ
#> Min. :0.0000 Min. :0.0000 Min. :0.1682 Min. :0.0164
#> 1st Qu.:0.2603 1st Qu.:0.0663 1st Qu.:0.3933 1st Qu.:0.4832
#> Median :0.7912 Median :0.4358 Median :0.5730 Median :0.6517
#> Mean :0.5840 Mean :0.4268 Mean :0.5380 Mean :0.5834
#> 3rd Qu.:0.8149 3rd Qu.:0.7942 3rd Qu.:0.6585 3rd Qu.:0.6684
#> Max. :0.9403 Max. :0.9420 Max. :0.9192 Max. :0.9749
#> NA's :1917 NA's :1917 NA's :1917 NA's :1917
#> ABCMN ABCMV ABCNR
#> Min. :0.0434 Min. :0.0000 Min. :0.0000
#> 1st Qu.:0.3367 1st Qu.:0.1098 1st Qu.:0.0000
#> Median :0.5309 Median :0.3065 Median :0.6000
#> Mean :0.4648 Mean :0.3356 Mean :0.5188
#> 3rd Qu.:0.6013 3rd Qu.:0.5514 3rd Qu.:1.0000
#> Max. :0.9868 Max. :0.9271 Max. :1.0000
#> NA's :1917 NA's :1917 NA's :1917pairs(Predictions)
library(sf)
#> Warning: package 'sf' was built under R version 4.1.3
#> Linking to GEOS 3.10.2, GDAL 3.4.1, PROJ 7.2.1; sf_use_s2() is TRUE
Coordinates <- st_as_sf(Velocity.example.df[,c("x", "y")], coords = c("x", "y"))
Colors.suitability <- colorRampPalette(c("red2", "darkorange", "gold", "green2", "dodgerblue"))(100)
# sf_points <- st_as_sf(data.frame(Example.hydrulic.simulation.depth[,c("x", "y")], Predictions), coords = c("x", "y"), crs = 32630)
Model.names <- sapply(colnames(Predictions), function(x){
x <- gsub(x = x, pattern = paste0(gsub(Selected.species, pattern = " ", replacement = "."), "."), replacement = "")
x <- gsub(x = x, pattern = paste0(Selected.size, "."), replacement = "")
x
})
op <- par(mfrow = c(ceiling(length(Selected.models$Models)^0.5), ceiling(length(Selected.models$Models)^0.5)), oma = c(0.5,0.5,2.5,0.5), mar = c(4,4,2,0.5))
for(i in 1:ncol(Predictions))
{
plot(st_geometry(Coordinates), pch = 15, cex = 0.5, bty = "n",
col = Colors.suitability[cut(Predictions[,i], breaks = seq(0, 1, length = 101), labels = 1:100, include.lowest = T)],
xlab = "X", ylab = "Y")
Axis(side = 1, cex.axis = 0.5)
Axis(side = 2, cex.axis = 0.5)
title(Model.names[i], font.main = 1, cex.main = 0.9)
}
par(op)
Species <- Selected.species
Size <- paste0(" - ", Selected.size)
title(substitute(expr = paste(italic(Species), Size, sep =" "), env = list(Species = Species, Size = Size)), outer = T, adj = 0.05, cex.main = 2)
par(op)Eugster, Manuel J. A., Friedrich Leisch, and Carolin Strobl. 2014. “(Psycho-)analysis of benchmark experiments: a formal framework for investigating the relationship between data sets and learning algorithms.” Computational Statistics & Data Analysis 71 (March): 986–1000. https://doi.org/10.1016/j.csda.2013.08.007.
Friedman, J H. 2001. “Greedy function approximation: A gradient boosting machine.” Annals of Statistics 29 (5): 1189–1232. https://doi.org/10.1214/aos/1013203451.
Greenwell, Brandon M. 2017. “pdp: An R Package for Constructing Partial Dependence Plots.” The R Journal 9 (1): 421–36. https://doi.org/10.32614/RJ-2017-016.
Muñoz-Mas, R, Rui Manuel Soares Costa, Francisco Martínez-Capel, and Juan Diego Alcaraz-Hernández. 2017. “Microhabitat competition between Iberian fish species and the endangered Júcar nase (Parachondrostoma arrigonis Steindachner, 1866).” Journal of Ecohydraulics 0 (0): 1–23. https://doi.org/10.1080/24705357.2016.1276417.
Muñoz-Mas, R, H Macian-Sorribes, F J Oliva-Paterna, L Sangelantoni, D Peano, M Pulido-Velazquez, and F Martínez-Capel. 2024. “Adaptation measures to global change in the Serpis River Basin (Spain): An evaluation considering agricultural benefits, environmental flows, and invasive fishes.” Ecological Indicators 161: 111979. https://doi.org/https://doi.org/10.1016/j.ecolind.2024.111979.
Muñoz-Mas, R, P Marcos-Garcia, A Lopez-Nicolas, F J Martínez-García, M Pulido-Velazquez, and F Martínez-Capel. 2018. “Combining literature-based and data-driven fuzzy models to predict brown trout (<i>Salmo trutta</i> L.) spawning habitat degradation induced by climate change.” Ecological Modelling 386: 98–114. https://doi.org/10.1016/j.ecolmodel.2018.08.012.