Marine Species Distributions: From data to predictive models Samuel Bosch
Marine Species Distributions: From data to predictive models
Samuel Bosch
Topics
• Introduction
• Invasive seaweeds
• Marine species distribution modelling
• Some future perspectives
Oceans • 70% of area • 40% of ecosystem value • 25% of species richness • > 200,000 registered species
Threats
Pollution
Overexploitation
Invasive species
Global climate change
© Hugo Ahlenius, UNEP/GRID-Arenda, 2008
Invasive marine species
Invasive seaweeds
Undaria pinnatifida Sargassum muticum Codium fragile
Caulerpa taxifolia Asparagopsis armata Dasysiphonia japonica
Introduction rate
Curated list of 153 introduced seaweed species in Europe
Introduction rate
Species Records
Introduction rate
Species Records
Invasive seaweeds: Vectors
Hull Fouling
Aquaculture
Suez Canal
a tale from
Monaco
Aquaria ?
and its ecological
conse-quence
Aquaria ?
Aquaria ?
Sampling
• 217 samples • 135 species
• 6 invasive or introduced • 40 possibly invasive
Present 2055
• Rich species diversity • Invasive species • Potential for new introductions
More …
• Chapter 5
Bosch, S., De Clerck, O. and Frédéric Mineur, F. Spatio-temporal patterns of introduced seaweeds in European waters, a critical review.
• Chapter 6
Vranken, S., Bosch, S., Peña, V., Leliaert, F., Mineur, F. and De Clerck, O. A risk assessment of aquarium trade introductions of seaweed in European waters.
Marine species distribution modelling
Image credit: Université de Lausanne
Species distribution modelling (SDM)
Species field observations
Environmental data Model fitting Predicted species
distributions
Ecological Niche
Hutchinson (1957) “… the hypervolume defined by the environmental dimensions within which that species can survive and reproduce.”
Abiotic
Movement
Biotic
GO
GI
Geographic area
Environmental data
Occurrences
SDM algorithm
Model
Absences
Output
Environmental data
Occurrences
SDM algorithm
Model
Absences
Output
Occurrences: Database
701 million occurrences
48.4 million occurrences of 123,287 marine species
Occurrences
But:
• Spatially uneven sampling and reporting
Occurrences
Himanthalia elongatha
Aiello-Lammens, M. E. et al. 2015. spThin: an R package for spatial thinning of species occurrence records for use in ecological niche models. - Ecography (Cop.). 38: 541–545.
Occurrences
But:
• Spatially uneven sampling and reporting
• Errors
– Taxonomic
• Misidentifications
• [cryptic] species complexes
– Geographic
• Typo’s, 0,0, generated coordinates, ….
Occurrences: (Eur)OBIS QC
Indicate the completeness and correctness
• Taxonomic
• Geographic
• Outliers
• Additional fields such as abundance
Occurrences: (Eur)OBIS QC
Outlier analysis on the dataset ‘ICES Biological community’
Environmental data
Occurrences
SDM algorithm
Model
Absences
Output
Absences
• Presence-only SDM
– Only presences
Absences
• Presence-only SDM
1. Only presences
2. Pseudo-absences
Environmental data
Occurrences
SDM algorithm
Model
Absences
Output
Environmental data
Salinity Bathymetry
Temperature Chlorophyll a
sdmpredictors
library(sdmpredictors)
# view all available layers
View(list_layers())
# load SST mean from Bio-ORACLE and
# bathymetry from MARSPEC as lat/lon data
x <- load_layers(c("BO_sstmean","MS_bathy_5m"),
equalarea = FALSE)
Which one ?
• Calcite • Chlorophyll A • Cloud fraction • Diffuse attenuation
coefficient at 490 nm • Dissolved oxygen • Nitrate • Photosynthetically
available radiation • pH • Phosphate
• Salinity • Silicate • Sea surface temperature • Bathymetry • East/West aspect • North/South Aspect • Plan curvature • Profile curvature • Distance to shore • Bathymetric slope • Concavity
library(marinespeed) # list all 514 species species <- list_species() view(species) help(marinespeed)
MarineSPEED
Predictor relevance
Predictor relevance
0
25
50
75
100
Sh
ore
dis
tan
ce
Ba
thym
etr
y
SS
T (
ran
ge
)
Sa
linity
Ca
lcite
pH
Ch
loro
ph
yll
a (
me
an
)
Ch
loro
ph
yll
a (
min
)
Ch
loro
ph
yll
a (
ma
x)
Ch
loro
ph
yll
a (
ran
ge
)
Diffu
se
atte
nu
atio
n (
me
an
)
Diffu
se
atte
nu
atio
n (
min
)
Diffu
se a
ttenuation (
max)
SS
T (
mean)
PA
R (
me
an
)
PA
R (
ma
x)
Ph
osp
ha
te
Nitra
te
Sili
ca
te
In s
pe
cie
s to
p 5
(%
)
Statistical variation
Biological variation
Environmental data
Occurrences
SDM algorithm
Model selection
Absences
Output
SDM algorithm
Model selection
metric
Validation dataset
Random Spatial
AUC Boyce
Kappa AIC
MaxEnt Random forests
GRaF
GLM
GAM
GARP
Visual
BIOCLIM
Ensemble
BRT
MARS Temporal
Environmental data
Occurrences
SDM algorithm
Model
Absences
Output
Output
• Maps
Output
• Response curves
Can we predict invasive seaweeds?
Abiotic
Movement
Biotic
GO
GI
Geographic area
Sargassum muticum
Codium fragile
Dictyota cyanoloma
Grateloupia turuturu
Undaria pinnatifida
Can we predict invasive seaweeds?
Can we predict invasive seaweeds?
Native Invasive
European Invasive non-European
1971
1941
Sargassum muticum
Can we predict invasive seaweeds?
Modelling in 1970
Sargassum muticum model fitted only with native records
Can we predict invasive seaweeds?
Modelling in 1970
Sargassum muticum model fitted with native records and Californian invasive records from before the European introduction
Europe in 2100 ?
Predicted changes in the range of 15 invasive seaweeds in Europe by 2100
Uncertainty
Uncertainty in the predicted ranges of 15 invasive seaweeds
More …
• Chapter 2 Vandepitte, L. et al. 2015. Fishing for data and sorting the catch: assessing the data quality, completeness and fitness for use of data in marine biogeographic databases. - Database
• Chapter 3 Bosch, S., Tyberghein, L., De Clerck, O. sdmpredictors: an R package for species distribution modelling predictor datasets
• Chapter 4 Bosch, S., Tyberghein, L., Deneudt, K., Hernandez, F., De Clerck, O. In search of relevant predictors for marine species distribution modelling using the MarineSPEED benchmark dataset
• Chapter 7 Bosch, S., Gomez Giron, E., Martínez, B., De Clerck, O. Modelling the past, present and future distribution of invasive seaweeds in Europe
Future perspectives
Future perspectives
• Traits data in WoRMS
Future perspectives
• New data in OBIS
Future perspectives
• Bio-ORACLE 2: including benthic layers
Surface layer
Difference between surface and benthic layer
Future perspectives
• Biotic interactions and knowledge transfer
Future perspectives
• Use MarineSPEED to study other aspects of SDM
Acknowledgement
The Great Wave off Kanagawa
“All models are wrong, but some are useful” – George Box