Diapositive 1

1
Factors influencing wild boar presence in
agricultural landscape a habitat suitability
modelling approach
Kevin Morelle Lejeune Philipppe
9th International Symposium on Wild Boar and
others Suids, Hannover 2012
2
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Wild boar (Sus scrofa) populations have increased
worldwide In parallel, distribution of the
species has enlarged, out of forest habitat ?
plasticity of the species can explain partly the
phenomenon Ability to make  home range shift 
Keuling et al. 2009 Consequently, agricultural
areas have become new  home  for wild boar,
providing cover and food
3
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Cultural cycle offers cover all over the year for
wild boar
4
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Why modelling distribution? Habitat management
policy Park at al. 2003 Conservation planning
Park at al. 2003 Species invasion Evangelista
et al. 2008 Forecast distribution (climate
change) Risk mapping - damage Saito et al.
2012 - disease transmission
Nexton-Cross et al. 2007 ? Give informations
on environmental correlates influencing the
patterns of distribution of a species
5
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Situation in Belgium
6
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
What are main drivers of wild boar distribution
in these agricultural landscape?
1 - identifying environmental variables that
explain seasonal distribution of the species 2 -
defining habitat suitability map in agricultural
landscape 3 - extrapolate the best model to the
north of Wallonia
7
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
STUDY AREA

• We used Condroz as study site to build
• our model
• agricultural area with patchily distributed
  forest
•  recently  (10-30 y) colonized by wild boar

8
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
DATASETS

• 2  presence  datasets agricultural damages
  hunting records
• covering same period (2009-2010)
• differences within year (april-october vs.
  october-december)

9
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
PREDICTORS
Set of 18 predictors defining habitat,
agricultural cover, topography and human
presence cell size of 300m (and landscape
metrics) were derivated using R packages raster
(Hijmans), SpatStat (Baddeley) and dismo.
Environmental predictors are represented as
raster thematic layers.
10
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
MODELING TECHNIQUE MaxEnt Phillips et al. 2006
MaxEnt is a program for modelling species
distribution from presence-only data ? minimizing
the entropy between two probability density,
presence background
From Elith et al. (2011)
11
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
MODELING TECHNIQUE MaxEnt Phillips et al. 2006
Training data to fit the model Test data to
evaluate the predictive ability of the model
(20) Background sample of 2000 points
hunting/damage records
Model evaluation receiver operating
characteristic (ROC) - Area under curve (AUC) ?
measure of the prediction success ? ROC curve is
obtained by plotting all true positive values
(sensitivity fraction) against their equivalent
false positive values (1-speci?city fraction)
12
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Hunting data
13
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Hunting data Response curve of distance to forest
variables
14
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Damage data Response curve
15
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Damage data - Response curves
Habitat
Cover fields
Road density
Potato fields
16
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Both dataset
17
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Both dataset Response curves
Road density
Distance to forest
18
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model evaluation Classical ROC curve analysis
AUC
19
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
20
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection Comparison with known presence
of wild boar
21
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
 Hunting model 
 Damage model 
22
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
 Both model 
 Damage model 
23
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection How to fix a probability
threshold to create a presence/absence map? ?
Theoritically maximizing sensitivity while
minimizing specificity Philips 2006
24
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection How to fix a probability
threshold to create a presence/absence map? ?
BUT to conservative approach! (175 km² of
predicted area vs. already 250 km² of presence
area)
25
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection How to fix a probability
threshold to create a presence/absence map? ?
BUT to conservative approach! (175 km² of
predicted area vs. already 250 km² of presence
area)
26
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
Current species range could increase up to 535
km² if wild boar occupies all the areas predicted
as suitable by the MaxEnt model
27
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
Current species range could increase up to 1116
km² if wild boar occupies all the areas predicted
as suitable by the MaxEnt model
28
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Model projection
35 km
Current species range could increase up to 879
km² if wild boar occupies all the areas predicted
as suitable by the MaxEnt model
29
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
Factors analysis Distribution model show
differences in environmental covariates
between ? autumn/winter decrease in cover/food
in agricultural plain acorn availability
switch to forest habitat after crop harvesting ?
spring/summer intensive use of fields providing
cover food BUTreliability of presence model
for a highly mobile species? How to take into
account movement ability of the wild boar? Model
prediction/projection Prediction show that range
could increase into suitable clustered patches
? now hunting pressure is high and maintain
population low, but ?
30
DISCUSSION
RESULTS
METHOD
OBJECTIVES
CONTEXT
References Evangelista, P. H., S. Kumar, T. J.
Stohlgren, C. S. Jarnevich, A. W. Crall, J. B.
Norman Iii, and D. T. Barnett. 2008. Modelling
invasion for a habitat generalist and a
specialist plant species. Diversity and
Distributions 14808-817. Mateo-Tomás, P. and P.
P. Olea. 2010. Anticipating Knowledge to Inform
Species Management Predicting Spatially Explicit
Habitat Suitability of a Colonial Vulture
Spreading Its Range. PLoS ONE 5e12374. Newton-Cro
ss, G., P. C. L. White, and S. Harris. 2007.
Modelling the distribution of badgers Meles
meles comparing predictions from field-based and
remotely derived habitat data. Mammal Review
3754-70. Park, C.-R. and W.-S. Lee. 2003.
Development of a GIS-based habitat suitability
model for wild boar Sus scrofa in the Mt.
Baekwoonsan region, Korea. Mammal Study
2817-21. Phillips, S. J., R. P. Anderson, and R.
E. Schapire. 2006. Maximum entropy modeling of
species geographic distributions. Ecological
Modelling 190231-259. Saito, M., H. Momose, T.
Mihira, and S. Uematsu. 2012. Predicting the risk
of wild boar damage to rice paddies using
presence-only data in chiba prefecture, Japan.
International Journal of Pest Management
5865-71.
31
Thank you for your attention
P. Taymans