Patterns and drivers of alien plant invasions · PDF filePatterns and drivers of alien plant invasions ... •Lantana camara – 3,021 ... Slide 1 Author: Neil Created Date:

Patterns and drivers of

alien plant invasions

LC Foxcroft, S MacFadyen,

P Pyšek, DM Richardson, M Rouget, C Hui

IAP distribution

Species characteristics / traits

• Propagule production

• Dispersal mode

• Resource demand

• Maximum growth rate

• Defences

• Competitiveness

• Seed-banking

• Seed size

Habitat susceptibility

• Bioclimatic suitability

• Resource availability, including disturbance

regime, fire regime, empty niche

• Local vector presence

• Predation / consumer pressure; including

release from natural enemies

• Heterogeneity (patchiness/gaps)

System context

• Pathways

• Propagule pressure

• Continuous influx from seed source

• Vector efficacy

• Global climate change (e.g. N deposition,

increased CO2, or altered rainfall patterns)

• Residence time

• Human values and perceptions

I. Veg/climate template

• # of records /cell

• % of species as

proportion of all

records /cell

• Relative

abundance

estimate River discharge

Rainfall

Distance to rivers

Distance to camp/ staff villages

by size of footprint

Distance to road & type

NDVI

Distance to closest boundary

Temperature

Fire frequency

Elephant density

Tree cover

Ungulate density II. Disturbance

Distance to nearest waterhole

III. Propagule

pressure/pathways

Objectives

• Separate & quantify

• by understanding the

relationship

• current distribution of

selected alien plant species

• help us gain understanding

of arrangement in the

landscape

• Topography

• Geology

• Soil

• Rainfall

• Temperature

• Habitat/vegetation

• Fauna

• Disturbance

• Fire

• Pathways

Abiotic template,

habitat features,

human drivers

Objectives cont..

• Compare these patterns over different

spatial scales

Data collection

• ~ 30,000 alien plant records

• Cybertracker & other sources

• Opuntia stricta – 20,808

• Lantana camara – 3,021

• Opuntia spp. – 1,493

• Xanthium spp. – 597

• Parthenium hysterophorus - 204

• MaxEnt

• Later – SEM & CART

14

Variables

Water

River density

River distance

Distribution in landscape

Climate

Long-term max temperature

Long-term min temperature

Long-term mean annual rainfall

NDVI

long-term standard dev.

long-term mean

Physical landscape

Landtype

Soils

Topography (20m DEM)

Disturbance

Density of roads

Distance to roads

Fire

MaxEnt

• 25% random records test data

• model species distributions from

presence-only species records

• estimate the relationship between

species records at sites and the

environmental and/or spatial

characteristics of those sites

Potential limitations

• data quality

• amount and accuracy of species data

• ecological relevance of predictor variables

• availability of information on disturbances

• dispersal limitations and biotic interactions

• modelling method

• scale of analysis

• Results in context of alien species

modeling (source, spread, etc)

Lantana camara

Variables

River density / distance

Max temp.

Mean annual rainfall

Xanthium spp.

Variables

River density / distance

Landscape water distribution

Opuntia spp.

Variables

Soils

River density

Min temp.

Opuntia stricta

Variables

Max temp.

Min temp.

Fire

Mean annual rainfall

Parthenium

hysterophorus

Variables

Distance to road

20 m DEM

Land type

Mean annual rainfall

Results

• L. camara:

• rivers main environmental variable

• also major pathway of continuous

invasion (circular problem)

• Xanthium spp.:

• rivers & other water sources (e.g. pans)

• annuals – disturbed areas

• O. stricta:

• invasion history overrides environmental

• P. hysterophorus:

• invasion history

• roadsides (disturbed)

• cars as vectors

Interpretation of Results

• AUCs vary depending whether a

species is widespread or narrowly

distributed, which in alien plant

modeling is confounded by spread

/dispersal mechanisms

• The choice of the “geographic

window of extent” for modeling

matters

• AUCs are typically inflated due to

spatial autocorrelation

NEXT..…

• Revisit modeling decisions and explore

how different setting choices affect

results e.g.

• explore additional covariates

• background data

• feature selection

• regularization settings

• sampling bias/effort……

• Comparing across scales

Sampling

effort

• Pseudo-

absence

points

Thank you