John Keane

Modelling weed spread
in heterogeneous landscapes
NZIMA weeds workshop
17 April 2007
John Kean (AgResearch, Lincoln)
Jake Overton (Landcare Research, Hamilton)
Peter Williams (Landcare Research, Nelson)
Rowan Buxton (Landcare Research, Lincoln)
Making a difference for a truly clean,
green and sustainable New Zealand
Overview
•
Why model weeds at the landscape scale?
•
Modelling weed spread in heterogenous landscapes
(e.g. PestSpread v.1)
•
Field data for
modelling
(e.g. hawthorn)
•
A model of a
weed model
What is a weed?
1. Any plant that is growing where it is unwanted
“A weed is a plant that has mastered every survival skill except for learning
how to grow in rows.” - Doug Larson
“What is a weed? A plant whose virtues have never been discovered.”
- Ralph Waldo Emerson
What is a weed?
1. Any plant that is growing where it is unwanted
“A weed is a plant that has mastered every survival skill except for learning
how to grow in rows.” - Doug Larson
“What is a weed? A plant whose virtues have never been discovered.”
- Ralph Waldo Emerson
2. A town in northern
California
Why model weeds?
(Feedback from DOC, Regional Councils, Biosecurity NZ)
• Prioritise pests and control efforts
• Transparency of decision-making
• Target surveillance
• Optimising control efficacy
• Support national and international cooordination
• Estimate and communicate the difference made
• Identify research needs
Weed prioritisations
• National Pest Plant Accord
(http://www.biosecurity.govt.nz/pests-diseases/plants/accord.htm)
• Regional Pest Management Strategies
(e.g. http://www.ecan.govt.nz/Plans+and+Reports
/pestAndWeeds/RPMS+2005.htm)
• National Pest Management Strategies
(e.g. http://www1.maf.govt.nz/pms/cgi/pms.pl)
Prioritisations are largely subjective:
expert opinion + qualitative weed risk assessments
Can we do better?
yr 10
yr 30
yr 20
current
distribution
yr 50
yr 40
potential
distribution
local population growth
(aging + local reproduction)
current
distribution
potential
distribution
dispersal of propagules
model
user
PestSpread v.1
predicted
distribution maps
stored
resources
web server
(with GIS)
modelling
modules
demography
modules
e.g. annual herb,
tree, vine etc
dispersal
modules
e.g. wind, bird,
water etc
species
setup files
e.g. gorse, pinus,
old man’s beard
model
core
species
distributions
current, potential,
pre-calculated
• species setup file
• distribution maps
• management file
other spatial
information
e.g. friction maps
for dispersal
Demography modules 1
sigmoid local
increase
± age-dependent
seed production
± persistent seed bank
Demography modules 2
Seeds
Seedlings
Juveniles
Adults
Dispersal modules 1
nearest neighbour
classical
dispersal
kernel
Dispersal modules 2
wind ± topography
Dispersal modules 3
water runoff: direction
+ flow rate
Dispersal modules 4
bird dispersal
= habitat preference
+ seed deposition
Case study weeds
Widespread species
Limited distribution species
Tree
Corsican pine
Pinus nigra
in Twizel
Sweet pittosporum
Pittosporum
undulatum
in Kaitaia
Shrub
Scotch broom
Cytisus scoparius
in Palmerston North
Spiny broom
Calicotome spinosa
in Palmerston North
Grass
Pampas
Cortaderia selloana
in Palmerston North
Pypgrass
Ehrharta villosa
in Palmerston North
Vine
Old man’s beard
Clematis vitalba
in Palmerston North
White bryony
Bryonia cretica
in Palmerston North
Case study weeds
Widespread species
Limited distribution species
Tree
Corsican pine
stage structured
wind
Sweet pittosporum
stage structured
bird dispersal
Shrub
Scotch broom
stage structured
neighbour + run-off
Spiny broom
stage structured
neighbour + run-off
Grass
Pampas
stage structured
classical kernel
Pypgrass
sigmoid
neighbour
Vine
Old man’s beard
stage structured
classical kernel
White bryony
sigmoid + age
bird dispersal
Pypgrass assumptions
SIGMOID local increase
NEAREST8 dispersal
(10% of cover)
Pypgrass predictions
Corsican pine life cycle
(NB. No seed bank)
Seedlings
< 2 yr
Juveniles
2 - 14 yr
Adults
>14 yr
Wind dispersal
Wind rose for Twizel in May
when wind speed > 5 m/s
and temperature > 15 °C
Corsican pine
Corsican pine
predicted % cover
for 2054
Old man’s beard life cycle
Seedling
< 1 yr
Juvenile
1 – 2 yr
250 m
Seed dispersal
Mature adult vine
> 3 yr
Old man’s beard
Old man’s beard predicted % cover
What next?
Robust pest prioritisation and risk assessment
= potential distribution (ultimate risk)
+ current distribution (scope for additional damage)
+ change over time (immediacy of risk)
PestSpread
v.1
+ management = cost/benefit of action
+ value of affected areas ($$ or NHMS)
+ impact on affected areas
PestSpread
v.2
Points to ponder
1. What is the appropriate spatial scale to be working at?
2. Can we just “scale up” from local models?
3. How much detail about the landscape do we need?
4. Can we really see the landscape from a plant's point of
view?
5. How does landscape affect competition/invasibility?
6. Can we legitimately extrapolate model results from one
landscape to another, or from one species to another?
Spread of hawthorn
Aims:
1. To identify the changing drivers determining
hawthorn spread
2. To predict hawthorn spread under different
landscape and management influences
Study site: Porters pass, Canterbury
Hawthorn ecology:
 Long-lived, slow to mature
 Abundant fleshy fruit spread by blackbirds
 Seedlings only partially grazing resistant
1908 (WA Taylor glass plate, Canterbury Museum)
1978 (Ian Whitehouse photo)
2005
Sampling hawthorn
The grand-daddy of them all
A successful day in the field
Predicting hawthorn age
90
90
y = 7.12e
y = 1.26x + 7.70
0.28x
2
R = 0.81
2
60
Age (years)
Age (years)
R = 0.70
30
0
60
30
0
0
3
6
Height (m)
9
12
0
10
20
30
40
Diameter (cm)
50
1930
1935
1940
1945
1950
1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
W
SW
SE
S
Direction from original tree
0
10
0
20
0
30
0
40
0
50
0
60
0
70
0
80
0
90
10 0
0
11 0
0
12 0
0
13 0
0
14 0
0
15 0
0
16 0
0
17 0
0
18 0
00
NW
trees
2006
Proportion
Prop. of of
2006
trees
Hawthorn spread
N
NE
E
0.16
0.12
0.08
0.04
0
Distance from original tree (m)
Effects of landscape
Landform
Intrinsic rate
of increase /yr
Hills
0.0833
Gullies
0.0707
Scarps
0.0662
High terraces
0.1341
Low terraces and riverbed
0.1925
Hawthorn invasion
(NB. Log scale)
Relative no trees present
1000
100
10
1
1925
1935
1945
1965
1955
Year
1975
1985
1995
Hawthorn invasion
(NB. Log scale)
Relative no trees present
1000
100
Phase 1
r = 0.036 /yr
10
1
1925
1935
1945
1965
1955
Year
1975
1985
1995
Hawthorn invasion
(NB. Log scale)
Relative no trees present
1000
Phase 2
r = 0.126 /yr
100
Phase 1
r = 0.036 /yr
cessation of burning
+ rabbit control
+ fertilisers
= blackbird nesting sites
10
1
1925
1935
1945
1965
1955
Year
1975
1985
1995
Potential risk of weed
=(
probability of
naturalisation
×[
potential
distribution
(+
×[
local rate
of increase
×
× (
economic
social
environmental
-
feasibility and cost
of eradication
)
climate
change
)-
current
distribution
dispersal
rate
-(
feasibility and
cost of control
impact on invaded
ecosystems
×
]
value of invaded
ecosystems
)
propagule
persistence
)]
Potential risk of weed
=(
probability of
naturalisation
×[
potential
distribution
(+
×[
local rate
of increase
×
× (
economic
social
environmental
-
feasibility and cost
of eradication
)
climate
change
)-
current
distribution
dispersal
rate
-(
feasibility and
cost of control
impact on invaded
ecosystems
×
]
value of invaded
ecosystems
)
propagule
persistence
)]
needs work
potential gains
well studied
Acknowledgements
Department of Conservation
Graeme Bourdot (AgResearch, Lincoln)
Shona Lamoureaux (AgResearch, Lincoln)
James Barringer (Landcare Research, Lincoln)
Stephen Ferriss (Landcare Research, Lincoln)
Mandy Barron (AgResearch, Lincoln)
Rowan invasion at Tekapo