Identification of dynamically resilient exploitationstrategies in complex landscape systems using sp

1
Identification of dynamically resilient
exploitation-strategies in complex landscape
systems using spatially explicit dynamical models
and multi-objective optimisation
Cameron Fletcher Andrew Higgins David Hilbert
Peter Roebelling
2
General understanding of land exploitation as a
dynamical system An improved understanding of
the trade-offs in land use triple bottom line
3
Project objectives
1. Develop an heuristic, phenomenological model
of local land-exploitation
2. Using the local model, develop a spatially
explicit model of landscape-level land-use
3. Develop techniques for multi-objective
optimisation to discover landscape-scale
exploitation strategies that are optimal in terms
of system-wide dynamical properties such as
resilience
4
4. Identify land-exploitation strategies that
lead to the best long-term environmental and
socio-economic outcomes as a function of regional
environmental and economic characteristics
5. Assess the utility of the multi-objective
optimisation techniques in collaboration with
other projects
5
A local land-exploitation model based on
ecological exploitation models
6
K
r.1 K100
N
7
Dynamics
Single point attractor Multiple attractors Stable
limit cycle
8
Land exploitation
Define N as aggregate, renewable natural capital
Define H as aggregate human-made capital and
labour
9
(No Transcript)
10
(No Transcript)
11
Strategies in the exploitation state space
non-extractive uses
Intensification
Ecosystem GS
12
Local optimisation
Intensification
Ecosystem GS
13
Subsidies Externalities
External Environment
Local Model
Represent spatial interactions in terms of their
effects on natural capital, economic value, and
biodiversity
14
(No Transcript)
15
Multi-objective optimisation
The Multi-objective Optimisation Problem can be
defined as finding a vector of decision
variables which satisfies constraints and
optimises a vector function whose elements
represent the objective functions.
Having several objective functions, the notion of
optimum changes, because we are trying to find
a set of compromises (a so called Pareto optimum)
rather than a single solution as in global
optimisation.
16
Optimisation using Genetic Algorithms
Generate a population of models with random
combinations of cmax, h and s
Run each model for some number of time steps
Calculate the models fitness from some
(muti-objective) criteria
Select some small number of models with the
highest fitness, eliminate the rest
Generate a new population of models by
recombination and mutation
17
Example
Fitnessmean production/standard error of N
45000
24
40000
22
35000
20
30000
N variance
Pmean
18
25000
16
20000
15000
14
10000
12
5000
10
0
0
10
20
30
40
50
0
10
20
30
40
50
Epoch
Epoch
0.12
0.1
0.08
exploitation per unit H
0.06
0.04
0.02
N
0
0
50
100
150
200
250
300
18
g
19
140
120
100
variance
N
80
60
mean
40
no forcing
20
0
2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
u
20
Parameters
21
(No Transcript)