Transcript Modeling Biocomplexity - Actors, Landscapes and

Envisioning Future
Landscape Trajectories
Using Multiagent-based models to
Simulate Dynamics of Landscape
Change
John Bolte
Biological & Ecological Engineering Department
Oregon State University
Landscape Planning and Simulation
Models
How can we use simulation models in landscape
planning to help people achieve desirable futures?

The future is uncertain

The future is certain to come

We have little control over much of what will happen

We have substantial control over some aspects of what will
happen

The choices we make today will affect the choices we have
tomorrow
Simulation models that project the future outcomes
of different possible actions are particularly useful
when the system is complex, relationships are
poorly understood, or uncertainties are high
Why models? Types and uses
In the most general sense, a model is anything used
in any way to represent anything else
• Physical model – a literal representation of something, generally
in miniature, to show its construction or appearance.

Conceptual model – a simplified, abstract representation of a
system or phenomena, typically of its components and the
relationships among them. Commonly used to illustrate, explore or
explain current understanding and gaps in understanding of a
system and/or to generate hypotheses about how the system works.

Quantitative model – a model that uses numeric representations
of system components and interactions among them to produce
quantitative outcomes from quantitative inputs.
Quantitative Models
•
Empirical model – a quantitative model based on empirical
observations rather than on mathematically describable relationships of
the system modeled.

Statistical model – a quantitative formalization of relationships between
variables in the form of mathematical equations.

Mechanistic model – a model that uses cause and effect logic to
describe the behavior of a system.

Simulation model – a computer program which attempts to simulate the
behavior of a particular system. May be used to project the outcomes
of specified interactions and to test the consequences of different
actions or scenarios

Spatially explicit model – a model that represents the behavior of a
physical system, including the spatial relationships among its
components.

Multiagent-based model – a model that represents the behaviors of one
or more “actors” in a system.
Building Models
“All models are wrong,
but some are useful.”
Box, G.E.P. 1976. Science and
Statistics. Journal of the American
Statistical Association 71: 791-799.

The essence of simulation models is to incorporate
critical/dominant system features so that projections
of the consequences of different scenarios can be
made with some desired level of accuracy in
representing likely real-world outcomes.
Alternative Futures Modeling

Examine multiple scenarios of trends and assumptions
about future conditions, generally using one or more
models of change,

Assist in incorporating stakeholder interactions to
define goals, constraints, trajectories, drivers, outcomes

Allow visualization of the results in a variety of types
and formats

Ultimately are intended to assist in improving land
management decision-making
Trajectories of Change and Alternative Futures
Source: Hulse et al. (2008), modified from Shearer (2005)
Approach: Multi-agent Modeling

Based on modeling behavior and actions of
autonomous, adaptive agents (actors)

Our approach: spatially explicit, represents land
management decisions of entities (actors) with
authority over parcels of land

Actor decisions implemented through policies that
guide & constrain potential actions

Autonomous processes (e.g. succession)
simultaneously modeled
A General Theory of Action
(Parsons and Shils 1951)
Systems
Personality
Social
Cultural

values  attitudes  action
“… values are abstract concepts, but not so abstract that they cannot
motivate behavior. Hence, an important theme of values research has
been to assess how well one can predict specific behavior knowing
something about a person’s values” (Karp 2001:3213).
Complex Theory of Action
Drivers
Context = difficulty, time, expense
Systems
Actor
personality
social
cultural
economic
biophysical
values
 beliefs 
norms
goals
I
information/matter/energy
attitudes behavior
plan  action
desires
intentions
Envision – Conceptual Structure
Actors
Decision-makers managing the
landscape by selecting policies
responsive to their objectives
Landscape Production Models
Landscape
Feedbacks
Multiagent
Decision-making
Scenario
Definition
Select policies and
generate land
management decision
affecting landscape
pattern
Generating Landscape Metrics Reflecting
Ecosystem Service Productions
Landscape
Spatial Container in
which landscape
changes, ES
Metrics are
Landscape
depicted
Feedbacks
Policies
Fundamental Descriptors of constraints and
actions defining land use management
decisionmaking
Autonomous Change Processes
Models of Non-anthropogenic Landscape
Change
ENVISION
– Triad of Relationships
Goals
•Economic Services
•Ecosystem Services
•Socio-cultural Services
Provide a common frame of reference
for actors, policies and landscape productions
Landscapes
Metrics of Production
Policy Definition
Landscape policies are decisions or plans of
action for accomplishing desired outcomes.
from:

Lackey, R.T. 2006. Axioms of ecological
policy. Fisheries. 31(6): 286-290.
Policies in ENVISION

Policies are a decision or plan of action for accomplishing a
desired outcome; they are a fundamental unit of computation in
Evoland

Describe actions available to actors

Primary Characteristics:


Applicable Site Attributes (Spatial Query)

Effectiveness of the Policy (determined by evaluative models)

Outcomes (possible multiple) associated with the selection and
application of the Policy
Example: [Purchase conservations easement to allow
revegetation of degraded riparian areas] in [areas with no built
structures and high channel migration capacity] when [native fish
habitat becomes scarce]
Models in ENVISION

Models are “plug-ins” of two types:
1)
Autonomous Processes: Represent processes
causing landscape changes independent of human
decision-making – e.g. climate change, vegetative
succession, forest growth, fire, flooding, ???
2)
Evaluative Models – Generate production statistics
and report back how well the landscape is doing a
producing metrics of interest – e.g. carbon
sequestration, habitat production, land availability,
risk, ???
Actors in Envision

Actors are entities that make decisions about landscape change

Any number of actors can be defined ( 0-???)

Actors can be defined in terms of


A set of IDU attributes

Prescribed areas on the landscape

Randomly
Each IDU is controlled by at most one Actor
Actors in Envision
(continued)

Actors have values that influence their decision-making
behaviors. These values reflect landscape productions

Actors make choices about landscape
management by selecting policies
based on some combination of:

Internal Values relative to Policy Intentions

Landscape Feedbacks/Emerging Scarcities
(dynamically generated during a run)

Global Policy Preferences (defined by
scenario)
Actor Decision-making

Step 1: For each location and each time step, collect all relevant policies based on
site attributes

Step 2: Score the policies with respect to:


1) How well the policy intentions “align” with the actors on values (Self-interest)

2) How well the policies align with emerging
landscape scarcities (Altruistic)

3) a “global preference” for the policy that can
be defined conditionally

4) a “scenario-specific preference” for the
policy

5) where an “lives” on a Self-Interest/Altruism
scale
Step 3: Stochastically select a policy based on a
multicriteria score reflecting the above factors
Actor Associations in Envision

Actor associations are “collections” of actors, defined in one
of three ways, based on:

a common landscape attribute or set of attributes

common values

Spatial proximity

Associations influence an actors decision-making process by
modifying the actors values

In theory, Envisions’s actor decision-making can be influenced
by their group affiliations, but in fact we’ve never done
anything with this.
ENVISION Actor Properties
Property
Meaning
Envision
Reactive
Responds to environment
Yes
Autonomous
Controls own actions
Yes
Social
Interact with other actors
Sort of
Goal-oriented
More than responsive to
environment
Yes
Temporally continuous
Agent behavior continuous
Communicative
Communicates with other agents
Mobile
Can transport self to other
locations
No
Flexible
Actions not scripted
Yes
Learning
Changes based on experience
No
Character
Believable personality or emotions
No
Adapted from Benenson and Torrens (2004:156)
Once/step
Sort Of
Inferring Values from Actions:
Votes on 1998 Environmental Ballot
Measures
Statewide
Ballot Measure
Statewide
Percent
Yes
Lane
County
Percent
Yes
Yes
Votes
No
Votes
56 (notification)
874547
212737
80
73
64 (timber)
215491
897535
19
21
66 (parks & salmon )
742038
362247
67
70
Definition of value categories including descriptive terms and text examples.
Value Category
Descriptive Terms
Economic
reflecting economic production of the landscape, job activity, productivity,
opportunities for capital production and revenue generation
Property Rights
concern is with the freedom to own and use private property as a landowner desires
Ecosystem
Health
ecological health, diversity of the landscape, environmental protection and
restoration
Nonmarket
reflecting aesthetics, scenic integrity, beauty, spiritual, future generations, “right
thing to do,” undiscovered utility, learning about and gaining connection with the
environment
Fairness
refers to actor perceptions about economic justice, winners and losers, fears about
litigation and its costs; unfair policies force an actor to do something she does not
want to do
Credibility
refers to policies are justified by scientific or other expertise, or to policies that lack
scientific or support by other expertise
Safety
concerned with human safety in jobs and activites, from chemicals, from natural
hazards
Recreation
emphasis on any type of recreational activity that could be helped or hurt by passage
of the ballot measure.
Value Frequencies in Ballot Measures
MEASURE
No
Economic
%
Notification
Private
property
rights
Ecosystem
health
Nonmarket
%
%
Fairness
%
Credibility
Safety
Recreation
%
%
%
9
67
63
0
15
100
15
0
0
Timber Pro
20
73
0
93
55
28
30
77
35
Timber Con
29
74
35
52
13
38
51
15
3
Salmon &
Parks
21
84
3
84
52
16
6
3
71
cell
ACTORWT_0
-2.00 - -1.76
-1.75 - -1.58
-1.57 - -1.22
-1.21 - -0.82
-0.81 - -0.61
-0.60 - -0.38
-0.37 - -0.11
-0.10 - 0.18
0.19 - 0.45
0.46 - 0.73
0.74 - 0.94
0.95 - 1.13
1.14 - 1.31
1.32 - 1.51
1.52 - 1.77
1.78 - 2.11
2.12 - 2.41
Economics Values
2.42 - 2.66
2.67 - 2.89
2.90 - 3.00
Integrated Decision Units (IDUs)
A spatial geometry to model both human decisions and successional processes
Each IDU described in GIS by a set of attributes used to model
climate effects, succession, wildfire and decisions
Envision Andrews Application
Data Sources
Evaluative Models
Parcels (IDU’s)
Mean Age at Harvest
Agent Descriptors
Autonomous Process
Models
Rural Residential
Expansion
ENVISION
Policy Set(s)
Carbon Sequestration
Forest Products Extraction
Harvested Acreage
Fish Habitat (IBI)
Vegetative Succession
Climate Change
Resource Lands Protection
Envision Andrews - Scenarios

Conservation - no Climate Change

Development - no Climate Change

Conservation - with Climate Change

Development - with Climate Change
Envision Andrews Study Area
Scenario Results – Forest Carbon
Scenario Results – Forest Product
Extraction
Scenario Results – Fish IBI
Envision Puget Sound Application
Data Sources
Evaluative Models
IDU’s – GSU/LULC/…
Impervious Surfaces
Agent Descriptors
Autonomous Process
Models
Rural/Urban Development
ENVISION
Policy Set(s)
Water Quality/Loading
(SPARROW)
Nearshore Habitat
(Controlling Factors Model)
INVEST Tier 1 Carbon
Expansion of
Nearshore Modifications
Resource Lands Protection
Population Growth
Residential Land Supply
Envision Puget Sound- Scenarios

Status Quo – continue current trends

Managed Growth – adopt a suite of additional
policies aimed at conserving/restoring habitats,
protecting resource lands, emphasizing denser
development pattern near urban areas

Unconstrained Growth – allow lower density
patterns, less habitat protection, less resource
land protection
Puget Sound
South Sound
Bainbridge Island
Ferry Terminal Area
more info at:
http://envision.bioe.orst.edu