Qualitative Reasoning About Population and Community Ecology

1
Qualitative Reasoning About Population and
Community Ecology

• Reha K. Gerçeker
• Bogaziçi University, 2005

2

• Qualitative Reasoning About Population and
  Community Ecology
• Paulo Salles, Bert Bredeweg. AI Magazine.
• Winter 2003. Vol. 24, Iss. 4 p. 77
• http//staff.science.uva.nl/bredeweg/pdf/aimag200
  3c.pdf

3
Simulation of Ecological Systems

• Interested in population dynamics
• Interested in interaction between different types
  of population (i.e. predation...)
• Tries to explain the mechanisms behind an
  observed behaviour

• Interested in population dynamics
• Interested in interaction between different types
  of population (i.e. predation...)
• Tries to explain the mechanisms behind an
  observed behaviour
• Ecological modelling is equivalent to
  mathematical modelling
• is it possible to capture accurate mathematical
  models?

• Interested in population dynamics
• Interested in interaction between different types
  of population (i.e. predation...)
• Tries to explain the mechanisms behind an
  observed behaviour
• Acquiring data of good quality requires
  long-term observations
• Data is mostly imprecise and incomplete

4
Why go Qualitative?

• Ecological data is more qualitative than it is
  quantitative
• Exact quantities are never available
• Exact quantities are not important either
• An ecologist is actually interested in
  qualitative simulation rather than quantitative
  simulation
• Qualitative models easily capture the knowledge
  in an ecologists mind
• Explicit and well-organized knowledge
• Computer processible

5
A Reasoning Engine GARP

• Bredeweg in 1992 has implemented a qualitative
  reasoning engine called GARP
• General Architecture for Reasoning about Physics
• It is based on Qualitative Process Theory by
  Forbus
• It has a compositional modelling approach like
  the QPT

6
The Growth Equation

• Nof(t 1) Nof(t) (B Im) (D E)

7
An Ecological Process Natality
8
Basic Processes

• Natality
• I(Nof, B), P(B, Nof)
• Mortality
• I(Nof, D), P(D, Nof)
• Immigration
• I(Nof, Im)
• Emigration
• I(Nof, E), P(E, Nof)

9
Quantity Space Resolution

• In physics specific landmarks exist
• i.e. a specific landmark for a temperature
  variable might be the boiling point
• In an ecological system, there are no specific
  landmarks to place inside the quantity spaces of
  Nof

10
GARPs Transition Rules

• QSIM and GARP differ in their transition rules in
  an interesting way
• GARP is concerned with neither time intervals nor
  intervals of landmarks
• Transitions seem to take place between time
  points only

11
Ambiguities

• According to the growth equation, Nof is
  influenced by several factors
• The effects of such numerous factors are combined
  by what Forbus calls influence resolution
• That is where ambiguities arise because the
  overall influence depends on the relative amounts
  of the factors (which are unknown)
• Ambiguities can cause the simulation to branch
  enormously

12
Ambiguities (contd)

• Ambiguity as a guide
• Ambiguity might act as a guide for an ecologist
  to acquire more information
• It might direct ecologists to fields of research
  where more work has to be done
• Ambiguity as a feature
• Ambiguity might sometimes be favorable
• That is how different branches of simulation come
  up after all
• Simplifying assumptions
• closed population (Im lt0, stdgt, E lt0, stdgt)

13
Interaction Between Populations
14
Interaction Types

• Interactions
• neutralism (0, 0)
• amensalism (0, )
• comensalism (0, )
• predation (, )
• symbiosis (, )
• competition (, )
• Another type of interaction is the absence of a
  population
• when there is no prey population, the predator
  population cannot survive

Modeled once and placed into the library of model
fragments
15
Simulating Predation
Causal Model for Predation Population 1
Predator Population 2 Prey
16
Simulating Predation (contd)
Start simulation with Nof1 ltnormal, ?gt and Nof2
ltnormal, ?gt
17
Cerrado Succession Hypotheses

• Brazilian cerrado vegetation
• There are different types of cerrado communities,
  characterized by the proportions of grass, shrubs
  and trees
• grass likes bright, warm, dry microenvironments
• trees like shaded, cold, moist microenvironments
• These communities have well-defined composition
  determined by
• fire frequency
• soil fertility
• water availability

The increases and decreases in populations of
cerrado communities is referred to as the Cerrado
Succession Hypotheses
18
Cerrado Causal Model
19
Simulating CSH
20
Conclusion

• Qualitative representation provides a rich
  vocabulary for describing
• objects
• situations
• causality
• mechanisms of change
• Conclusions relevant to ecologists can be derived
  automatically using only qualitative data
• Qualitative models prove to be a valuable
  complement to mathematical approaches in
  ecological modeling

21
Conclusion (contd)

• Compositional approach enables reusability
• lets the modeler use parts of his previously
  defined models
• lets the modeler to increase the complexity of
  his models gradually
• basic models represent fundamental knowledge that
  explain more complex systems

22
Future Work

• Apply same approach to represent and understand
  behaviour of other large communities
• Develop tools to support educational and
  management activities