HR: 17:35h
AN: NG14A-07 [Abstracts]
TI: Maximisation Principles in Foodwebs and Daisyworlds
AU: * Ackland, G J
EM: gjackland@ed.ac.uk
AF: School of Physics,
University of Edinburgh, 5411 JCMB
Kings Buildings, Edinburgh, EH9 3JZ
United Kingdom
AU: Gallagher, I D
EM: s9812798@sms.ed.ac.uk
AF: School of Physics,
University of Edinburgh, 5411 JCMB
Kings Buildings, Edinburgh, EH9 3JZ
United Kingdom
AB:
Using computer simulation we investigate whether the steady-state time
averaged state of a self-organising system with many internal degrees
of freedom can be described by optimising a single quantity. Our open
systems follow evolutionary dynamics hence the conservation laws and
energy-based state probabilities which underpin Hamiltonian dynamics
do not apply. We find that these dynamics observe a novel optimality
principle, that the system self-organises to a state which maximises
the sustainable amount of replicating objects.
We have studied a number of mathematical models of evolving
replicating systems: daisyworlds[1], logistic map and generalized
Lotka Volterra foodwebs[2]. Each is characterised by being (1) "open"
- resources flow into and out of the system. (2) "self-regulating" -
the inflow/outflow of resources is not fixed externally. (3)
"evolving" - the increase in population at the next timestep depends
on the population at the current timestep. These properties violate
the assumptions made in deriving optimality principles such as free
energy minimisation, maximum/mimimum entropy production etc., so it is
unsurprising that they are not observed.
The absence of a Hamiltonian for ecosystems is particularly
problematic for coupled models of life and the environment - moreover
there is ambiguity in defining an entropy for an ecosystem. By
considering large and small species within the 2D daisyworld model we
show that the appropriate measure comes from the interaction with the
rest of the system, not the information theoretic entropy of the daisy
field.
We introduce evolution within the classic Lotka-Volterra model for
interaction between species in an ecosystem. Generalisation to many
species is straightforward, but the resulting network is usually
unstable. By restricting the number of links between species it is
possible to form a stable network by evolution - allowing some species
to go extinct. This method can be used to generate arbitrarily large
network, from which a treelike structure of trophic levels emerges,
but typically the number of connection is much smaller than in real
ecosystems. Here, we show that applying evolution to the strength of
the links, rather than simply their existence, stabilises the entire
network and generates a power-law distribution of link strengths. The
network dynamics are chaotic, but as a whole tend towards maximising
the use of resources. If the dynamics are linearised to remove the
chaos, the scale-free link strengths also disappear.
[1] Maximisation Principles and Daisyworld G.J. Ackland J.Theo.Bio. 227, 121, (2004)
[2] Stabilization of large generalized Lotka-Volterra foodwebs by evolutionary feedback
G.J. Ackland and I.D. Gallagher
Phys Rev Lett 93 158701 2004
UR: http://www.ph.ed.ac.uk/nania/examples.html
DE: 3265 Stochastic processes (3235, 4468, 4475, 7857)
DE: 4430 Complex systems
DE: 4480 Self-organized criticality
DE: 4485 Self-organization
SC: Nonlinear Geophysics [NG]
MN: Fall Meeting 2005