Artificial Life in Animal Ecology
|has title::Can IBM’s be suitable tools for Artificial Life in Animal Ecology?|
|Master:||project within::Computational Intelligence and Selforganisation|
|Student name:||student name::David Wendt|
|Second reader:||has second reader::Jacintha Ellers and Matty Berg (Animal Ecology)|
Can IBM’s be suitable tools for Artificial Life in Animal Ecology?
Individual-Based Simulation of Population Dynamics and Plasticity as a response to Fluctuating A-biotic Factors.
A classical mathematical method for analysis within the field of population ecology and ecological modelling has been and still is an important approach in the study of patterns and evolution of life history. However, the assumptions of classical ecological models speak for themselves, where only average individuals are considered and only feature characteristics of the population as a whole (Uchmanski and Grimm, 1996). Nowadays, individual-based modelling has been considered as a promising qualitative and quantitative methodology for social sciences and increasingly used for spatial simulation of a wide variety of ecological, environmental, urban and economical phenomena’s (Parker et. al., 2003) (Bousquest and Page, 2004) (Batty, 2005). This shift of the emphasis to individuals in ecological modelling was possible because of the development of computers. However, classical models are simpler and allow perfectly for analytical solutions (Uchmanski and Grimm, 1996). This raises the question whether it pays more to use simple global or more costly individual approach?
From a Computer Scientist perspective we are interested in individual-based modelling. Therefore, it would be interesting to question whether this modelling approach could give us insight into the predictive value of supporting biologists, philosophers, or sociologists in validating their theories, even when these theories concern phenomena showing nontrivial dynamics. Here, we would like to investigate whether individual-based modelling can be a suitable tool for modelling, exploring and answering biological questions. Therefore, this study attempts to give answer to a more specific question: “Can we develop an individual-based model that includes domain-knowledge and empiric field-data in order to provide Biologists better insight and understanding of the dynamics and phenomena that occur in a specific case-study in Animal Ecology?”
To address this, in collaboration with research group Animal Ecology (VU), a specific case study at the green beach at Schiermonnikoog has been chosen in order to answer this research question. The behaviour of the two most dominant species and the ecological composition has been modelled for simulating, verification and analyzing the emerging population effects at this green beach. We are interested if this study could show us how easy ecological problems can be represented and explored with individual-based modelling and if the results of the simulation are consistent with what the domain-experts claim.
In order to be certain that we face a valid model, interviews are taken with domain-experts, extensive literature study and fieldwork has been done to verify the model at a local rule level. Then, to analyse the simulation model at individual– and population level, the predicate logical Temporal Trace Language TTL (Bosse et al., 2006) was used to formalize the hypotheses by specifying a number of global dynamic properties that are relevant for the Biologists to check against given simulations runs.
From an Artificial Intelligence perspective:
1. Can Individual-Based Modelling (IBM) be a suitable tool for modelling and exploring ecological problems that occur at the Green Beach at Schiermonnikoog?
From an Animal Ecology perspective:
2. How is the population distribution in terms of abundance and fitness related to gradient and plasticity as a response to abiotic factors?
If prey is more sensitive to salt-ratio than predators than we expect that
- a. predators to occur over a larger salt range than prey species.
- b. fitness divergence over the whole salt range to be greater for prey than for predators, because prey are directly influenced by salt levels, and predators only indirectly through abundance of prey.
- c. Large population fluctuations in prey and predator abundance, because predators are better able to cope with salt and will reduce prey levels over the entire range.
If prey is less sensitive to salt-ratio than predators than we expect that
- d. prey to occur over a larger salt range than predator species.
- e. fitness divergence over the whole salt range to be greater for predators than for predators, because predators are directly influenced by salt levels, and prey only indirectly through mortality risk by predators.
- f. Reduced population fluctuations in prey and predator abundance, because prey have a refuge from predation in the salt range where predators cannot live.
It turned out that:
- habitat quality plays an integral role in success of population dynamics in terms of fitness and abundance,
- species occur over a larger salt range once they are more plastic than their opponent – except in cases when prey are fresh water specialists,
- salt-water predator specialists do not benefit enough to become generalist, only if they become extreme generalist at once,
- fresh-water prey specialists do not perform better when they become generalist,
- salt-water prey specialists will benefit quickly when they become generalist,
- predator fresh water specialists perform best if they are able to limit their sensitivity for salt-ratio from 1 to 2.
- finding food for predator is less certain than for prey, therefore predators are more influenced by salt penalties than prey and finally
- population dynamics in this simulation are never approaching an equilibrium state or are limiting to certain bandwidth.
- The 10% chance for predators to catch prey could be discussed further at the green beach of Schiermonnikoog.
In general, the results suggest that dispersion-pressure is stronger for salt-water specialists than for fresh-water specialists. Here, food location and spatial vegetation structure are important factors for developing this plasticity. Since the food quality increases as the patches become fresher, the need to be more plastic is stronger at the salty patches than at the fresh patches. Here, it is better for salt-water specialists to become generalist under these circumstances in order to increase the chance of survival. Fortunate, there is also something to say about gender and plasticity. We have seen that predators are more influenced by salt penalties than prey and therefore predators preserve a greater need to be plastic than prey. Furthermore, we have seen that the population dynamics occurring in this simulation never approaches to an equilibrium state. This seems rather strange at first glance, because most historical and contemporary opinions from biologists and environmentalists have adopted this view (Hastings, 1993) and (Pimm, 1991). However, nothing is as changeable as Nature and stable oscillations can occur in predator-prey systems, with the peak of species lagging slightly behind the opponent oscillation (Smith, 1973).
In summary, was this individual-based modelling a suitable tool for modelling, simulating and analysing this specific case study? Besides both supporting and contradict the hypotheses of this study, did individual-based modelling provide a suitable work-around for exploring these problems? Beyond the scope of this study, the answer whether it pays to use a simple classical or a complex individual-based approach lies more between using both. As pointed out before, individual bottom-up approaches will never lead to theories and classical top-down approaches never fundaments the whole theory. Both individual-based and classical approaches have their own momentum and therefore using both approaches can maintain a prolific direction towards solving the problem (Grimm and Railsback, 2005). Unfortunately, using solely individual-based modelling approaches for exploring population dynamics has some downsides that on the long run have let to more complexness and dilatation than simplicity and completeness. These characteristics are in line with our own experiences indicating that the excessive validation of the model (at local, individual and global level) – and the process of constantly abstracting the domain – and constantly adjusting parameters to represent reality, brought huge work overload. In respect to accuracy (Bosse et al., 2008) and gaining valuable research time, we would next time seriously consider classical approaches for exploring population effects. Starting with a classical approach for gaining fast contextual insight of the problem and use an individual-based approach for detailed excessive problems that makes understanding difficult.
“Every problem has its solution, and every solution has its problems” TomSaidak, blogger Peakoil.com
- Bakker, J.P., Veeneklaar, R.M., Jansen, A. and Samwel A. (2005). Een nieuw Groen Strand op Schiermonnikoog, De levende Natuur 2005, pp. 151-155
- Bousquet, F., and Page, C.L. (2004). Multi-agent simulations and ecosystem management: A review. Ecological Modelling 176:313-332.
- Bosse, T., Jonker, C.M., Meij, L. van der, Sharpanskykh, A. and Treur, J. (2006). Specification and Verification of Dynamics in Coginitive Agent Models. In: Proceedings of the Sixth International Conference on Intelligence Agent Technology, IAT06. IEEE Computer Society Press, 2006, pp. 247-254.
- Cole, L.C., (1954). The population consequences of life history phenomena. Q. Rev. Biol., 29:103–137, 1954.
- Jong, de,. (2005). Evolution of phenotypic plasticity: patterns of plasticity and the emergences of ecotypes. New Phytol. 166: 101-117
- DeWitt,T.J. et.al., (1998) Costs and limits of phenotypic plasticity. Trends Ecol. Evol. 13, 77–81
- Krebs, J.R., and Davies, N.B. (1993). An Introduction to Behavioural Ecology (3nd ed.). Oxford: Blackwell.
- Liefting, M., and J. Ellers. 2008. Habitat-specific differences in thermal plasticity in natural populations of a soil arthropod. Biol. J. Linn. Soc. 94:265–271.
- Nylin, S., and Gotthard, K. (1998) Plasticity in Life-history traits. Annu. Rev. Entomol. 43, 63-83
- Parker, D. C., S. M. Manson, M. A. Janssen, M. J. Hoffman, and P. Deadman. (2003). Multi-agent systems for the simulation of land-use and land-cover change: a review. Annals of the Association of American Geographers 93(2): 316–340.
- Uchmanski, J., and V. Grimm. (1996). Individual-based modelling in ecology: what makes the difference? Trends in Ecology and Evolution 11:437-441.