Gecko's promise lies in the ability to study food webs. One opportunity is exploring their tolerance to perturbation, and how to dampen perturbations. Such perturbations might include removing the top level predators, cycling resource levels, and reducing site area or chopping it into parcels. As suggested in the introduction, this is an area where individual-based models offer great promise over population-based models. Before perturbations can be studied, we need dynamically stable systems as controls for comparison.
Figure 2 and Figure 4 show such a stable system of three trophic levels in three species. This system persists for 4000 rounds in 93% of runs, based on 100 trials.
There are several sources of randomness in a Gecko run. First, although all members of one species are introduced with identical attributes and life-cycle state (newborn), they are placed randomly about the site. Their neighborhoods, and thus opportunities, differ. This isn't unique to species introduction---all newborn locations have a random component. Second, when a Grazer faces a choice of plants to eat, he selects randomly among the choices. Third, at each step, there is a random component to the Grazer's direction. All other factors are deterministic. However, these sources produce considerable variation between runs. A dynamically stable scenario such as this is tolerant of the variation.
Table 1 gives the parameters used for this scenario. Some of these parameter choices are discussed above with the attributes of Creatures, Grazers, and Carnivores. The selection of assimilation efficiency is discussed below. In brief, after some trial and error, the minimum sizes for plants and herbivores were hand-set, as was the site size. The smallest feasible site and population size was desired. This was partly for pragmatic reasons---smaller scenarios run faster. But also, the desire was to deemphasize the role of luck in system persistence, focusing instead on the species' abilities. The larger the field, the larger role random chance plays in stabilization. Assimilation efficiencies were considered given, and the site production function is effectively a unit declaration of 1.0.
A number of parameters remain, all underdetermined. This is a common bane of simulators which aspire to realism. These parameters were set close to stable by trial and error, then tuned algorithmically with an Evolution Strategy based search engine [12,17].
How much ecosystem-level realism is achieved by this scenario? The previous section discusses the trophic cascade test. Here we compare Gecko's realized energy budgets with those measured in nature.
One striking feature of the biomass graph in Figure 3b is that the biomass of plants is much higher than that of grasshoppers and spiders. (The runs of Figures 3 and 4 are equivalent after the third species is introduced.) If we use the assimilation efficiencies to compare standing biomass of the three species, we can equate one spider unit biomass (cproteins) to 1.2 grasshopper units (proteins) at 85%, and 3.5 plant units (carbos) at sequential 85% and 33% efficiencies. This is the scale used in Figure 3b. In these cprotein equivalents, the average standing biomasses of the three species form a proportion of 86.4::2.6::1.0. In other words, the living grasshoppers embody 3% of the energy embodied by plants, and spiders 40% of that in grasshoppers, or only 1% of that in plants.
The assimilation efficiencies of 33% for grasshoppers and 85% for spiders are based on Figures 1 and 2 in Hairston and Hairston [HH] [2]. These figures summarize the trophic energy transfers through temperate forest and grassland, based on a number of detailed studies. For herbivores, I used the forest assimilation efficiency, as the grassland figure clearly indicates that 60-81% of herbivory was below ground, and I felt Gecko Grassland was unequal to that comparison. The 85% efficiency for carnivores is taken from grasslands.
Hairston and Hairston place the energy content of the living forest food chain at a proportion of 83::2::1, and the temperate grassland food chain at 67::1.5::1 (my restatement of their data). Both ecosystems are dominated by the detritus cycle rather than the living food chain. Carnivores in both systems receive only 5-6% of their energy by eating herbivores, the remainder from detritivores. Considering how simplistic Grassland is, and assuming average proportional biomass is equivalent to average proportional energy flow, the agreement between Grassland's trophic pyramid of 86.4::2.6::1.0, and the proportions of 83::2::1 and 67::1.5::1 reported for real terrestrial forest and grassland systems, is remarkable.