Banas NS (2011) Effects of adding complex trophic interactions to a size-spectral plankton model: Emergent variability and emergent structure. Ecol. Modelling 222:2663–2675.
ASTroCAT is implemented in Java in the Processing environment, as part of this NPZ visualizer. Start there to play with this and other nutrient-phytoplankton-zooplankton models and to download a version you can modify yourself.
This poster describes a newer version of the model, that runs in a slightly more complex testbed and includes some new experiments with adaptive phytoplankton dynamics.
If you want a streamlined version that runs ASTroCAT only, download ASTroCAT16.zip. This contains a complete sketch that runs inside Processing. Or you can browse the source code right here: ASTroCAT16 | Projects | ac_diversity | ac_expts | ac_model | ac_npz | core_archive | core_eco | core_env | core_varFluxEtc | gr_axes | gr_network | gr_view | gui | gui_generic | utils_misc
ASTroCAT is a new model in the NPZ (nutrient–phytoplankton–zooplankton) style, mechanistically simple but with 40 size classes each of phytoplankton (1–20 μm) and small zooplankton (2.1–460 μm), in order to resolve one level of trophic interactions in detail. General, empirical allometric relationships are used to parameterize both the optimal prey size and size selectivity for each grazer class, as is rarely done. This inclusion of complex predator–prey linkages and realistic prey preferences yields a system with an emergent pattern of phytoplankton diversity consistent with global ocean observations, i.e., a parabolic relationship between diversity (as measured by the Shannon evenness) and biomass.
It also yields significant long-term time evolution, which places limits on the extent to which the community response to nutrient forcing can be predicted from forcing in a pragmatic sense. When a simple annual cycle in nutrient supply is repeated exactly for many years, transient fluctuations up to a factor of two in spring bloom magnitude persist for 10–20 years before a stable seasonal biomass cycle is achieved. When the amplitude of the nutrient-supply annual cycle is given a random interannual modulation (left), these long-lived transients add significant noise to the relationship between annual-mean nutrient supply and annual-mean biomass.
This noise is 20% of total interannual variance in the model base case (left), and ranges from 0% to 40% depending on the grazer size selectivity. In general, unpredictability on the bloom timescale is damped when food-web complexity is increased by making grazers less selective, while unpredictability on the interannual scale shows the opposite pattern, increasing with increasing food-web complexity up to a high threshhold, past which community structure and biomass time evolution both suddenly simplify.
These results suggests a new strategy for ensemble ecosystem forecasting and uncertainty estimation, analogous to methods common in circulation and climate modeling, in which internal variability (predator–prey interactions in the biological case; eddies and climate-system oscillations in the physical case) are resolved and quantified, rather than suppressed.