Steven F. Railsback

Agent-based Simulation Platforms: Review and Development Recommendations

Five software platforms for scientific agent-based models (ABMs) were reviewed by implementing example models in each. NetLogo is the highest-level platform, providing a simple yet powerful programming language, built-in graphical interfaces, and comprehensive documentation. It is designed primarily for ABMs of mobile individuals with local interactions in a grid space, but not necessarily clumsy for others. NetLogo is highly recommended, even for prototyping complex models. MASON, Repast, and Swarm are “framework and library” platforms, providing a conceptual framework for organizing and designing ABMs and corresponding software libraries. MASON is least mature and designed with execution speed a high priority. The Objective-C version of Swarm is the most mature library platform and is stable and well organized. Objective-C seems more natural than Java for ABMs but weak error-handling and the lack of developer tools are drawbacks. Java Swarm allows Swarm’s Objective-C libraries to be called from Java; it does not seem to combine the advantages of the two languages well. Repast provides Swarm-like functions in a Java library and is a good choice for many, but parts of its organization and design could be improved. A rough comparison of execution speed found MASON and Repast usually fastest (MASON 1-35% faster than Repast), Swarm (including Objective-C) fastest for simple models but slowest for complex ones, and NetLogo intermediate. Recommendations include completing the documentation (for all platforms except NetLogo), strengthening conceptual frameworks, providing better tools for statistical output and automating simulation experiments, simplifying common tasks, and researching technologies for understanding how simulation results arise.

Pattern-oriented modelling: a 'multi-scope' for predictive systems ecology.

Modern ecology recognizes that modelling systems across scales and at multiple levels—especially to link population and ecosystem dynamics to individual adaptive behaviour—is essential for making the science predictive. ‘Pattern-oriented modelling’ (POM) is a strategy for doing just this. POM is the multi-criteria design, selection and calibration of models of complex systems. POM starts with identifying a set of patterns observed at multiple scales and levels that characterize a system with respect to the particular problem being modelled; a model from which the patterns emerge should contain the right mechanisms to address the problem. These patterns are then used to (i) determine what scales, entities, variables and processes the model needs, (ii) test and select submodels to represent key low-level processes such as adaptive behaviour, and (iii) find useful parameter values during calibration. Patterns are already often used in these ways, but a mini-review of applications of POM confirms that making the selection and use of patterns more explicit and rigorous can facilitate the development of models with the right level of complexity to understand ecological systems and predict their response to novel conditions.

Concepts from complex adaptive systems as a framework for individual-based modelling

Abstract Individual-based models (IBMs) have long been proposed as a key tool for understanding and predicting ecosystem complexities, yet the contribution of this approach to basic or applied ecology has been less than anticipated. Fundamental reasons for the disappointing contribution of IBMs have been, in the current absence of a theoretical foundation for IBMs, conceptual flaws in model formulation and the failure to address critical computer implementation issues. Researchers in the new field of Complex Adaptive Systems (CAS) study how complex behaviors emerge in systems of relatively simple interacting individuals. Research on CAS, while still new and informal, has identified key concepts for making individual-based systems realistic. I propose that explicit consideration of the following concepts from CAS should make the design of IBMs less ad hoc and more likely to produce models of value for basic and applied ecology: (1) Emergence: what behaviors and population dynamics should emerge from the models mechanistic representation of key processes vs. being imposed on the model as empirical relations? How should individual traits be modeled so that realistic population responses emerge?; (2) Adaptation: given the models temporal and spatial scales, what adaptive processes of individuals should be modeled? What mechanisms do individuals use to adapt in response to what environmental forces?; (3) Fitness and strategy: what measures of fitness are appropriate to use as the basis for modelling decision making? Should fitness measures change with life history state?; (4) State-based responses: how should decision processes depend on an individuals state?; (5) Prediction: anticipating decision outcomes appears essential for modelling many behaviors; what are realistic assumptions about how organisms predict the consequences of decisions?; (6) Computer implementation: what user interfaces are necessary to make the model, and especially individual behaviors, observable and testable? How will the models full design and computer implementation be documented and tested so results are reproducible and valid?

Movement rules for individual-based models of stream fish

Abstract Spatially explicit individual-based models (IBMs) use movement rules to determine when an animal departs its current location and to determine its movement destination; these rules are therefore critical to accurate simulations. Movement rules typically define some measure of how an individual’s expected fitness varies among locations, under the assumption that animals make movement decisions at least in part to increase their fitness. Recent research shows that many fish move quickly in response to changes in physical and biological conditions, so movement rules should allow fish to rapidly select the best location that is accessible. The theory that a fish’s fitness is maximized by minimizing the ratio of mortality risk to food intake is not applicable to typical IBM movement decisions and can cause serious errors in common situations. Instead, we developed fitness measures from unified foraging theory that are theoretically and computationally compatible with individual-based fish models. One such fitness measure causes a fish to select habitat that maximizes its expected probability of survival over a specified time horizon, considering both starvation and other risks. This fitness measure is dependent on the fish’s current state, making fish with low energy reserves more willing to accept risks in exchange for higher food intake. Another new measure represents the expectation of reaching reproductive maturity by multiplying expected survival by a factor indicating how close to the size of first reproduction the fish grows within the time horizon. One of the primary benefits of the individual-based approach is avoiding the need for simplifying assumptions; this benefit is best realized by basing movement decisions on such simple, direct measures of fitness as expected survival and expected reproductive maturity.

Bioenergetics Modeling of Stream Trout Growth: Temperature and Food Consumption Effects

Abstract We investigated bioenergetics modeling of growth as an approach for assessing the effects of temperature changes on stream dwelling rainbow trout Oncorhynchus mykiss. Study objectives were (1) to determine the relative effect of temperature versus food consumption on model-predicted growth and (2) to identify relationships between model-predicted food consumption and commonly measured environmental variables. A bioenergetics model for rainbow trout was calibrated to apparent age-1 growth in summer and fall–spring periods for 10 years at eight Sierra Nevada, California, study sites. Model analyses showed that the observed year-to-year variation in summer growth was related to food consumption but not to temperature and that temperature was more important, but still of secondary importance, to observed variation in fall–spring growth. Growth at all sampling sites appeared lower and more variable in summer than in other seasons, and variation among sites and years in the food consumption parameter P...

WHAT CAN HABITAT PREFERENCE MODELS TELL US? TESTS USING A VIRTUAL TROUT POPULATION

Habitat selection (“preference”) models are widely used to manage fish and wildlife. Their use assumes that (1) habitat with high animal densities (highly selected habitat) is high quality habitat, and low densities indicate low quality habitat; and (2) animal populations respond positively to the availability of highly selected habitat. These assumptions are increasingly questioned but very difficult to test. We evaluated these assumptions in an individual-based model (IBM) of stream trout that reproduces many natural complexities and habitat selection behaviors. Trout in the IBM select habitat to maximize their potential fitness, a function of growth potential (including food competition) and mortality risks. We know each habitat cells intrinsic habitat quality, the fitness potential a trout in the cell would experience in the absence of competition. There was no strong relation between fitness potential and the density of fish in the IBM; cells where fitness potential was high but density low were com...

Tests of theory for diel variation in salmonid feeding activity and habitat use

For many animals, selecting whether to forage during day or night is a critical fitness problem: at night, predation risks are lower but feeding is less efficient. Habitat selection is a closely related problem: the best location for nocturnal foraging could be too risky during daytime, and habitat that is safe and profitable in daytime may be unprofitable at night. We pose a theory that assumes animals select the combination of daytime and night activity (feeding vs. hiding), and habitat, that maximizes expected future fitness. Expected fitness is approximated as the predicted probability of surviving starvation and predation over a future time horizon, multiplied by a function representing the fitness benefits of growth. The theorys usefulness and generality were tested using pattern-oriented analysis of an individual-based model (IBM) of stream salmonids and the extensive literature on observed diel behavior patterns of these animals. Simulation experiments showed that the IBM reproduces eight diverse patterns observed in real populations. (1) Diel activity (whether foraging occurs during day and/or night) varies among a populations individuals, and from day to day for each individual. (2) Salmonids feed in shallower and slower water at night. (3) Individuals pack more tightly into the best habitat when feeding at night. (4) Salmonids feed relatively more at night if temperatures (and, therefore, metabolic demands) are low. (5) Daytime feeding is more common for life stages in which potential fitness increases more rapidly with growth. (6) Competition for feeding or hiding sites can shift foraging between day and night. (7) Daytime feeding is more common when food avail- ability is low. (8) Diel activity patterns are affected by the availability of good habitat for feeding or hiding. We can explain many patterns of variation in diel foraging behavior without assuming that populations or individuals vary in how inherently nocturnal or diurnal they are. Instead, these patterns can emerge from the search by individuals for good trade- offs between growth and survival under different habitat and competitive conditions.

Making predictions in a changing world: The benefits of individual-based ecology

Ecologists urgently need a better ability to predict how environmental change affects biodiversity. We examine individual-based ecology (IBE), a research paradigm that promises better a predictive ability by using individual-based models (IBMs) to represent ecological dynamics as arising from how individuals interact with their environment and with each other. A key advantage of IBMs is that the basis for predictions—fitness maximization by individual organisms—is more general and reliable than the empirical relationships that other models depend on. Case studies illustrate the usefulness and predictive success of long-term IBE programs. The pioneering programs had three phases: conceptualization, implementation, and diversification. Continued validation of models runs throughout these phases. The breakthroughs that make IBE more productive include standards for describing and validating IBMs, improved and standardized theory for individual traits and behavior, software tools, and generalized instead of system-specific IBMs. We provide guidelines for pursuing IBE and a vision for future IBE research.

Effects of land use on bird populations and pest control services on coffee farms

Significance How land use can support both natural and agricultural resources is a key sustainability question. This question is more complicated when nature and agriculture support each other; here, birds consume pest insects, and coffee provides bird habitat. We used a model based on field research to study the effects of agricultural and natural land uses on birds, the pest control services that birds provide, and crop production on Jamaican coffee farms. Our conclusions address whether small forest-like areas are more valuable for pest control than large intact forest, how the size and number of natural patches affect birds and crop production, and whether replacing sun coffee (a monoculture) with more diverse and tree-rich shade coffee increases pest consumption to a degree sufficient to offset the reduced yield. Global increases in both agriculture and biodiversity awareness raise a key question: Should cropland and biodiversity habitat be separated, or integrated in mixed land uses? Ecosystem services by wildlife make this question more complex. For example, birds benefit agriculture by preying on pest insects, but other habitat is needed to maintain the birds. Resulting land use questions include what areas and arrangements of habitat support sufficient birds to control pests, whether this pest control offsets the reduced cropland, and the comparative benefits of “land sharing” (i.e., mixed cropland and habitat) vs. “land sparing” (i.e., separate areas of intensive agriculture and habitat). Such questions are difficult to answer using field studies alone, so we use a simulation model of Jamaican coffee farms, where songbirds suppress the coffee berry borer (CBB). Simulated birds select habitat and prey in five habitat types: intact forest, trees (including forest fragments), shade coffee, sun coffee, and unsuitable habitat. The trees habitat type appears to be especially important, providing efficient foraging and roosting sites near coffee plots. Small areas of trees (but not forest alone) could support a sufficient number of birds to suppress CBB in sun coffee; the degree to which trees are dispersed within coffee had little effect. In simulations without trees, shade coffee supported sufficient birds to offset its lower yield. High areas of both trees and shade coffee reduced pest control because CBB was less often profitable prey. Because of the pest control service provided by birds, land sharing was predicted to be more beneficial than land sparing in this system.

Exploring the Persistence of Stream-Dwelling Trout Populations under Alternative Real-World Turbidity Regimes with an Individual-Based Model

Abstract We explored the effects of elevated turbidity on stream-resident populations of coastal cutthroat trout Oncorhynchus clarkii clarkii using a spatially explicit individual-based model. Turbidity regimes were contrasted by means of 15-year simulations in a third-order stream in northwestern California. The alternative regimes were based on multiple-year, continuous monitoring in two streams. Turbidity affected model fish by reducing both their risk of predation and their reactive distance to drifting prey. It did not affect their ability to locate nondrifting food, such as invertebrates on the stream bottom. Under a calibration scenario that assumed trout predominantly consume drifting prey, the less-turbid real-world regime produced relatively stable abundance across years (similar to field observations) whereas the more-turbid regime (under otherwise identical physical conditions) resulted in extinction within the 15-year simulation period. Additional simulations revealed sensitivity to the relat...