Vincent Ginot

A multi-agents architecture to enhance end-user individual-based modelling

Abstract The increasing importance of individual-based modelling (IBM) in population dynamics has led to the greater availability of tools designed to facilitate their creation and use. Yet, these tools are either too general, requiring the extensive knowledge of a computer language, or conversely restricted to very specific applications. Hence, they are of little help to non-computer expert ecologists. In order to build IBMs without hard coding them nor restricting their scope too much, we suggest a component programming, assuming that each elementary task that forms the behaviour of an individual often follows the same path: an individual must locate and select information in order for it to be processed, then he must update his state, the state of other individuals, or the state of the rest of the ‘world’. This sequence is well suited to translation into elementary computerised components, that we call primitives. Conversely, task building will involve stringing out well-chosen primitives and setting their parameter values or mathematical formulae. In order to restrict the number of primitives and to simplify their use, ‘information’ must be carried through well defined structures. We suggest the use of the multi-agents system paradigm (MAS) which originates from the distributed artificial intelligence and defines agents as autonomous objects that perceive and react to their environment. If one assumes that a model can be described entirely with the help of agents, then primitives only handle agents, agent state or history. This greatly simplifies their conception and enhances their flexibility. Indeed, only 25 primitives, split into six groups (locate, select, translate, compute, end, and workflow control) proved to be sufficient to build complex IBMs or cellular automata drawn from literature. Furthermore, such a primitive-based multi-agents architecture is very flexible and facilitates all the steps of the modelling process, in particular the simulation engine (agents call and synchronisation), the results analysis, and the simulation experiments. Component programming may also facilitate the design of a domain specific language in which these models could be written and exported to other simulation platforms.

The early drug selection of nematodes to anthelmintics: stochastic transmission and population in refuge.

We have developed an individual-based model to reflect the complexity of the early phase of drug resistance selection in a nematode/sheep model. The infection process consists of the stochastic ingestion of infective larvae spatially aggregated in clumps. Each clump corresponds to infective larvae, which are the offspring of the mature nematodes from a given sheep. We studied the dynamics of the parasitic population and the frequency of the recessive resistance alleles during selection by anthelmintic treatments. The interaction between genetic and demographic processes illustrated the trade-off between the control of the infection and the delay of resistance selection. We confirmed the importance of the number of treatments and their timing. The same treatment frequency may result in different outcomes on resistance selection in relation to the size of the refuge (infective larvae on pasture). Treatment applied during the summer (when the mortality of infective larvae on pasture was high), may lead to a rapid selection of drug resistance and a lack of control of sheep and pasture contamination. We showed that higher stocking rates were also a force in promoting the resistance allele selection.

Modelling macroparasite aggregation using a nematode-sheep system: the Weibull distribution as an alternative to the negative binomial distribution?

Macroparasites are almost always aggregated across their host populations, hence the Negative Binomial Distribution (NBD) with its exponent parameter k is widely used for modelling, quantifying or analysing parasite distributions. However, many studies have pointed out some drawbacks in the use of the NBD, with respect to the sensitivity of k to the mean number of parasites per host or the under-representation of the heavily infected hosts in the estimate of k. In this study, we compare the fit of the NBD with 4 other widely used distributions on observed parasitic gastrointestinal nematode distributions in their sheep host populations (11 datasets). Distributions were fitted to observed data using maximum likelihood estimator and the best fits were selected using the Akaikes Information Criterion (AIC). A simulation study was also conducted in order to assess the possible bias in parameter estimations especially in the case of small sample sizes. We found that the NBD is seldom the best fit for gastrointestinal nematode distributions. The Weibull distribution was clearly more appropriate over a very wide range of degrees of aggregation, mainly because it was more flexible in fitting the heavily infected hosts. Moreover, the Weibull distribution estimates are less sensitive to sample size. Thus, when possible, we suggest to carefully check on observed data if the NBD is appropriate before conducting any further analysis on parasite distributions.

Improving mesocosm data analysis through individual-based modelling of control population dynamics: a case study with mosquitofish (Gambusia holbrooki)

Experimental ecosystems such as mesocosms have been developed to improve the ecological relevance of ecotoxicity test. However, in mesocosm studies, the number of replicates is limited by practical and financial constraints. In addition, high levels of biological organization are characterized by a high variability of descriptive variables. This variability and the poor number of replicates have been recognized as a major drawback for detecting significant effects of chemicals in mesocosm studies. In this context, a tool able to predict precisely control mesocosms outputs, to which endpoints in mesocosms exposed to chemicals could be compared should constitute a substantial improvement. We evaluated here a solution which consists in stochastic modelling of the control fish populations to assess the probabilistic distributions of population endpoints. An individual-based approach was selected, because it generates realistic fish length distributions and accounts for both individual and environmental sources of variability. This strategy was applied to mosquitofish (Gambusia holbrooki) populations monitored in lentic mesocosms. We chose the number of founders as a so-called “stressor” because subsequent consequences at the population level could be expected. Using this strategy, we were able to detect more significant and biologically relevant perturbations than using classical methods. We conclude that designing an individual-based model is very promising for improving mesocosm data analysis. This methodology is currently being applied to ecotoxicological issues.

Mobidyc, a Generic Multi-Agents Simulator for Modeling Populations Dynamics

We present a simulator designed to help scientists who are not experts in computing, such as biologists, to build up, run, and update their models in the field of fish population dynamics. This simulator, written in Smalltalk, is built upon multi-agents concepts and is well suited to model systems where individual or spatial aspects are involved. It provides tools to set the state and the behavior of the agents and to set their environment. The pre-defined components should preferably be used, but help for customizing components or programming new ones i s supplied, in order to build more complex agents. The simulator manages qualitative and quantitative variables with simple or composed units, dependencies on parameters, synchronous and asynchronous modes within and between agents, and the importation of ASCII files for the scenario scripts or for the initial stocking of the agents. Four examples of applications are presented.