Virtual Salmon: New Model Simulates a Century of Life, Genes and Survival in Endangered Fish

Atlantic salmon have become one of the most closely watched fish on the planet, and for good reason. Their populations are squeezed from every direction by warming rivers, dams that block ancient migration routes, and fishing pressure that removes the largest and oldest individuals before they can spawn. Predicting how these iconic fish will cope with such pressures has long frustrated scientists, because salmon lives weave together genetics, behavior, and shifting environments across two very different worlds: freshwater rivers and the open ocean. A research team now unveils a major upgrade to a simulation tool designed to capture that complexity in unprecedented detail, and its first century-long virtual experiment offers a striking preview of what modern evolutionary modeling can do.

The tool, called IBASAM, short for Individual Based Anadromous SAlmonids Model, is described in the journal Individual-based Ecology. It belongs to a class of models known as demo-genetic agent-based models, or DG-ABMs, in which every fish is simulated as an individual agent carrying its own genes, body size, fat reserves, and behavioral rules. Rather than imposing a predefined measure of evolutionary fitness, these models let fitness emerge from the bottom up, as individuals make decisions about growth, migration, and reproduction and then live with the consequences. That mechanistic approach is what sets IBASAM apart from simpler population models that treat salmon as statistical averages rather than living, choosing animals.

The new version solves several long-standing limitations of its predecessors. Earlier iterations could simulate only a single river population or, in the metapopulation extension, run several independent copies in parallel with fish exchanged between them. The upgraded model is fully spatially explicit: rivers can be divided into reaches separated by weirs with user-defined passability, multiple rivers can be linked into true metapopulations with density-dependent dispersal, and the ocean is no longer a single uniform box but a set of distinct marine regions with different growth and survival conditions. Fish can now move within rivers, disperse between river basins, and shift between sea areas, each decision shaped by their genes and by the landscape around them.

Perhaps the most consequential upgrade lies in the genetics. The model now simulates heritable traits using a bi-allelic multilocus system that combines quantitative genetics with Mendelian inheritance, allowing variation in the effect sizes of individual loci and in dominance effects within loci. Crucially, some traits share a single genetic basis across sexes and life stages, with expression modulated by sex or stage. This mirrors real salmonid genomics, where a major locus known as vgll3 exerts sex-dependent dominance on the age at maturity in Atlantic salmon. The maturation threshold, the seaward migration threshold, growth potential in river and sea, and the propensity to return to natal waters are all genetically coded and transmitted from parents to offspring, while a neutral trait tracks genetic drift without affecting any process.

Phenology, the timing of life history events, also gains new flexibility. Migration, maturation, and reproduction now unfold within user-defined time windows, so the model can capture the seasonal shifts that field biologists increasingly document as climates change. Dispersal itself is under genetic control: each fish carries a philopatric propensity that determines how likely it is to stray to a non-natal river, and the destination is chosen by a dispersal kernel weighting distance and attractiveness. This means that straying, long treated as a nuisance parameter in salmon management, becomes an evolvable adaptive trait that can respond to spatially variable pressures such as localized fishing.

To demonstrate the model’s realism, the team calibrated it against one of the best-monitored salmon populations in Europe: the Atlantic salmon of the Scorff River in Brittany, France, subject to a continuous survey spanning nearly three decades. The researchers simulated the river as 22 reaches separated by 15 weirs, with two distinct sea regions representing conditions experienced by fish that spend one winter at sea versus those spending several. Parameterization combined values drawn from the vast salmonid literature with an Approximate Bayesian Computation procedure that tested 80,000 parameter combinations sampled by Latin hypercube design across 32 simulated years, comparing 32 summary statistics against field observations.

The result is a remarkable validation. Across 50 independent simulation replicates, the model reproduced the observed trajectories of 30 demographic and phenotypic time series, including abundances of juvenile parr, migrating smolts, and returning adults, as well as body sizes, weights, sex ratios, proportions of repeat spawners, and the dates of key life events. The team then extended the simulations to a full century under a neutral scenario of constant fishing and no climatic trend. As expected under stabilizing selection, population abundance, mean sea age, maturation thresholds, and allele frequencies settled into stationary distributions, although a slight decline in the juvenile growth potential hinted at subtle ongoing dynamics even in a stable environment.

Some of the most intriguing findings are patterns that were never programmed in but emerged spontaneously from the simulated ecology. Although the 20 simulated loci were unlinked, the model generated linkage disequilibrium, a statistical covariation among alleles at different loci, for traits under selection, exactly as predicted by classical population genetics through the Bulmer effect. No such disequilibrium appeared for the neutral trait, confirming that selection, not artifact, was responsible. The strength of that selection differed between sexes: the maturation threshold of adult females showed markedly stronger negative disequilibrium than that of males, echoing the sex-dependent genetic architecture documented in real salmon. Correlations also emerged between juvenile growth potential and the migration threshold, suggesting that fish with genetically high migration thresholds need robust growth to cross them and head to sea.

A sensitivity analysis using variance decomposition across another 80,000 parameter sets revealed which knobs matter most. Parameters governing the trade-off between rapid growth and survival in freshwater dominated the model’s behavior, along with the mean genetic growth potentials set at initialization and summer growth activity. Other outcomes depended on more specific parameters, such as the mortality of adults after spawning, which strongly shaped the proportion of repeat spawners. The analysis also uncovered a counterintuitive computational quirk: subdividing a river into more reaches actually speeds simulations up, because the sexual selection process operates within each section rather than across the whole population, cutting the runtime of a 50-year simulation from roughly 45 minutes to about 10.

The implications reach well beyond one French river. Because the model is open-source, runs on the Capsis platform with a user-friendly graphical interface, and can be launched from R for large-scale experiments, the authors intend it as a shared community tool adaptable to other anadromous salmonids, from brown trout to charr. Managers could use it to test dam removal scenarios, fishing regulations, or hatchery interventions before committing real populations to the experiment. Researchers can push it further by adding epistasis, pleiotropy, or genetically based spawning timing, and by exploring how portfolio effects in migration timing buffer populations against environmental shocks. In an era when salmon conservation decisions carry enormous economic and cultural weight, a model that can replay a century of evolution in minutes may prove to be one of the most valuable instruments in the toolbox.

Subject of Research: An individual-based demo-genetic model for simulating the eco-evolutionary dynamics of anadromous salmonid metapopulations

Article Title:  IBASAM : a spatially explicit demo-genetic individual-based model for anadromous salmonids (meta)populations

Article References: Lamarins, A., de Coligny, F., Labonne, J., Prévost, E., Piou, C., Primmer, C. R., & Buoro, M. (2025). IBASAM: a spatially explicit demo-genetic individual-based model for anadromous salmonids (meta)populations. Individual-based Ecology, 1, Article e153305. https://doi.org/10.3897/ibe.1.153305

Image Credits: AI Generated

DOI: 10.3897/ibe.1.153305

Keywords: Atlantic salmon, individual-based model, eco-evolutionary dynamics, metapopulation, genetic architecture, salmonid conservation, dispersal, phenology, Approximate Bayesian Computation, linkage disequilibrium, fisheries management, river connectivity