To improve our understanding and management of species’ response to environmental and anthropogenic changes, there is a critical need for integrative modeling approaches that encompass demographic, ecological and evolutionary processes in a unified framework. This is especially crucial for anadromous salmonid fishes that hold significant socio-economic value. Mechanistic models, particularly Demo-Genetic Agent-Based Models (DG-ABMs), are promising tools to address this challenge. Building on previous work, we present IBASAM, an Individual Based Anadromous SAlmonids Model. It incorporates significant advancements, including new processes such as dispersal and in-river movement, enhanced flexibility for simulating diverse spatial configurations- from river dendritic networks to metapopulations and multiple marine areas- and the integration of variation in life history phenology and the genetic architecture of traits. Additionally, a user-friendly interface has been developed to facilitate broader application. This improved versatility enables IBASAM application to a number of populations and species of anadromous salmonids, making it well-suited for a wide range of ecological and evolutionary studies. The model is particularly suited for investigating (meta)population eco-evolutionary dynamics, persistence, and adaptation to environmental changes, as well as exploring management options including fisheries regulation and river connectivity restoration. We summarize the model’s structure and illustrate some of its novelties, parameterization, and emerging patterns with the case study of a naturally reproducing Atlantic salmon population simulated over 100 years, with constant fishing rates and no climatic trend. Our simulations successfully reproduced 30 variables observed over 30 years in the Scorff population. IBASAM opens up new opportunities for research on the response of salmonid meta-populations to environmental changes, the exploration of management practices, with further developments possible.

Demo-Genetic Agent-Based Model, Anadromous salmonids, Life history strategies, Evolution, Genetic architecture, Phenotypic plasticity, Dispersal, Metapopulation, Management

Species are encountering rapid environmental changes and multiple anthropogenic pressures that threaten their persistence. Assessing their ability to respond necessitates the integration of eco-evolutionary processes and their underlying mechanisms into prospective models, alongside environmental dynamics. Phenotypic plasticity, genetic adaptation of phenotypic and phenological traits, and dispersal are the fundamental components of species’ adaptive capacity (Beever et al. 2016; Thurman et al. 2022), enabling populations persistence and resilience to environmental changes. This highlights the need for modeling approaches that span scales, from genetic mechanisms at the individual level to dynamics at the metapopulation scale (Baguette et al. 2017; Romero-Mujalli et al. 2019; Thorogood et al. 2023). Ultimately, such comprehensive models provide valuable insights into the interplay of eco-evolutionary processes and environmental changes, enhancing our ability to manage species responses to future challenges (Seaborn et al. 2023).

Demo-Genetic (or Eco-Genetic) Agent-Based Models (DG-ABMs) offer promising avenues to tackle these challenges (Xuereb et al. 2021). These mechanistic models are well-suited for evaluating species’ persistence because they explicitly (i) include eco-evolutionary processes and their underlying mechanisms, and (ii) account for inter-individual variation and interactions that are key drivers of selection. Importantly, DG-ABMs do not rely on predefined fitness functions; instead, fitness and population-level properties emerge from individual decisions, interactions, and feedbacks, shaping genetic diversity and population dynamics (Lamarins et al. 2022b). Various generic DG-ABM platforms exist for simulating (meta)populations, such as NEMO (Guillaume and Rougemont 2006), SLIM (Messer 2013), or CDMetaPOP (Landguth et al. 2017). While flexible (e.g., allowing addition of life history traits or life cycle events), these generic models may not be suitable for species with complex life cycles.

Anadromous salmonid fishes, which migrate between freshwater and marine habitats, exhibit complex life cycles and diverse life history strategies (Erkinaro et al. 2019; Persson et al. 2023). Hence, their response to threats such as migration barriers (Lindley et al. 2006), climate change (Crozier et al. 2021), and overfishing (Christensen et al. 2003), are challenging to predict. Beyond their ecological significance, numerous anadromous salmonid species have substantial socio-economic value due to their cultural heritage and their importance in both recreational and commercial fisheries across the globe (Woo and Noakes 2014; Kochalski et al. 2022). Additionally, they are emblematic model species for studying eco-evolutionary dynamics (e.g., Hendry and Stearns 2004; Primmer 2011), and evidence of dispersal also suggests metapopulation dynamics (Keefer and Caudill 2014; Lamarins et al. 2024) despite their strong philopatric behavior. Given the extensive knowledge of salmonid population dynamics, from genetic architecture to demographic processes, a comprehensive simulation tool is critically needed to integrate this information holistically. Such a tool would enhance our understanding of their future trajectories under environmental and anthropogenic pressures while enabling the evaluation of management and conservation strategies.

IBASAM (Individual Based Atlantic SAlmon Model, Piou and Prévost 2012) and its extension MetaIBASAM (Lamarins et al. 2022a), were developed to assess the effect of stressors on evolutionary demography of a single population or a set of interconnected Atlantic salmon populations respectively. However, they have key limitations. First, they are not fully spatially explicit and are limited to the river scale, preventing the simulation of finer-scale structures (e.g., upstream / downstream gradient, estuaries, dendritic networks) and within-river movements influenced by river connectivity. Also, MetaIBASAM simulates metapopulations by running several IBASAM models in parallel with individuals exchanged between them, which limits the addition of dispersal mechanisms requiring explicit inter-population feedbacks such as density-dependent immigration (Berdahl et al. 2016). Additionally, it assumes a single marine domain, despite evidence of multiple migration and growth areas with varying conditions (Rikardsen et al. 2021). Second, IBASAM’s temporal scale limits its ability to capture variation in the phenology of life histories, despite growing evidence of phenological shifts in response to environmental changes (Kovach et al. 2013; Woods et al. 2022). Third, previous versions lack the capacity to simulate complex genetic architectures, relying on a polygenic framework without locus-specific or dominance effects. Yet, advances in genomics highlight the role of complex genetic architectures of traits and major-effect loci, such as sex-dependent dominance in salmon age at maturity (Barson et al. 2015), which have implications for genetic diversity, evolutionary potential, and conservation in salmonids (Waples et al. 2022). Simulating such architectures of adaptive traits could advance our understanding of species’ adaptive capacities (Xuereb et al. 2021). Lastly, IBASAM currently lacks an end-user graphical interface, limiting accessibility for a wider scientific audience, management applications and stakeholder engagement (McInerny et al. 2014; Gaydos et al. 2021).

Here, we present a major revision of IBASAM (now standing for Individual Based Anadromous SAlmonids Model) addressing previous limitations and providing enhanced flexibility in various aspects of salmonid life cycle representation, within a fully spatially explicit framework. Key advances include flexible spatial scaling (from river sections with variable connectivity to metapopulations and multiple marine areas), improved phenology, complex genetic architectures of traits, and a user-friendly interface. We provide a summarized description of the model, with additional details available in the Suppl. material 1. To demonstrate its capabilities, we simulate ~ 30 years of past and 70 years of future evolution of a virtual population mimicking a comprehensively monitored population of Atlantic salmon in France, assuming constant fishing rates and no climatic trends. However, the model is adaptable to other anadromous salmonid species and river systems. Finally, we discuss the model’s new features and highlight opportunities for further development.

IBASAM is implemented in the Capsis simulation platform (https://capsis.cirad.fr, Dufour-Kowalski et al. 2012), offering a user-friendly interface (Suppl. material 1: note 1). The Capsis platform can also be launched from R, enabling users to run large-scale simulations and perform more advanced statistical analyses. The model executable file, Java source codes, R scripts to create initialization files and launch the model, as well as a comprehensive user guide are all available at https://doi.org/10.5281/zenodo.14825431. A complete, detailed model description, following the ODD (Overview, Design concepts, Details) protocol for describing individual- and agent-based models (Grimm et al. 2006, 2020) is provided in Suppl. material 1: note 2. Here, we provide an overview of the model features (Fig. 1 and Suppl. material 1: fig. S9).

Figure 1. 

Visualization of IBASAM following the vODD format (Szangolies et al. 2024). At initialization, fish individuals with their state variables are generated in river sections (reaches, separated by weirs and eventually grouped by populations) and sea regions. Several submodels (circled numbers) are run on a daily basis following the salmon life cycle. Population and individual metrics can then be derived from a simulation run (e.g., 50 years) according to several scenarios.

The overall purpose of IBASAM is to investigate how selective fishing, environmental changes (e.g., variation in water temperature, discharge, and growth conditions), and management actions (e.g., spatially defined fishing, conservation measures, and connectivity restoration by dam removal) influence the eco-evolutionary dynamics of salmonid populations. This version of IBASAM enables the exploration of two key aspects: (1) the potential effects of metapopulation dynamics through dispersal on the evolution of life history traits, population stability, persistence, and resilience, and (2) the influence of complex genetic architectures underlying life history traits. To consider our model realistic enough for its purpose, one can use patterns of population dynamics, including changes in demographic, phenotypic and genotypic (e.g., allele frequencies) patterns over time. Here, we evaluate the model’s ability to reproduce a variety of patterns using observed time-series of abundance, size/mass/date distributions at different ages/life stages, and life history strategies (i.e., proportion of early maturation, years at sea) of a long-term monitored population (see Fig. 2).

The model includes the following entities: individuals are fish, redds (spawning nests) and fry siblings; and spatial units (river, population, reach, weir and sea region; Fig. 1). The state variables characterizing these entities are listed in Suppl. material 1: note 4, table S1. Fish belong to a population (a gene pool) and have a status (parr, smolt, anadromous), unique values of body size, weight, fat reserve, dates of life history events (e.g., migration, reproduction) and a genetic material coding for both genotypic and phenotypic values of heritable traits (neutral trait, growth potential, philopatry propensity, migration and maturation thresholds). Redds are created after reproduction and record the genetic material of the mother and potentially multiple fathers, along with the number and weight of the fertilized eggs. To limit computational time and because offspring suffer high mortality during the first weeks after emergence, redds produce new super-individuals called fry siblings, separated by sex, representing the total number of offspring issuing from the fertilized eggs of two parents before being individualized as fish. Each fry sibling is characterized by attributes similar to fish with an additional number of siblings. Regarding the spatial and temporal resolution and extent, this version of IBASAM is spatially explicit, capable of simulating complex watersheds comprising multiple rivers and/or interconnected basins, as well as several marine regions. Rivers can be further subdivided into single or multiple reaches, separated by weirs with user-defined passability (connectivity). By default, a river is a population, but it can contain several populations (groups of reaches). Reaches and sea regions are the spatial units where processes unfold (Fig. 1). Each reach and sea region is defined by distinct characteristics, including habitat area, environmental conditions (e.g., water temperature and growth conditions), and fishing intensity. These conditions may vary between spatial units but are assumed to be uniform within each unit. The basic time step for all individual processes is daily, with some processes occurring only within a time window (i.e., maturation, migration and reproduction processes). Simulation can be extended over multiple years to generate successive generations and monitor long-term dynamics (e.g. ~ 20 generations over 50 years for a southern European Atlantic salmon population).

The most important processes of the model, which are repeated every time step, are ten submodels representing critical life events and processes of salmonid life cycle: 1) reproduction and redd (nest) creation, 2) emergence from the redd, 3) genetic coding and transmission, 4) growth, 5) survival, 6) seaward migration decision, 7) sexual maturation, 8) migration (including inter-basin dispersal), 9) movement (downstream/upstream, including intra-basin dispersal) and 10) fishing (Fig. 1). Processes are applied daily, tailored to the location (river or sea, population), individual status (e.g., juvenile, anadromous, or mature), and temporal factors (e.g., season, defined time windows). The simulation begins on an initial date specified by the user. The model follows a predefined sequence of processes, starting with the river phase and then the marine phase (Box 1). A detailed explanation of the underlying processes and mechanisms is presented in Suppl. material 1: note 2.

Figure 2. 

Times series of simulation results (distribution over 50 replicates, orange) with the retained parameters values (Suppl. material 1: table S3) versus observations from the monitoring of the Scorff river (distribution over observations, black) per life stage and metrics. Metrics include the number of individuals (A; simulation results scaled up to the real population size), the body size (B) and weight (C), the proportion of mature individuals (D1–D2) and the proportion of males (D3–D6), the proportion of iteroparous individuals (D7), and the date of life history events (E). Distributions show the median (point) and inter-quantile intervals (50% and 95%, lines). Horizontal dashed black lines represent the assumed date of emergence (E1) and number of parr 1+ (A2), based on expert knowledge.

The most important design concepts of the model are emergence, adaptation, and observation. All individual life history traits (growth, survival, maturation, migration, reproduction) emerge from a mechanistic framework that integrates environmental variables, interactions with other individuals, demographic dynamics and genetic structures, themselves emerging from individual life histories via an eco-evolutionary feedback (Lamarins et al. 2022b). Indeed, maturation, migration, philopatry and growth potential are adaptive traits which influence individual fitness. They result from the interplay of a genetic basis (e.g. via thresholds) and environmental conditions (e.g. via environmental cue) through adaptative phenotypic plasticity. All individual traits or summary metrics (e.g. number of individuals, median size) can be recorded at specific time steps (e.g., day 91 and 274 on the calendar of each year) or at life-cycle events (e.g. during seaward / riverward migration, reproduction) defined at initialization by the user.

The model is initialized with a spatial structure (numbers and locations of rivers, reaches, populations, weirs, and sea regions), environmental conditions and fishing rates in river reaches and sea regions, river weirs passability rates, genetic architecture of traits, and virtual populations with individual fish at every individual life stage (parr, smolt, anadromous). Redds and fry siblings are not created at initialization. Model dynamics are driven by population parameters values (Suppl. material 1: table S3), spatial structure and genetic architecture of traits. Input data consist of daily times’ series of reach (water temperature, water flow, fishing rates), weir (passability rates) and sea region (growth condition, fishing rates) conditions (Suppl. material 1: note 5).

Key processes that were added in the model are the transmission of heritable traits with flexible genetic architecture, phenology variation, dispersal and in-river movement. Individual traits such as maturation and migration thresholds, growth potential and philopatry propensity are genetically heritable, combining quantitative genetics with Mendelian inheritance (Suppl. material 1: fig. S4). Trait phenotypic expression results from the additive effects of genetic and environmental components, with genotypic values controlled by multiple independent bi-allelic loci. While the model excludes mutation (given the low number of generations) and inter-locus interactions (epistasis), it allows for variation in locus effect sizes and intra-locus interactions (i.e. dominance effects). Some traits exhibit sex-, life stage-, or location-dependent expression. This update aligns with recent advances in understanding the genetic architecture of maturation in male parr (Lepais et al. 2017; Debes et al. 2021) and anadromous salmon (Barson et al. 2015; Sinclair-Waters et al. 2020). Maturation threshold now has a unique genetic basis, but its expression varies by sex and life stage (parr vs anadromous), allowing for the implementation of sex-dependent dominance patterns. Considering recent work on the sex-specific migration propensity of brown trout (Kane et al. 2022; Lavender et al. 2024), a similar framework has been applied to the expression of seaward migration decision. The migration threshold also has a unique genotypic basis with sex-specific phenotypic expression. Similarly, the river and sea growth potentials have a unique genotypic basis, but their phenotypic expression varies between river and sea. This update differs from previous IBASAM versions, which treated traits independently by sex and life stage. The genetic architecture can also differ from one population to another (named as gene pools), either by only adjusting allele frequencies or by modifying the entire genetic map (allele effects, dominance, etc.). A neutral trait is also genetically heritable but does not influence any model processes - it serves to track genetic drift. Additionally, this version includes variation in life history phenology, such as migration timing, which depends on the selection of time windows and parameter values (e.g. migration thresholds). Finally, dispersal between rivers basins, along with in-river movements processes and intra-basin dispersal, are key additions to the model. Dispersal decision is driven by individual philopatric propensity - which is here genetically determined - and the selection of the destination river is based on distance and attractiveness following a dispersal kernel. In-river movements are simulated for both smolts downstream migration and anadromous upstream migration, using implicit or explicit methods based on habitat attractiveness (altitude, size, sex ratio) and weir passability trials.

IBASAM (now standing for Individual Based Anadromous SAlmonids Model) introduces several major advancements compared to the previous release as outlined in Suppl. material 1: table S17, including the implementation of key processes such as dispersal and in-river movement. The model offers enhanced flexibility to simulate a wide range of spatial configurations, from dendritic river networks to metapopulations and multiple marine areas. IBASAM also incorporates variation in life history traits, phenology, and underlying genetic architectures. These additions significantly expand the model’s spatial and temporal realism, without increasing computational cost or sacrificing generality. This enhanced versatility facilitates the adaptation of IBASAM to diverse populations and species of anadromous salmonids, making it well-suited for a broad spectrum of ecological and evolutionary studies. To support wider adoption, a user-friendly interface has been developed and the model’s integration within the Capsis platform facilitates code and library sharing among a community of modelers, fostering collaboration and iterative improvements. We therefore strongly recommend that users adopt this version of IBASAM.

We used the example of a French Atlantic salmon population to demonstrate the ability of IBASAM to successfully reproduce 30 demographic and phenotypic patterns of field observations over approximately 30 years. Under a neutral scenario (i.e., no climatic trends or changing anthropogenic pressures), we highlighted emerging properties at demographic, phenotypic, and genotypic levels, including sex differences in sea age (or age at maturity) and in the intensity of selection on the maturation threshold. The emerging correlations among traits or loci (linkage disequilibrium) may be particularly relevant, as they could influence the population’s response to environmental or anthropogenic selective pressures. However, the simulations presented here are inherently influenced by simplifying assumptions (e.g., no emigration/immigration, no climate change). Consequently, our findings should not be interpreted as exact representations of the natural population’s dynamics taken as a reference but rather as an illustration of IBASAM’s potential applications. This example serves as a foundation for further applied or hypothesis-driven studies under realistic environmental scenarios.

The model was coded on the Capsis-4 modelling platform (http://capsis.cirad.fr) maintained by François de Coligny and Nicolas Beudez. We thank Miguel de Navascués for discussions on the ABC approach. We acknowledge CSC (IT Center for Sciences, Finland) for access to computational resources.