Here we emphasize on the main features of the model structure.IBASAM is fully described by Piou & Prévost(2012).
In the river phase, individuals grew in weight according to individual and stage-dependent growth capacity and influenced by water temperature, population density and river flow. Growth increments in weight were then allocated to fat reserves (Fat) or somatic growth through an increase in body length depending on a variable individual propensity to accumulate fat. Survival in the river was phase dependent, with higher mortality for maturing individuals and during winter. The triggering of sea migration was size dependent 6 months before the run. The smoltification process allows an individual that was in the river (‘parr’) to become physiologically ready to run into the sea (as ‘smolt’). The probability of smolting for an individual followed a reaction norm based on its body length (Buoro, Prévost & Gimenez 2010).
Once at sea, individuals grew in weight following a Gompertz function depending on specific individual characteristics and overall oceanic conditions. In this function, a NoiseSeat factor repre- sented a daily environmental condition for growth. In the absence of EC, NoiseSeat was simply a normal random number centred on MeanNoiseSea = 1 and of variance 0 1. With EC, MeanNoiseSea decreased through time and NoiseSeat was a normal random number with variance 0 1 but centred on the MeanNoiseSea of the year. Each replicate of simulations had consequently different environmental conditions. Fat and body length accumulation were allocated as in the river phase. Survival at sea was size dependent with a clear disadvantage for small individuals.
Six months before reproduction time for riverine parr or returning time for anadromous individuals, the increase in fat content over a temporal window was used as conditioning variable triggering maturation. For an individual j, the binary indica- tor of the maturation status Matj was set to 1 when its ProjectedFatTheoryj value was above the threshold of maturation pFmidj. This threshold was represented as the phenotypic expres- sion of an underlying genetically coded trait gFmidj. The later varied individually with an average level dependent on the sex and location of the individual (in river vs. at sea). The ProjectedFatTheoryj was calculated as a linear projection of Fatj over 6 months given the daily rate of change in Fatj of the individual over an evaluation window (61 days in spring for riverine parr, 47 days in autumn for anadromous individuals at sea) (Thorpe et al. 1998).
The various thresholds (sex and location dependent:
gFmida = male parr in river, gFmidb = female parr in river, gFmidc = anadromous males at sea, gFmidd = anadromous females at sea) were transmitted by the two parents of a fertilized egg to their progeny according to a bi-allelic multilocus system. Thus, for each threshold, each individual had two branches of 20 loci of 0 (unfavourable) or 1 (favourable) to code for the genetic value. The link between the genetic parameters and their phenotypic expression was controlled by a heritability value ($h^2$). This heritability served to decompose the phenotypic variance of the thresholds between genetic and environmental components at the initialization of the simulations. The environmental variance was then applied on the phenotypic expression of the genetically coded thresholds (pFmid expression of gFmid with environmental noise). No mutations were assumed in the model.
The reproduction events were simulated according to relevant S. salar literature (Fleming 1996). The mating system allowed mature male parr to fertilize a fraction of the eggs. It selected anadromous males following size dominance while allowing satel- lite males to fertilize a significant fraction of the eggs. The number of egg per female was size dependent. Egg-to-emergence survival was water temperature, river flow and density dependent.
The simulations presented hereafter were conducted with the parameterization proposed by Piou & Pr evost (2012). The initial simulated population was randomized in structure and size to correspond to 25% of mid-1990s average Scorff salmon popula- tion (Piou & Pr evost 2012). As in Piou & Pr evost (2013), fishery mortality on returning individuals was simulated by the removal of a proportion of each sea age class (1SW or MSW, base pro- portion of 15%) from the population at the end of summer step, that is after the observation of number of returns and propor- tions of MSW.