Initialisation Conclusion

Initialisation is a core component of OSMOSE rather than a secondary technical detail. Because the model is stochastic, school-based, and structured by size, age, and species interactions, the starting state of a simulation must represent much more than total biomass by species. It must provide, either directly or indirectly, a coherent ecological description of how populations enter the simulation.

OSMOSE 4.4 currently supports three main conceptual routes for doing so. Direct school-level initialisation provides the most explicit representation, because the initial schools are given directly in a restart-like file. Reconstructed population-level initialisation provides a practical class-based representation from which the OSMOSE core simulator creates schools at runtime. Egg seeding initialisation provides an indirect, reproduction-based route that can help establish or maintain populations when mature spawners are absent. In addition, the codebase includes utility procedures that allow movement between school-level and population-level representations.

Among these methods, reconstructed population-level initialisation is generally the most suitable default workflow for ordinary applications in OSMOSE 4.4. It offers the most useful balance between realism, control, and practical usability. In particular, it allows users to exploit the standard OSMOSE species parameterisation together with biomass and fishery information to generate a coherent starting state without preparing a complete school-by-school file. For this reason, it occupies the central place in the present documentation.

At the same time, the current initialisation module reflects the historical development of the model. It combines methods conceived for different purposes, along with later extensions introduced to broaden applicability. The resulting flexibility is valuable, but it also makes careful terminology, explicit user guidance, and systematic inspection of the initial state especially important.

For users of OSMOSE 4.4, the main practical recommendation is therefore straightforward: choose the initialisation representation that matches the information available, document clearly which representation is actually being used, and inspect the resulting initial state explicitly before treating it as the starting point of a simulation. In most cases, this means working through reconstructed population-level initialisation and preserving both the inputs used for reconstruction and the generated population.initialization.* block.

A more systematic redesign of the initialisation module is planned for a later release, but the present implementation remains fully usable provided it is approached with this level of clarity. The purpose of this document is precisely to support that use: to make the current behaviour of OSMOSE 4.4 explicit, understandable, and operational for both technical and applied modelling workflows.