Initialisation Introduction

Initialisation is a fundamental part of any OSMOSE simulation. Before the model can simulate growth, mortality, reproduction, movement, predation, and fishing, it must be provided with a starting ecological state. In a school-based, stochastic, multispecies model such as OSMOSE, that starting state is not a trivial quantity. It is not enough to specify only a total biomass by species. The model requires a representation of how populations are distributed across demographic structure and, depending on the method, across schools and space as well.

The purpose of this document is to provide a technical description of the initialisation module in OSMOSE 4.4 and, at the same time, practical documentation for users who need to configure it. The initialisation module has evolved over time, and the current implementation combines methods developed at different stages of the model history. Some are legacy routes retained for backward compatibility or for specific use cases. Others, especially reconstructed population-level initialisation, now provide a more practical and coherent default workflow for most applications.

This document therefore has four main objectives.

First, it explains the conceptual role of initial conditions in OSMOSE and why they matter for the behaviour of the model.

Second, it describes the initialisation methods currently implemented in OSMOSE 4.4, including their conceptual logic, their main differences, and their practical role in the modelling workflow.

Third, it documents the parameters used by these methods, distinguishing carefully between the different layers of configuration involved in the current implementation.

Fourth, it provides practical guidance to users on how to choose an initialisation route, how to set the relevant parameters correctly, and what checks should be performed before accepting an initial state for simulation.

A central theme of the document is the distinction between different representations of the initial state. OSMOSE 4.4 supports direct school-level initialisation, in which the starting schools are provided explicitly through a restart-like file; reconstructed population-level initialisation, in which a class-based population state is reconstructed and then translated into schools by the OSMOSE core simulator; and egg seeding initialisation, in which early population establishment can emerge indirectly through seeded reproduction. These are not merely different file formats. They correspond to different practical and conceptual ways of specifying the starting state of the model.

Among these routes, reconstructed population-level initialisation is the most suitable default workflow for many ordinary applications in OSMOSE 4.4. It uses standard species parameterisation together with biomass and fishery information to generate a coherent population-level starting state, without requiring the user to build a full school-level initial-condition file. For this reason, the document gives particular attention to this method while still documenting the other implemented routes in full.

At the same time, the purpose of this document is descriptive rather than prescriptive in a narrow sense. It aims to document the OSMOSE 4.4 implementation as it currently exists. This is especially important because the initialisation module has accumulated extensions over time, particularly to broaden the applicability of reconstructed population-level initialisation to more data-poor situations. Some aspects of the present configuration structure are therefore historical and may not reflect the cleanest possible design.

For that reason, this document should also be read as documentation of the current state of the codebase prior to future redesign. A more systematic recoding of the initialisation module is planned for a later release, but the need to document the current implementation remains essential. Users of OSMOSE 4.4 require a clear account of what the available methods are, how they operate, and how their parameters should be interpreted in practice.

The remainder of the document is organised as follows. Section 2 introduces the conceptual background of initialisation in OSMOSE. Section 3 provides an overview of the methods currently implemented. Section 4 describes reconstructed population-level initialisation in detail, from its data layer and internal reconstruction procedure to the generation of the population.initialization.* block and its use by the OSMOSE core simulator. Section 5 documents the other implemented methods, namely egg seeding initialisation, direct school-level initialisation, and restart-derived generation of population-level initialisation inputs. Section 6 provides user guidance and parameter documentation. Section 7 discusses limitations of the OSMOSE 4.4 implementation and the rationale for future recoding.