Model development to predict environmental concentrations of chemical substances in marine sediment when the substance is applied via feed in marine aquaculture: Task 1.4. Model setup and analysis

Abstract The European Food Safety Authority (EFSA) assesses the safety of fish feed additives used in aquaculture. To evaluate their potential environmental risks, it is essential to calculate the predicted environmental concentrations (PEC) and compare them to known exposure thresholds. This report outlines the setup and analysis of a model established in earlier phases of the project to get it functioning with default data and parameter sets, and to explore its sensitivity and applicability. The model is configured using environmental input data (derived from large‐scale ocean and atmosphere models across Europe) and applies standardised parameter sets representing the most common finfish aquaculture species (salmon, sea bass, and sea bream). It is implemented within a containerised system, ensuring portability and ease of deployment across computing environments. The automated modelling pipeline is broadly applicable, although performance can be challenged in areas with complex bathymetry; in these cases, targeted adjustments to default settings and workflow configurations have been explored. Model results indicate that predicted environmental concentrations (PECs) are sensitive to specific parameters, particularly those governing sediment consolidation and additive mixing depth. These parameters should therefore be prioritised in future calibration and validation efforts. The approach adopts high‐resolution, process‐based modelling across a range of representative environmental conditions to derive PEC estimates. Compared with more simplified (idealised) modelling approaches, this framework better reflects real‐world conditions, enables direct comparison with site‐specific observations, and produces outputs that are more readily interpretable for end users and decision‐makers. However, these advantages come with practical challenges, including the need to ensure robust automation across diverse environmental settings and the potential for bias in the selection of representative sites.