A Mechanistic Dynamic Model of an Aquaponic RAS: Multi-Cycle Fish-Growth Assessment and Sensitivity Analysis

Aquaponic systems couple fish and plant production in recirculating loops, yet quantitatively assessed dynamic models for engineering analysis, scale-up, and operation under realistic conditions remain limited. Here, a modular process-based MATLAB R2026a framework is developed for the recirculating aquaponic system operated at the ASTREDHOR facility (France). The model links hydraulic transport with fish metabolism, nitrification, solids removal, and plant nitrate uptake, using monitoring-derived boundary conditions for temperature, dissolved oxygen, pH, and electrical conductivity. The fish-growth component was calibrated and evaluated against archived, temporally reconstructed biomass trajectories derived from campaign-based biometrics in three production cycles with different fish compositions and environmental regimes. Tank-wise R2 values were 0.865–0.952 in the calibration windows and 0.700–0.921 in the fixed-parameter prediction windows, with prediction-period NRMSE values of 0.64–3.91%. These descriptive metrics quantify agreement on the reconstructed evaluation grid rather than performance over independently retained biometric sampling occasions. Complete corresponding time series were unavailable for TAN, NO2−, NO3−, total suspended solids, and plant uptake; these simulated outputs were therefore used only for mechanistic consistency assessment and exploratory scenario analysis, rather than independent validation. Local sensitivity analysis showed limited effects of temperature sensitivity (αT), optimal temperature (Topt), and minimum dissolved oxygen (DOmin) under observed conditions, whereas the feeding ratio (TR) and metabolic scaling exponent (n) strongly influenced simulated fish growth and nitrogen loading. Parametric sweeps provided preliminary, model-derived indications of feeding and biofilter-sizing limits under intensified loading; these thresholds require confirmation against independent water-quality measurements. The resulting framework is positioned as an off-line digital shadow with a fish-growth component assessed against reconstructed biomass trajectories and exploratory water-quality simulations.