Logo image
From manual to semi-automated calibration: A practical framework for calibrating Atlantis ecosystem models
   

From manual to semi-automated calibration: A practical framework for calibrating Atlantis ecosystem models

Alaia Morell, Isaac C. Kaplan, Hem Nalini Morzaria-Luna, Alberto Rovellini, Chris J. Harvey, Javier Porobic, Elizabeth A. Fulton Ricardo Oliveros-Ramos
Ecological Modelling, Vol.520
2026
Marine ecosystem models, Complex model optimization, EBFM, End-to-end models, Inverse parameter estimation, Multi-objective evolutionary algorithm
Marine ecosystem models have expanded rapidly over the past decade, as their scope increasingly aligns with system-scale questions emerging in management decision-making. Nevertheless, they remain challenging to implement, and the calibration of complex marine ecosystem models continues to be a major bottleneck in model development, particularly for end-to-end frameworks such as the Atlantis modelling platform. Here, we present a semi-automated calibration workflow that integrates the R package calibrar with Atlantis. The automated component of the approach uses a multi-objective evolutionary algorithm to optimize a likelihood-based objective function, quantifying the goodness of fit between model outputs and multiple data types (biomass, weight-at-age, and catch). Our method enables simultaneous calibration across functional groups and indicators, improving overall model fit and reducing calibration time by a factor of two to four compared to manual approaches. This work formalizes Atlantis calibration as a reproducible and portable optimization problem, moving beyond traditional manual heuristics while acknowledging the components of calibration that still require manual calibration. The approach is readily transferable to other ecosystem modelling platforms of comparable complexity. By reducing the effort and subjectivity of the calibration process, our workflow supports broader adoption of complex models for strategic ecosystem-based management applications.

(2)

url
Find in HAL
url
https://doi.org/10.1016/j.ecolmodel.2026.111688
Published (Version of record)
Logo image