BioEcoOcean Workshop on Model Validation Protocols
We would like to share that the video recording of the Workshop on Model Validation Protocols held on 26 November 2025 is available on our website! The BioEcoOcean workshop brought together international teams working on marine ecosystem modelling, from lower trophic levels to top predators, carbon-export processes, species distribution models, or integration of satellite and in situ data. The goal was to exchange on current validation practices, identify common data gaps, and explore how to build shared validation standards across models and trophic levels.
Diversity of models and approaches
Talks covered:
- Lower-trophic modelling (SEAPODYM-LMTL, NUM–SISSOMA), simulating zooplankton, micronekton and their role in carbon export.
- Higher-trophic modelling (SEAPODYM HTL, STAMM), linking prey fields to fish, turtles and tuna dynamics.
- Global size-structured modelling (FEISTY), using only two calibration parameters to maintain global consistency.
- Species distribution modelling (MPA Europe), with machine-learning models and multi-level validation.
- Autonomous observations (BGC-Argo), offering depth-resolved biogeochemical data to improve NPP estimation and ecosystem model validation.
Common challenges and validation needs
Despite their diversity, the models share several major validation challenges:
1. Insufficient, uneven, or biased observational datasets
- Zooplankton & micronekton (net tows): scarce time series, poor temporal resolution, lack of day/night sampling, incomplete depth information.
- Higher trophic levels: strong observational biases (strandings, bycatch), confidential data, poor spatial resolution (either spatially disaggregated data or data covering continental shelves)
- Particulate organic matter (POM): poorly constrained POM processes (fragmentation, solubilization).
2. Lack of standardization in data formats and validation metrics
- Participants highlighted heterogeneous input formats, inconsistent units, and diverse metrics (RMSE, NRMSE, correlations, CBI, residual analysis).
- A need for shared validation protocols and minimal common standards was strongly emphasized.
3. Dependence on proxies rather than direct measurements
Many models rely on:
- climatologies (forcing fields, aggregated observational data due to poor spatio-temporal resolution),
- prey proxies for predators (zooplankton fields not mature enough),
- satellite-derived NPP despite known surface biases,
- fisheries data with confidentiality or quality issues.
This increases uncertainty and reveals the need for more direct, vertically resolved, and less biased observations.
4. Increasing importance of autonomous platforms
BGC-Argo floats emerged as a key future tool for:
- resolving vertical structure of phytoplankton and NPP,
- improving export-flux estimation,
- providing gridded 3D/4D products for model evaluation
New sensors (UVP, hyperspectral, acoustic micro-sonar) are expected to greatly enhance validation capacity.
Key outcomes of the workshop
- Strong agreement on developing a shared BioEcoOcean validation protocol, including common metrics, datasets, and EOV requirements.
- Clear prioritization of data gaps, summarized in a “Christmas wish list”: day/night and depth-resolved zooplankton/micronekton data monthly time series, gelatinous zooplankton size-structured fields, NPP corrected with in-situ observations, functional-group or species biomass datasets over continental slope, improved POM‐process measurements, etc.
- Recognition of the need for closer collaboration between modellers and observational teams, to steer future sampling strategies and database development.
- Anticipated major role of BGC-Argo-based 3D gridded products in future ecosystem model validation.
The workshop was organized as part of the Pelagic Ocean Living Lab 1 activities by Sarah Albernhe and Patrick Lehodey from Mercator Ocean International.
