Gauging the biochemistry of the Ocean is crucial to understanding how climate change affects the health of these precious resources. However, as the numerical models used for the task become more complex, evaluating their accuracy becomes increasingly challenging. A fruitful collaboration between Mercator Ocean International (MOi), the Laboratoire d’Océanographie de Villefranche sur Mer and the National Institute of Oceanography and Applied Geophysics (OGS) of Trieste led to a scientific paper describing a new method based on the use of machine learning and of Biogeochemical-Argo (BGC-Argo) floats to assess the performance of biogeochemical models.
The scientific paper, published by the European Geosciences Union, is co-authored by our experts oceanographers Alexandre Mignot, Elodie Gutknecht, Julien Lamouroux, and Coralie Perruche.
Biogeochemical models: how accurate are they?
A performing biogeochemical model can allow a better ocean health and biodiversity monitoring. This includes, for instance, better predicting extreme changes in Ocean health and eventually support the production of alert bulletins, useful for decision makers. Assessing biogeochemical models is challenging because we don’t always have enough data to do a thorough evaluation. Data sources such as satellite observations or data collected during research cruises for instance, may not have sufficient temporal and spatial detail, and they may not provide all the information needed to evaluate how well the models represent important climate-relevant processes. This is where the scientific paper intervenes, offering an innovative approach to tackle this challenge.
Biogeochemical Argo floats: a game changer
The Biogeochemical-Argo programme provides a solution by developing a global network of biogeochemical sensors on Argo profiling floats, able to retrieve in situ data. These provide greater accuracy and reliability in data, all year round and anywhere the BCG floats are deployed, therefore pushing the limits of quality data acquisition. The availability of this new type of data is particularly valuable for model evaluation. In fact, the BGC-Argo dataset contains more observations than any other dataset available.
On the other hand, BGC models generate vast amounts of data that can be challenging to interpret. To address this, scientists used a machine-learning algorithm to group similar areas in the global ocean based on biogeochemical variables, shown in the picture below.
Recommendation for more observations in the Arctic region
The paper demonstrates that the model is less accurate in the Arctic and Antarctic zones compared to other regions. It denotes the importance of keeping high levels of in situ observations on the Arctic region, while still upholding the already established robust density of BGC-Argo observations in the Southern Ocean (also known as the Antarctic area).
The Arctic region is an area that could benefit from increased BGC-Argo observations. Deploying more BGC floats in this region would help the scientists to enhance the accuracy of the model. The Polar Regions are zones of the globe with a high cloud cover most of the year, which makes it more difficult to retrieve data from satellites. Consequently, the use of BGC-Argo measurements would ensure extra accuracy.
In summary, to refine biogeochemical models, MOi scientists use satellite observations of chlorophyll-a. However, cloud cover prevents some satellite measurements of chlorophyll-a in the Arctic and Antarctic regions. This limitation means that, because of the prevalent cloud cover in these high latitude regions for a considerable part of the year, satellite data cannot substantially be used to improve the precision of BGC models in these zones. This indicates the necessity for other data sources to better regulate such models. Consequently, incorporating BGC-Argo measurements would help scientists in refining the model’s accuracy.
Useful resources
- Take a deep dive into the scientific paper Using machine learning and Biogeochemical-Argo (BGC-Argo) floats to assess biogeochemical models and optimize observing system design: https://bg.copernicus.org/articles/20/1405/2023/
- Learn more about the BGC-Argo program website: https://biogeochemical-argo.org
- Check where the BGC-Argo floats are located in real time: https://biogeochemical-argo.org/float-map-network-status-maps.php
About the authors
Mercator Ocean International’s experts and oceanographers Alexandre Mignot, Elodie Gutknecht, Julien Lamouroux, and Coralie Perruche wrote the article with the participation of Hervé Claustre, Gianpiero Cossarini, Fabrizio D’Ortenzio, Paolo Lazzari, Stefano Salon, Raphaëlle Sauzède, Vincent Taillandier, and Anna Teruzzi.