Estimating carbon pools in the shelf sea environment: reanalysis or model-informed machine learning?
This work provides a cost-effective alternative to expensive reanalyses for estimating carbon pools in shelf seas, which is important for understanding carbon sequestration and climate scenarios, though it is incremental as it applies existing machine learning methods to a specific domain.
The authors tackled the problem of estimating carbon pools in shelf seas, where observations are sparse and reanalyses are costly, by using a deep ensemble of neural networks trained on a model simulation to reproduce reanalysis outputs for the North-West European Shelf, achieving accurate predictions with uncertainty information.
Shelf seas are important for carbon sequestration and carbon cycle, but shelf sea observations for carbon pools are often sparse, or highly uncertain. Alternative can be provided by reanalyses, but these are often expensive to run. We propose to use an ensemble of neural networks (i.e. deep ensemble) to learn from a coupled physics-biogeochemistry model the relationship between the directly observable variables and carbon pools. We demonstrate for North-West European Shelf (NWES) sea environment, that when the deep ensemble trained on a model free run simulation is applied to the NWES reanalysis, it is capable to reproduce the reanalysis outputs for carbon pools and additionally provide uncertainty information. We focus on explainability of the results and demonstrate potential use of the deep ensembles for future climate what-if scenarios. We suggest that model-informed machine learning presents a viable alternative to expensive reanalyses and could complement observations, wherever they are missing and/or highly uncertain.