Preprints
https://doi.org/10.5194/egusphere-2024-4026
https://doi.org/10.5194/egusphere-2024-4026
04 Feb 2025
 | 04 Feb 2025
Status: this preprint is open for discussion and under review for Biogeosciences (BG).

Optimisation of the World Ocean Model of Biogeochemistry and Trophic-dynamics (WOMBAT) using surrogate machine learning methods

Pearse James Buchanan, P. Jyoteeshkumar Reddy, Richard J. Matear, Matthew A. Chamberlain, Tyler Rohr, Dougal Squire, and Elizabeth H. Shadwick

Abstract. The introduction of new processes in biogeochemical models brings new model parameters that must be set. Optimisation of the model parameters is crucial to ensure model performance based on process representation, rather than poor parameter values. However, for most biogeochemical models, standard optimisation techniques are not viable due to computational cost. Typically, (tens of) thousands of simulations are required to accurately estimate optimal parameter values of complex non-linear models. To overcome this persistent challenge, we apply surrogate machine learning methods to optimise the model parameters of a new version of the World Ocean Model of Biogeochemistry and Trophic dynamics (WOMBAT), which we call WOMBAT-lite. WOMBAT-lite has undergone numerous updates described herein with many new model parameters to prescribe. A computationally inexpensive surrogate machine learning model based on Gaussian Process Regression was trained on a set of 512 simulations with WOMBAT-lite. These simulations explored model fidelity to 8 observation-based target datasets by varying 26 uncertain parameters across their a priori ranges. The surrogate model, trained on these 512 simulations, facilitated a global sensitivity analysis to identify the most important parameters and facilitated Bayesian parameter optimisation. Our approach returned optimal posterior distributions of 13 important parameters that, when input to WOMBAT-lite, ensured excellent fidelity to the target datasets. This process improved the representation of chlorophyll-a concentrations, air-sea carbon dioxide fluxes and patterns of phytoplankton nutrient limitation. We present an optimal parameter set for use by the modelling community. Overall, we show that surrogate-based calibration can deliver optimal parameter values for the biogeochemical components of earth system models and can improve the simulation of key processes in the global carbon cycle.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Share
Pearse James Buchanan, P. Jyoteeshkumar Reddy, Richard J. Matear, Matthew A. Chamberlain, Tyler Rohr, Dougal Squire, and Elizabeth H. Shadwick

Status: open (until 18 Mar 2025)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
Pearse James Buchanan, P. Jyoteeshkumar Reddy, Richard J. Matear, Matthew A. Chamberlain, Tyler Rohr, Dougal Squire, and Elizabeth H. Shadwick

Model code and software

WOMBAT-lite code Pearse J. Buchanan https://github.com/pearseb/WOMBAT_dev

Interactive computing environment

Analysis of sections 3.1 and 3.4 Pearse J. Buchanan https://github.com/pearseb/WOMBAT-lite_optimisation_analysis

Pearse James Buchanan, P. Jyoteeshkumar Reddy, Richard J. Matear, Matthew A. Chamberlain, Tyler Rohr, Dougal Squire, and Elizabeth H. Shadwick

Viewed

Total article views: 31 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
27 2 2 31 1 0 0
  • HTML: 27
  • PDF: 2
  • XML: 2
  • Total: 31
  • Supplement: 1
  • BibTeX: 0
  • EndNote: 0
Views and downloads (calculated since 04 Feb 2025)
Cumulative views and downloads (calculated since 04 Feb 2025)

Viewed (geographical distribution)

Total article views: 56 (including HTML, PDF, and XML) Thereof 56 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 06 Feb 2025
Download
Short summary
We developed a cost-effective method to improve ocean models for studying the global carbon cycle. Using machine learning, we optimized parameters in the WOMBAT-lite model, enhancing its accuracy in predicting chlorophyll levels, air-sea carbon dioxide exchange, and phytoplankton nutrient use. This approach increases model reliability and offers a pathway for scientists to better understand and predict ocean changes, contributing to improved insights into Earth's climate system.
Share