Abstract:

The marine environment, both globally and in Denmark, is severely challenged by expanding hypoxia in coastal waters, with devastating consequences for biodiversity and ecosystem functioning. Proactively addressing this issue requires a data-informed, model-based understanding of its scale and, importantly, cost-effective mitigation strategies.  At the European scale, Denmark presents a particularly complex case, with some of the most challenging marine conditions. Despite extensive monitoring efforts over the past 30 years and a state-of-the-art national biogeochemical model, current approaches are limited by the fine granularity and hydrodynamic complexity of Danish waters, leading to computational constraints. Consequently, the ecosystem models cannot be calibrated for the entire Danish domain, restricting their capacity for future scenario projections. Additionally, the national model currently in use is not open-source, reducing scientific transparency and limiting collaborative research efforts. 
 
This project addresses these challenges by developing an open-source, computationally efficient ocean-biogeochemical model. Leveraging Backward Automatic Differentiation (AD) and GPU acceleration, the model will enable orders-of-magnitude faster parameter calibration, real-time adaptability, and extensive scenario testing. Unlike traditional variational data assimilation methods, AD-based optimization allows for direct gradient-based parameter estimation, improving model efficiency and predictive accuracy. The framework builds on Julia’s Oceananigans.jl, ClimaOcean.jl, and OceanBioMe.jl, ensuring modularity, extensibility, and integration with Denmark’s extensive long-term observational datasets. 
 
Beyond this project, the model aims to foster scientific collaboration, public transparency, and data-driven decision-making. By enabling researchers to integrate novel modules and conduct independent scenario testing, it will serve as a community-driven tool for advancing coastal ocean modelling. Additionally, a user-friendly interface will support engagement with policymakers, municipalities, and the public, ensuring that scientific evidence informs environmental management. 
Ultimately, this research lays the foundation for a transparent, high-performance coastal modelling framework, helping Denmark—and other regions—tackle future climate and ecological challenges with a data-driven approach.