Use of genetic algorithms for ocean model parameter optimisation: a case study using PISCES-v2_RC for North Atlantic particulate organic carbon

Abstract

Abstract. When working with Earth system models, a considerable challenge that arises is the need to establish the set of parameter values that ensure the optimal model performance in terms of how they reflect real-world observed data. Given that each additional parameter under investigation increases the dimensional space of the problem by one, simple brute-force sensitivity tests can quickly become too computationally strenuous. In addition, the complexity of the model and interactions between parameters mean that testing them on an individual basis has the potential to miss key information. In this work, we address these challenges by developing a biased random key genetic algorithm (BRKGA) able to estimate model parameters. This method is tested using the one-dimensional configuration of PISCES-v2_RC, the biogeochemical component of NEMO4 v4.0.1 (Nucleus for European Modelling of the Ocean version 4), a global ocean model. A test case of particulate organic carbon (POC) in the North Atlantic down to 1000 m depth is examined, using observed data obtained from autonomous biogeochemical Argo floats. In this case, two sets of tests are run, namely one where each of the model outputs are compared to the model outputs with default settings and another where they are compared with three sets of observed data from their respective regions, which is followed by a cross-reference of the results. The results of these analyses provide evidence that this approach is robust and consistent and also that it provides an indication of the sensitivity of parameters on variables of interest. Given the deviation in the optimal set of parameters from the default, further analyses using observed data in other locations are recommended to establish the validity of the results obtained.

Influence of GEOTRACES data distribution and misfit function choice on objective parameter retrieval in a marine zinc cycle model

“…This finding supports previous studies suggesting that identification of optimised parameter values at prescribed boundaries, as well as the occurrence of lower misfits outside the prescribed and supposedly realistic parameter space, may point to deficiencies in biogeochemical model structure, a wrong choice of parameters to be optimised, or bias in the physical circulation (e.g. Kriest et al, 2017;Falls et al, 2022) and highlights the importance of well-considered boundaries for interpretability of results.…”

“…Because of this expected concentration dependence of the residuals, several studies suggest that log transformation might be appropriate if there is such a wide variability in concentrations (e.g. Stow et al, 2009;Seegers et al, 2018;Falls et al, 2022); instead, VolRMSE exacerbates this concentration dependence for elements with a nutrient-like distribution. When there are differences in the underlying physical model (synObs_seas and synObs_circ), VolRMSE optimisation consistently leads to the smallest globally integrated Zn export flux (Table S2).…”

Recommendations for the formulation of grazing in marine biogeochemical and ecosystem models

“…In other algorithms, movement across parameter space is more stochastic, mimicking the evolutionary process by selecting for optimal genomes (i.e. parameter sets) from a population of initial estimates and passing on their parameters (sometimes with mutations) to future 'generations' (Falls et al, 2022). Either way, if search schemes are 'pointed' in the wrong direction, say by a partial with a large magnitude or a mutation with strong fitness, then they may take much longer to compute or, worse, never converge on the 'true' optimal solution.…”

Chromophoric dissolved organic matter dynamics revealed through the optimization of an optical-biogeochemical model in the NW Mediterranean Sea

“…A major challenge when working with complex coupled hydrodynamic-BGC numerical models is the need to make an appropriate choice of model parameter values that ensure optimal performance in terms of reproducing observational data. Genetic algorithms are commonly used by BGC modellers to solve 125 nonlinear optimization problems such as parameter estimation (e.g., Kriest et al, 2017;Kuhn and Fennel, 2019;Falls et al, 2022;Wang et al, 2020). In this work, we calibrated the model by using the Parallel Sensitivity Analysis and Calibration utility (ParSAC, Bruggeman and Bolding, 2020), which implements the Differential Evolution (DE) algorithm (Storn and Price, 1997)…”