Biological data assimilation for parameter estimation of a phytoplankton functional type model for the western North Pacific

Abstract

Abstract. Ecosystem models are used to understand ecosystem dynamics and ocean biogeochemical cycles and require optimum physiological parameters to best represent biological behaviours. These physiological parameters are often tuned up empirically, while ecosystem models have evolved to increase the number of physiological parameters. We developed a three-dimensional (3-D) lower-trophic-level marine ecosystem model known as the Nitrogen, Silicon and Iron regulated Marine Ecosystem Model (NSI-MEM) and employed biological data assimilation using a micro-genetic algorithm to estimate 23 physiological parameters for two phytoplankton functional types in the western North Pacific. The estimation of the parameters was based on a onedimensional simulation that referenced satellite data for constraining the physiological parameters. The 3-D NSI-MEM optimized by the data assimilation improved the timing of a modelled plankton bloom in the subarctic and subtropical regions compared to the model without data assimilation. Furthermore, the model was able to improve not only surface concentrations of phytoplankton but also their subsurface maximum concentrations. Our results showed that surface data assimilation of physiological parameters from two contrasting observatory stations benefits the representation of vertical plankton distribution in the western North Pacific.

Assessing the value of biogeochemical Argo profiles versus ocean color observations for biogeochemical model optimization in the Gulf of Mexico

“…Finally, about 24 % of primary production is removed from the upper 200 m through gravitational sinking. The simulated export ratio of 24 % is within the wide range of reported export ratios, from 6 % to 43 %, at 120 m depth in the Gulf of Mexico (see Table 3 of Hung et al, 2010). Despite the high degree of uncertainty that exists when estimating export ratios (e.g., the global mean export ratio varies from ∼ 10 % Henson et al, 2012;Lima et al, 2014;Siegel et al, 2014 to ∼ 20 % Henson et al, 2015;Laws et al, 2000), it is obvious that only experiment C reproduced an export ratio of a reasonable magnitude.…”

Assessing the value of BGC Argo profiles versus ocean colour observations for biogeochemical model optimization in the Gulf of Mexico

“…In this study, a 1D model is used to optimize the biological parameters of the 3D model. This approach has been successfully used previously (Hoshiba et al, 2018;Kane et al, 2011;Oschlies and Schartau, 2005).…”

“…On the one hand, the optimized parameters from the 1D model may have been adjusted to compensate for biases in the biological properties caused by neglecting advection and, as a result, this may degrade the 3D simulations (Kane et al, 2011). On the other hand, counter examples exist where the 3D simulations outperform the 1D models (Hoshiba et al, 2018). Secondly, the spatial heterogeneity of parameters (e.g., Kuhn and Fennel 2019) is another issue that influences the portability of parameters from 1D to 3D models.…”

“…Secondly, the spatial heterogeneity of parameters (e.g., Kuhn and Fennel 2019) is another issue that influences the portability of parameters from 1D to 3D models. In some studies, the parameter optimization has been performed at several contrasting stations with the goal of using different parameter sets in different regions of the 3D model (Hoshiba et al, 2018). In other studies different https://doi.org/10.5194/bg-2020-137 Preprint.…”

“…Since the high computational cost and storage demands of 3D models make direct application of most parameter optimization techniques difficult (but see Mattern et al, 2012;Mattern and Edwards, 2017;Tjiputra et al, 2007 for exceptions), they are typically applied in computationally efficient 1D models before using the resulting parameters in 3D version (e.g. Hoshiba et al, 2018;Kane et al, 2011;Kuhn and Fennel, 2019;Schartau and Oschlies, 2003). We follow the latter approach here.…”