Optimality-Based Non-Redfield Plankton-Ecosystem Model (OPEMv1.0) in the UVic-ESCM 2.9. Part II: Sensitivity Analysis and Model Calibration

Abstract

We analyse 400 perturbed-parameter simulations for two configurations of an optimality-based plankton-ecosystem model (OPEM), implemented in the University of Victoria Earth-System Climate Model (UVic-ESCM), using a Latin-Hypercube sampling method for setting up the parameter ensemble. A likelihood-based metric is introduced for model assessment and selection of the model solutions closest to observed distributions of NO 3 -, PO 4 3 -, O 2 , and surface chlorophyll a concentrations. 5According to our metric the optimal model solutions comprise low rates of global N 2 fixation and denitrification. These two rate estimates turned out to be poorly constrained by the data. For identifying the "best" model solutions we therefore also consider the model's ability to represent current estimates of water-column denitrification. We employ our ensemble of model solutions in a sensitivity analysis to gain insights into the importance and role of individual model parameters as well as correlations between various biogeochemical processes and tracers, such as POC export and the NO 3 inventory. Global O 2 varies by 10 a factor of two and NO 3 by more than a factor of six among all simulations. Remineralisation rate is the most important parameter for O 2 , which is also affected by the subsistence N quota of ordinary phytoplankton (Q N 0, phy ) and zooplankton maximum specific ingestion rate. Q N 0, phy is revealed as a major determinant of the oceanic NO 3 pool. This indicates that unraveling the driving forces of variations in phytoplankton physiology and elemental stoichiometry, which are tightly linked via Q N 0, phy , is a prerequisite for understanding the marine nitrogen inventory.The basic structure of most marine ecosystem models has been designed for resolving mass fluxes between nutrients, phy-25 toplankton, zooplankton and detritus, typically referred to as NPZD models. Mathematical formulations that describe growth and fate of marine phytoplankton and zooplankton biomass have been successfully applied over a range of scales, from local 0D-ecosystem models (e.g., Fasham et al., 1990;Edwards, 2001) to global 3D models (Sarmiento et al., 1993;Keller et al., 2012;Nickelsen et al., 2015). However, most of these NPZD models lack a sound mechanistic foundation, preventing them from explicitly accounting for the organisms' regulation of their internal physiological state. For example, N 2 fixation by algae 30 is often diagnosed from the availability of dissolved nutrients, so that it only occurs when the ratio of nitrate-to-phosphate concentrations falls below the Redfield ratio of 16:1 (Deutsch et al., 2007;Ilyina et al., 2013). As these assumptions neglect a number of environmental and ecological controls (e.g., grazing, often also temperature), they do not adequately describe the behaviour of plankton organisms and their sensitivity to changes in their environment. With the introduction of refined mechanistic (physiological) descriptions we here aim at alleviating this deficiency. In this study we introduce a new marin...

Optimality-Based Non-Redfield Plankton-Ecosystem Model (OPEMv1.0) in the UVic-ESCM 2.9. Part I: Implementation and ModelBehaviour

“…Negative concentrations are also generated in the main biogeochemical module of UVic (subroutine npzd_src), owing to the long time-steps (we use 0.5 times the physical time step of 30 h and, if this would generate negative tracer concentrations, subcycle with 0.25 times the physical time step) and the Euler scheme used for calculating the sources-minus-sinks terms. For many cases (parameter settings), phytoplankton and/or diazotrophs can end up negative everywhere, compromising our calibration procedure, which depends on the reliability of simultaneous evaluation of simulation ensembles (see Section 2.4 below and Chien et al, 2019). We have addressed the problem by limiting the biological tracer fluxes of the sub-cycled biological time step at every grid box, so that not more than 90 % of any tracer is removed within any grid box during one time step.…”

“…where v 0 = 6 m d −1 is the sinking velocity at the surface, z is depth and a v = 0.06 d −1 the rate of increase in v sink with depth (Kriest, 2017 (Chien et al, 2019). Symbol descriptions are given in Table 1.…”

“…Then we set up an ensemble of 400 parameter sets and ran both of our model configurations into steady state for all parameter sets. We select two reference simulations, one each from the OPEM and OPEM-H ensembles, according to a cost function and the ability to predict realistic levels of water-column denitrification (Chien et al, 2019). The cost function quantifies the model-data misfit by a measure of the discrepancies between observed and simulated O 2 , NO 3 -, PO 4 3 -, and Chl, considering also correlations and covariances (Chien et al, 2019).…”

“…Here we compare the behaviour of the OPEM with that of a previous UVic configuration, described in Nickelsen et al (2015), modified with several improvements and bug fixes as described below. Since the calibration of the OPEM embedded in UVic presents a major challenge, it is dealt with in the companion paper (Chien et al, 2019).…”

“…Vertically-integrated and temporally-averaged phytoplankton (top) and diazotroph biomass (centre) and difference between diazotroph and phytoplankton net relative growth rates (bottom), in the original, OPEM, and OPEM-H UVic simulations. Note that the positive growth-rate differences for the original UVic in the Arctic are spurious as they result from µdia = 0 d −1 and µphy < 0 d −1 k Fe, phy , whereas k Fe, dia < k Fe, phy in the original UVic), and the model calibration yielded a higher values of the prey-capture coefficients for diazotrophs (Table 2, see also Chien et al, 2019). Both OPEM and OPEM-H have a similar phytoplankton biomass and distribution (Fig.…”