Journal Pre-proof Modelling the mediterranean pelagic ecosystem using the POSEIDON ecological model. Part II: Biological dynamics

A three-dimensional coupled hydrodynamic/biogeochemical numerical model, currently in 5 operation as part of the POSEIDON forecasting system, was implemented in the Mediterranean 6 Sea. The model was assessed regarding its efficacy in representing the main biogeochemical 7 components and seasonal dynamics of the Mediterranean planktonic system. Model outputs were 8 compared with available historical data in a proposed objective eco-regionalization of the 9 Mediterranean basin, based on productivity. The simulated production (primary, bacterial) and 10 planktonic biomass (phytoplankton, zooplankton and bacteria), as well as particulate and 11 dissolved organic carbon, were found consistent with observational estimates in different areas 12 of the Mediterranean Sea, following large scale productivity gradients. Although some 13 limitations of the model were identified, the observed variability of the phytoplanktonic 14 community structure was reasonably well-reproduced, simulating a dominant microbial food 15 web with an intermittent development of the classical food web. The generic POSEIDON 16 ecosystem model skill assessment provides a benchmark for future model improvement, 17 highlighting the need to expand the range of biological variables sampled for further model 18 calibration and validation.

carbon fluxes by merging information from model simulations and ocean colour observations. 83 However, this study was focused on the time-averaged, surface estimates of phytoplanktonic 84 groups and, therefore, could not represent the seasonal variability of the planktonic community    As mentioned above, the most productive regions of the Mediterranean pelagic system were 155 identified in areas characterized by lateral nutrient inputs or strong vertical mixing (Table S1 in   156 Supplementary Material, eco-regions #1, #2) in contrast to the oligotrophic nutrient-depleted 157 eco-regions #3 and #4.

158
Model results are in good agreement with field measurements of the surface NPP along the 159 west-east sampling transect of the TransMediterranean Cruise (Table S2 in (Table 1). In particular, Moutin and Raimbault (2002) 163 reported higher mean integrated in-situ NPP measurements compared to the model output (Table   164 1) in the north-western basin (eco-region #2 probably be attributed to some specific event, triggering an increased primary production. We 175 should note that productivity in this sampling area is significantly influenced by the BSW inflow 176 , which presents an important inter-annual variability. However, this is not 177 taken into account in the model, as BSW nutrient inputs are based on mean climatologic data.

178
In order to give a synthetic view of the Mediterranean phytoplanktonic community structure, 179 the contribution of each phytoplanktonic group to the total phytoplanktonic biomass, integrated 180 to 100m, is shown in Figure 5. According to the model results, and in agreement with previous  In Table 2, the model-simulated phytoplanktonic community structure is compared against    The comparison of observed and model-simulated DOC concentrations (Table 5)   As mentioned above, the horizontal variability of the contribution of different zooplanktonic 299 groups in the total zooplanktonic biomass is determined by the variability of their preys (see prey 300 preferences in Table S4, Supplementary Material of Part I companion paper). Model-simulated 301 HNAN was the dominant zooplanktonic group in the MS (~52 %, Figure 7a) with a relatively 302 higher contribution in the oligotrophic eco-regions #3 and #4 (Table S1) (Table S1). In particular, 306 mesozooplankton's higher contribution was simulated in areas with a considerable abundance of 307 diatoms and dinoflagellates -their main preys -i.e. eco-regions #1 and #2, showing an increasing 308 gradient toward coastal areas that receive lateral nutrient inputs. Microzooplankton presented a 309 more uniform distribution, with a slightly higher abundance in eco-region #1, due to the wide 310 distribution of its preferable preys, HNAN and nanophytoplankton.

311
Model-simulated mesozooplankton biomass presents a rather good agreement with the field 312 measurements, as shown in Table 6, except for some deviations found in the Alboran Sea. In this 313 area, the model simulated a significantly lower mesozooplankton biomass, as compared to the 314 observed one in winter 1997 (eco-region #1, Youssara and Gaudy (2001), Table 6), whilst a 315 model overestimation was found, as compared to the SESAME EU project field data (Mazzocchi Seas. This is related to the increased hydrodynamic variability, influencing nutrient availability 320 and thus mesozooplankton's prey (i.e. phytoplankton) availability in these areas (Siokou-  (Table 6), unlike the measured values during the SESAME field experiment in spring and 324 autumn of 2008 (Table S4). This model deviation is related to the variability of the river loads  surface-to-volume ratio (Legendre and Rassoulzadegan, 1995). This is reflected by their 454 increased growth rates in the model (picophytoplankton: 3.3 day -1 , nanophytoplankton: 2.9 day -1 , 455 diatoms: 2.5 day -1 , dinoflagellates: 1.5 day -1 in model). As mentioned above, another important 456 factor that contributes to the size-based spatial variability in the phytoplanktonic community 457 structure is grazing control (e.g. Thingstad, 1998), which is reflected by the size-based decrease 458 of zooplankton maximum growth rate (HNAN: 4 day -1 , microzooplankton: 1.2 day -1 , 459 mesozooplankton: 0.7 day -1 ; see Table S4 in Supplementary Material of Part I companion paper).

460
The top-down control is thus stronger for picophytoplankton, serving as prey to the faster   food web, since mixotrophs can profoundly influence the cycling of carbon and trophic dynamics 546 among bacteria, primary and secondary producers, both in oligotrophic waters (Stoecker,et al,547 2017) and waters characterized by harmful algal bloom events (Mitra and Flynn, 2010). 548 However, limited information is available on the mixotrophs ecophysiology, particularly those