Ecosystem modelling

Introduction

Marine ecosystem modelling involves identifying and understanding the mechanisms that govern the way the marine environment works, and translating them into mathematical expressions. In REMSEM, different types of ecosystem models are developed: 3D coupled hydrodynamic-biogeochemical models with the aim to support eutrophication management, Larval Transport Models (LTMs) which help to assess larval dispersal pattern and connectivity between spawning grounds and nurseries, Individual-Based Models (IBMs), and trait-based models (TBMs).

Ecosystems models deal with physical variables (currents, temperature, salinity), chemical variables (major nutrients), microbiological variables (phytoplankton, mesozooplankton, copepods, recycling bacteria) (Example in Fig. 1), and macrobiological variables (fish or invertabrate’s larvae, jellyfish). All these variables are linked to one another and the degree of complexity and uncertainty increases from the physical to the biological variables. Assumptions must sometimes be made to bypass processes that are not well understood.

Figure 1: Example of model results showing the spatial distribution of surface salinity, Dissolved Inorganic Nitrogen (µmolN L^-1), Phosphate (µmolP L^-1) and chlorophyll a (µg L^-1) the 1st of May 2006 as simulated with 3D-MIRO&CO. (Click to enlarge.)

The theoretical results produced from the models are carefully compared with all the available observations (see the Belgian Marine Data Center (BMDC) and the Remote Sensing section). This is the validation phase (see examples in Fig. 2 and Fig. 3). An unsatisfactory validation is a sign that the mathematical expressions chosen to describe physical, chemical and biological processes are not correct. A satisfactory validation indicates that the models can be used for management purposes and to a certain extent for forecasting.

Modelled chlorophyll a time series at station 330 over the period 1993-2006 — Figure 2: Modelled chlorophyll a time series at station 330 (51°26.00’N, 2°48.50’E) over the period 1993-2006. The dots represent the in situ observations recorded by the BMDC. (Click to enlarge.)

Figure 3: Modelled chlorophyll a (top) and satellite chlorophyll a (bottom). Maps represent the mean in the period 2003-2005 of annual chlorophyll a percentile 90 in the English Channel and the Southern Bight of the North Sea. (Click to enlarge.)

Expertise in ecosystem modelling

Eutrophication management

As a consequence of eutrophication, the spring bloom in the Belgian Continental Shelf (BCS) is dominated almost every year by the nuisance colonial species Phaeocystis. The eutrophication status of the BCS is mainly determined by the nutrient loads from the rivers Scheldt, Seine, Somme and Rhine. According to the OSPAR strategy (2008), the Belgian coastal zone is still characterised as a “problem area”. It has been shown that efforts should be directed to reduction of N-loads (not only P-loads) in order to decrease Phaeocystis blooms without negatively affecting diatoms (Gypens et al. 2007, Lacroix et al. 2007a). The phosphorus reductions of the last decades caused an increase in riverine N:P ratio (Brion et al. 2008), which is favorable to Phaeocystis. The excessive nitrogen concentration allows Phaeocystis colonies to reach large sizes (when abundances become larger than 6 millions cells L^-1), making them inedible to zooplankton. The resulting decrease in zooplankton grazing constitutes a loss for higher trophic levels, and is considered an ecological nuisance (Lancelot et al. 2009). The nuisance of Phaeocystis bloom is usually measured by the intensity of its spring maximum and by the duration of the bloom. It may also be estimated with the annual integral of excessive abundance, IEB (Fig. 4). To study these processes in the BCS, the 3D-MIRO&CO mathematical model was developed in the frame of AMORE project with a full description of hydrodynamical and ecological processes (Lacroix et al. 2007b).

Figure 4: Modeled time series of Phaeocystis colonies abundance in the BCS for an oligotrophic station (dotted line), and a coastal eutrophicated station (black line). The horizontal line indicates the abundance above which colonies become inedible for copepods (6 millions cells L-1). The grey area is a measure of Phaeocystis ecological nuisance, the integral of excessive abundance (IEB).

The spatial distribution of the Phaeocystis spring maximum illustrates the role of the rivers in marine eutrophication, as the maximum values are located close to the estuaries and along the coastal zones (Fig. 5). The spatial distribution of Phaeocystis nuisance also shows that, in areas further from the rivers influence, Phaeocystis may in some locations never reach problematic abundances.

Figure 5: 3D-MIRO&CO model results in the English Channel and the Southern Bight of the North Sea for the year 2000. Left: map of *Phaeocystis* spring maximum abundance (not synoptic). Right: annual map of *Phaeocystis* IEB (see Fig.4). The white area depicts the area where *Phaeocystis* abundance never exceeds 6 millions cells L^-1 during the year. (Click to enlarge.)

Yet, a reduction in riverine nutrient loads may have a significant impact on phytoplankton production in remote marine areas (Lenhart et al. 2010), illustrating the link between continental freshwater management and offshore marine ecology. As Atlantic waters circulate through the Channel in the northward direction, terrigenous materials entering the coastal zones are eventually exported offshore. The 3D-MIRO&CO model allowed to carry on a trans-boundary nutrient transport study (TBNT). The principle of these simulations is to track the nutrients coming from specific sources during several years, and to analyse their remote impact on ecology. That way, it is possible to identify the different contributive sources to eutrophication in a specific area. This greatly helps to understand the system, and to support the decisions at policy and management levels.

Relative contributions of rivers and boundaries to DIN and PO4 — Figure 6: Relative contributions of rivers and boundaries to DIN (top) and PO₄ (bottom) in specific target areas during the year 2002 (BC1 and BO1 are respectively the Belgian coastal and the Belgian offshore zones). Each coloured bar represents the influence of a group of rivers or the influence of one boundary (RBE = Belgian rivers, RNL1 = Dutch rivers, RFR1 = French rivers, RUK1 = UK rivers, WBC = Western Boundary of the model in the English Channel, NBC = Northern Boundary of the model in the Central North Sea). (Click to enlarge.)

The Figure 6 illustrates that the Belgian coastal zone (BC1 on the axes) receives DIN from several contributors in 2002, firstly from the English Channel boundary, secondly from the French rivers, and thirdly from the Belgian and Dutch rivers in almost equal amounts. In the same area, the major contributor to PO₄ is the English Channel in 2002. In the Belgian offshore zone (BO1 on the axes), the French rivers are almost an equal contributor as the English Channel to DIN, while PO4 is almost only imported via the Channel waters. The TBNT study also allowed to estimate the total contribution of rivers to nutrient increase in the sea, by substracting the import of oceanic nutrients at the Southwestern boundary from the export of nutrients at the Northern boundary. For the year 2002, the total nutrient increase in the sea due to rivers is estimated at 76% for DIN and at 54% for PO₄ by comparison to oceanic background, which leads to an increase of 14% of the N:P ratio in the Central North Sea due to rivers.

The 3D-MIRO&CO model is currently being coupled to Riverstrahler (Billen et al., 1994), a river basin model, in the frame of EMoSEM, with the aim to test the impact of realistic nutrient reduction scenarios at the watershed level on eutrophication in the coastal area.

Literature

Billen G, Garnier J, Hanset P, Descy JP, Reynolds C, and Padisak J. Modelling phytoplankton development in whole drainage networks: The riverstrahler model applied to the seine river system. Hydrobiologia 289 (1994), 119-137.

Brion N, Jans S, Chou L and Rousseau V (2008) Nutrient loads to the Belgian coastal zone, in Current Status of Eutrophication in the Belgian Coastal Zone. Rousseau V, Lancelot C, Cox D (eds), Presses Universitaires de Bruxelles, Brussels, pp. 17-43. N° dépôt legal: D/2006/1191/45.

Gypens N, Lacroix G and Lancelot C (2007) Causes of variability in diatom and Phaeocystis blooms in Belgian coastal waters between 1989 and 2003: a model study. Journal of Sea Research 57, 19-35.

Lacroix G, Ruddick K, Gypens N and Lancelot C (2007a) Modelling the relative impact of rivers (Scheldt/Rhine/Seine) and Western Channel waters on the nutrient and diatoms/Phaeocystis distributions in Belgian waters (Southern North Sea). Continental Shelf Research 27, 1422-1446.

Lacroix G, Ruddick K, Park Y, Gypens N and Lancelot C (2007b) Validation of the 3D biogeochemical model Miro&Co with field nutrient and phytoplankton data and Meris-derived surface chlorophyll a images. Journal of Marine Systems 64, 66-88.

Lancelot C, Rousseau V and Gypens N (2009) Ecologicaly based indicators for Phaeocystis disturbance in eutrophied Belgian coastal waters (Southern North Sea) based on field observations and ecological modelling. Journal of Sea Research 61, 44-49.

Lenhart HJ, Mills DK, Baretta-Bekker H, van Leeuwen SM, van der Molen J, Baretta JW, Blaas M, Desmit X, Kühn W, Lacroix G, Los HJ, Ménesguen A, Neves R, Proctor R, Ruardij P, Skogen MD, Vanhoutte-Brunier A, Villars MT and Wakelin SL (2010) Predicting the consequences of nutrient reduction on the eutrophication status of the North Sea. Journal of Marine Systems 81, 148-170.

OSPAR Commission (2008) Second OSPAR integrated report on the eutrophication status of the OSPAR maritime area. EUC 09/2/Info.1-E (English only). ISBN 978-1-906840-13-6.

Larval transport models

Effective management of fish stocks requires understanding of how spawning grounds and nurseries are connected and what processes influence larval retention, dispersal and recruitment. The transport of fish larvae from the spawning grounds to the nurseries is driven by hydrodynamic processes, but the final dispersal pattern and larval survival/abundance might be influenced by both behavioural and environmental factors. Individual-based coupled physical–biological models have become one of the standard tools for studying larval dispersal and recruitment.

REMSEM has developed a larval transport model by coupling a Lagrangian particle-tracking model to the 3D hydrodynamic COHERENS model of the North Sea to calculate the movement of individual particles in space and time. In addition, the development of an Individual-Based Model (IBM) allows to take into account the individual life-history of organisms. The IBM includes for instance larval stage-related duration, behaviour such as vertical migration, mortality... The larval transport model is used to investigate trajectories and dispersal (Fig. 7) of fish or invertebrate’s eggs and larvae and to study the impact of interannual variability and behaviour on dispersal pattern (ex. application to Solea solea in Savina et al., 2010; Lacroix et al., 2013). The model is also used to calculate the connectivity (Fig. 9) between spawning areas and nurseries (Fig. 8) and to quantify larval recruitment at nurseries as a support to fisheries management.

Frequency of connectivity matrix — Fig. 9. The frequency of connectivity matrix represents the occurrence (number of years over the period 1995-2006) of connections or retention (self-recruitment on the diagonal) predicted by the model. Red: never (0/12), orange: sometimes (1-6/12), yellow: often (7-11/12), green: always (12/12).

Results of Solea solea larval transport model — Fig. 7. Results of *Solea solea* larval transport model. Left: Trajectories of the centre of mass (6 spawning grounds) for 12 years (1995-2006). Black line: average. Origin: circle. End of trajectories: squares. Right: Larval abundance at the end of the pelagic phase (average 1995-2006). (Click to enlarge)

Distribution of the main spawning areas of sole / Nursery areas — Fig. 8. Left: Distribution of the main spawning areas of sole (*Solea solea*) in the southern North Sea and the eastern English Channel with contour plots of the mean daily egg production (redrawn from Loes *et al.*, in prep). Abbreviations of the spawning grounds are: EC (Eastern channel), BC (off the Belgian coast), Tx (off Texel), GB (inner German Bight), N (Norfolk banks), Th (Thames estuary). Right: Nursery areas of sole defined as the coastal area with a depth less than 20 m and a substrate of either mud to sand or sand to mud (<5 % gravel). Abbreviations of the nursery grounds are: FR (French nursery), BE (Belgian nursery), NL (Dutch nursery), GE (German nursery), No (Norfolk nursery), Tha (Thames nursery). (Click to enlarge)

Literature

Lacroix G., Maes G.E., Bolle L.J., Volckaert F.A.M. Modelling dispersal dynamics of the early life stages of a marine flatfish (Solea solea L.). Journal of Sea Research 84:13-25.

Luyten, P., Jones, J.E., Proctor, R., Tabor, A., Tett, P., Wild-Allen, K., 1999. COHERENS - A Coupled Hydrodynamical-Ecological model for Regional and Shelf seas: User Documentation. MUMM Report, Management Unit of the Mathematical Models of the north Sea, Belgium. 911 p.

Savina M., Lacroix G., Ruddick K., 2010. Modelling the transport of common sole larvae in the Southern North Sea: influence of hydrodynamics and larval vertical movements. Journal of Marine Systems, 81: 86-98.

Trait-based models

While statistical models establish correlations between the variables, they are not suitable to explain the mechanisms that link the variables together in the natural environment. Trait-Based model are dynamical models with a large number of state variables (i.e. species) linked by variation in physiological traits, and constrained by trade-offs between trait-values and ecosystem response. They are developed to test hypotheses about the mechanims that may explain the emergence of particular species or communities under changing environmental conditions. The spatial distribution of benthic communities in the Belgian Continental Shelf is under study with a trait-based model (currently under development).