Use of sensitivity and comparative analyses in constructing plausible trophic mass-balance models of a data-limited marine ecosystem — The KwaZulu-Natal Bight, South Africa
Research highlights
► Plausible ecosystem models can be constructed using local and non-local data. ► Models showed sensitivity to non-local apex predator data due to top-down balancing. ► Models reproduce known differences in functioning with a Southern Benguela model. ► Models described an ecosystem reliant on import and cycling of riverine detritus.
Introduction
Understanding how marine ecosystems function as a whole is a challenge for data-poor systems. However, a holistic overview of an ecosystem is necessary for understanding how they function and how anthropogenic activities could potentially impact them. Bays and bights are ecosystems often influenced by anthropogenic activities due to their proximity to the coast. Firstly, most are affected by commercial, recreational and/or subsistence fisheries (Jennings and Kaiser, 1998, Pauly et al., 1998, Pauly et al., 2005). Secondly, some are influenced by the outflow of large rivers (Diaz and Rosenberg, 2008, Gillanders and Kingsford, 2002, Lamberth et al., 2009).
River-influenced bays and bights are usually more productive than the adjacent ocean due to an inflow of nutrients and detritus from topographical upwelling and rivers (Wollast, 1998). However, there is an increasing need to place impoundments onto large rivers for inland water-use, and to increase inshore fishery catches for local consumption and export. Therefore it is important to understand how these ecosystems function at present in order to predict what effects these activities will have on production in the system.
The KwaZulu-Natal Bight (KZN Bight) on the east coast of South Africa is an example of a river-influenced bight (Fig. 1). It is also a data-poor ecosystem with sparse quantitative data on biotic and abiotic components. Biomasses are available for plankton groups and other quantitative data are available for important linefishing species, large sharks, and cetaceans only. Oceanic waters off south-east Africa are typically oligotrophic (Lutjeharms, 2006a). However, the waters of the bight are slightly less so than the bordering Agulhas Current (Lutjeharms et al., 2000, Meyer et al., 2002). This is due to nutrient inputs from an episodic upwelling off Richards Bay, a lee eddy off Durban and rivers along the coast, particularly the Thukela (Carter and D'Aubrey, 1988, Lutjeharms et al., 1989, Pearce et al., 1978). The Thukela River is the third largest river in southern Africa and has a sediment output of 6.79 × 103 m3 year− 1 (Birch, 1996). This outflow creates a turbid area in the central bight which is home to South Africa's only prawn fishery (Fennessy and Groeneveld, 1997). The rivers and estuaries along the coast aid the recruitment of many targeted fishery species by providing nursery grounds (Lamberth et al., 2009, Wallace et al., 1984, Wallace and van der Elst, 1975, Whitfield, 1998). To understand and predict the impact of current and future anthropogenic activities (e.g. fishing and water impoundments on the Thukela River (DWAF, 2004)) on this ecosystem as a whole it is important to understand the functioning of this ecosystem as a whole.
Ecosystem modelling allows a system to be studied as a whole. The system can be constructed as a network using information on nodes (biomasses), links between nodes i.e. trophic flows (diet compositions), production, consumption, and fishery landings of biotic groups. If biomass, production or consumption is missing for a group then this can be estimated using the mass-balance approach, most commonly used in Ecopath with Ecosim (EwE) software (Christensen and Pauly, 1992). Thus this approach lends itself to the modelling of ecosystems such as the KZN Bight where many biomasses are unknown. For example, EwE has been used to model coral reef ecosystems and historical representations of ecosystems (Heymans and Pitcher, 2002, Morato and Pitcher, 2005, Polovina, 1984). From this network system metrics can be calculated which describe the system in terms of energy flows, energy cycling and ecosystem services provided by the system (Ulanowicz, 1986). These metrics enable an understanding of how the ecosystem functions and the identification of system level characteristics. In addition, comparisons of marine ecosystems from different areas and of different spatial scales can be carried out. These comparisons can aid the modelling of a data-limited ecosystem if functional differences to another ecosystem are known a priori.
The east coast of South Africa is typically oligotrophic with low fishery catches and is known to be influenced by the many rivers/estuaries flowing into coastal waters. This is in contrast to the west coast of South Africa, comprising the Southern Benguela upwelling ecosystem, which is nutrient rich with large plankton biomass and fishery landings (Shannon et al., 2003). A direct comparison of models of these systems would aid in assessing the plausibility of models of the data-limited KZN Bight. Plausible models of the KZN Bight would produce known differences in functioning to the Southern Benguela.
In this paper, we describe the development and analysis of models of the KZN Bight. The aims of this study are to demonstrate that plausible representations of data-poor river-influenced bights can be constructed, to better understand how the KZN Bight functions by gaining a holistic overview of the system and to build a framework for future models of the KZN Bight. We make comparisons between:
- a)
several versions of KZN Bight models to determine the levels of uncertainty associated with non-local input data,
- b)
the KZN Bight models and a model of the southern Benguela, an upwelling system on the west coast of South Africa, to determine if the KZN Bight models reproduce known differences in functioning between these systems.
Section snippets
Modelling approach
Mass-balanced models of the KZN Bight were constructed and analysed using Ecopath with Ecosim software, version 5.1 (Christensen et al., 2005). Underlying Ecopath with Ecosim are two equations ensuring the mass-balance or energy-balance of each biotic group. The first describes the production of a group:
P/Bi is the production/biomass ratio of group i; Bi is the biomass of group i; Q/Bj is the consumption/biomass ratio of predator j; DCij is the proportion
Parameterization
Ecopath routines allowed missing parameters, biomasses and EEs, to be estimated. These were similar, if not the same, between models for all groups except macrobenthos which ranged between 25.6 and 38.94 t km− 2 (Fig. 2). The system was clearly dominated by macrobenthos in terms of biomass.
Sensitivity of models
KZN Bight models were most sensitive to changes in the input parameters of apex and benthic-feeding chondrichthyans. Small changes (±10%) of these parameters caused the largest number of missing parameters
KZN Bight models
This was the first set of ecosystem models of the entire KwaZulu-Natal Bight on the South African east coast. Due to the scarcity of local data, non-local data were used for many functional groups. In cases of high uncertainty or several poorly defined functional groups, the development of multiple models spanning the range of potential system states is important (Essington, 2007, Fulton et al., 2003). It should be noted however that due to limited data this set of models may not necessarily
Conclusions
Whilst the models may not have predicted ‘accurate numbers’, they have identified data gaps in the literature for the KZN Bight and given an overview of its functioning. Expanding on the steps for ecological network construction in Fath et al. (2007), the following summarises steps taken in this study to construct the KZN Bight models which can serve as recommendations on how models for data-limited ecosystems may best be achieved:
- 1.
Define model domain (spatial boundary and time period).
- 2.
Define
Acknowledgements
We are thankful to the African Coelacanth Ecosystem Project (ACEP II) and the River Influenced Bays and Bights Project (RIBBS), CSIR, for financial assistance; Sean Fennessy from the Oceanographic Research Institute (ORI), Durban for advice; Bruce Mann and Jade Maggs (ORI) for facilitating access to recreational NMLS data; Geremy Cliff and Sheldon Dudley at the Natal Sharks board for shark net landings and advice; and finally the librarians at ORI for their assistance.
References (121)
- et al.
Interactions between a natural food web, shellfish farming and exotic species: the case of the Bay of Mont Saint Michel (France)
Estuarine, Coastal and Shelf Science
(2008) - et al.
Phytoplankton pigments, functional types, and absorption properties in the Delagoa and Natal Bights of the Agulhas ecosystem
Estuarine, Coastal and Shelf Science
(2008) - et al.
Phytoplankton production and physiological adaptation on the southeastern shelf of the Agulhas ecosystem
Continental Shelf Resarch
(2010) - et al.
Selective harvesting by small-scale fisheries: ecosystem analysis of San Miguel Bay, Phillipines
Fisheries Research
(2001) - et al.
Age and growth of the queen mackerel Scomberomorus plurilineatus from KwaZulu-Natal, South Africa
Fisheries Research
(1999) - et al.
Ecopath with Ecosim: methods, capabilities and limitations
Ecological Modelling
(2004) - et al.
Food and feeding of the Indian Ocean bottlenose dolphin off southern Natal, South Africa
- et al.
Ecological network analysis: network construction
Ecological Modelling
(2007) Measures of ecosystem structure and function derived from analysis of flows
Journal of Theoretical Biology
(1976)- et al.
Contribution of ecosystem analysis to investigating the effects of changes in fishing strategies in the South Brazil Bight coastal ecosystem
Ecological Modelling
(2004)
Spatial dynamics of coccolithophore communities during an upwelling event in the Southern Benguela system
Continental Shelf Research
Mortality and biological reference points for the king mackerel (Scomberomorus commerson) fishery off Natal, South Africa (based on a per-recruit assessment)
Fisheries Research
Network analysis of the northern Benguela ecosystem by means of NETWRK and ECOPATH
Ecological Modelling
The effects of altered freshwater inflows on catch rates of non-estuarine-dependent fish in a multispecies nearshore linefishery
Estuarine, Coastal and Shelf Science
The hydrography and water masses of the Natal Bight, South Africa
Continental Shelf Research
The nutrient characteristics of the Natal Bight, South Africa
Journal of Marine Systems
Age and growth of Rhabdosargus sarba (Pisces: Sparidae), from KwaZulu-Natal, South Africa
Fisheries Research
Trophic flows in the southern Benguela during the 1980s and 1990s
Journal of Marine Systems
Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 9. The spinner shark Carcharhinus brevipinna (Müller and Henle)
South African Journal of Marine Science
Occurrence and population structure of pilchard Sardinops ocellatus, round herring Etremus whiteheadi, and anchovy Engraulis capensis off the east coast of southern Africa
South African Journal of Marine Science
Seasonal occurrence of the pilchard Sardinops ocellata on the east coast of South Africa
Aspects of the biology of the galjoen Coracinus capensis (Cuvier) off the South-Western Cape, South Africa
South African Journal of marine Science
Quaternary Sedimentation off the East Coast of Southern Africa (Cape Padrone to Cape Vidal), Bull.118
A preliminary investigation of age and growth of Otolithes ruber from KwaZulu-Natal, South Africa
Western Indian Ocean Journal of Marine Science
Estimates of biomass, consumption and production of Octopus vulgaris Cuvier off the east coast of South Africa
An evaluation of primary productivity studies in the continental shelf region of the Agulhas Current near Durban (1961–1966)
An ecosystem model of San Pedro Bay, Leyte, Phillipines: initial parameter estimates
Factors affecting the development and distribution of marine plankton in the vicinity of Richards Bay
Inorganic nutrients in Natal continental shelf waters
ECOPATH II — A software for balancing steady-state ecosystem models and calculating network characteristics
Ecological Modelling
Ecopath with Ecosim: a User's Guide
Ecopath with Ecosim version 6
Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 8. The great hammerhead shark Sphyrna mokarran (Ruppell)
South African Journal of Marine Science
Sharks caught in the protective gill nets off Natal, South Africa. 4. The Bull shark Carcharhinus leucas Valenciennes
South African Journal of Marine Science
Sharks caught in the protective gill nets off Natal, South Africa. 5. The Java shark Carcharhinus amboinensis (Müller & Henle)
South African Journal of Marine Science
Sharks caught in the protective gill nets off Natal, South Africa. 2. The great white shark Carcharadon carcharias (Linnaeus)
South African Journal of Marine Science
Sharks caught in the protective gill nets off Natal, South Africa. 3. The shortfin mako shark Isurus oxyrinchus (Rafinesque)
South African Journal of Marine Science
Seasonal distribution and density of common dolphins Delphinus delphis off the south-east coast of southern Africa
South African Journal of Marine Science
Age, growth and food of Cheimerius nufar (Ehrenberg, 1820) (Sparidae), collected off St Croix Island, Algoa Bay
South African Journal of Zoology
Preliminary study of the male reproductive cycle in common dolphins, Delphinus delphis, in the eastern North Atlantic
Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 11. The scalloped hammerhead shark Sphyrna lewini (Griffin and Smith)
African Journal of Marine Science
Spreading dead zones and consequences for marine ecosystems
Science
Sharks caught in the protective gill nets off Natal, South Africa. 7. The blacktip shark Carcharhinus limbatus (Valenciennes)
South African Journal of Marine Science
Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 10. The dusky shark Charcharhinus obscurus (Lesueur 1818)
African Journal of Marine Science
Thukela Water Project Decision Support Phase
Evaluating the sensitivity of a trophic mass-balance model (Ecopath) to imprecise data inputs
Canadian Journal of Fisheries and Aquatic Science
The biology of Parablennius cornutus (L.) and Scartella emarginata (Gunther) (Teleostei: Blenniidae) on a Natal Reef
Cited by (20)
Network construction, evaluation and documentation: A guideline
2021, Environmental Modelling and SoftwareCitation Excerpt :It is difficult to identify a single common documentation format because networks may have very different structures, data sources, and purposes. However, this does not preclude the documentation of network construction and evaluation to be presented in detail and to standardise documentation where possible (e.g. Ayers and Scharler, 2011; Bonet et al., 2014; Grüss et al., 2017; Gurney et al., 2014; Hoch et al., 1998; Schmolke et al., 2010). Articles that present new network models must include a full description of the network construction to enable peer review and evaluation.
Bibliometric review of ecological network analysis: 2010–2016
2018, Ecological ModellingCitation Excerpt :Despite the potential influence of aggregation issues, Fath et al. (2007) argued that for ENA applications to be most useful as a systems analysis tool, it is essential to include all components of the ecosystem in the model – even if this means creating aggregated functional groups. A second form of sensitivity and uncertainty analysis considers the network structure largely fixed and focuses on the uncertainty in the flux magnitude estimations (e.g., model parameterization) (Ayers and Scharler, 2011; Guesnet et al., 2015; Hines et al., 2018, 2015; Kones et al., 2009). This is largely accomplished by selected flux perturbations to an initial model using a Monte Carlo approach (Bodini et al., 2012; Heymans et al., 2016; Salas and Borrett, 2011) or using Monte Carlo methods coupled to a modeling procedures such as a regionalized sensitivity analysis (Borrett and Osidele, 2007) or linear inverse modeling (Chaalali et al., 2016; Guesnet et al., 2015; Kones et al., 2009; Pacella et al., 2013) to sample the space of plausible network model parameterizations.
Uncertainty analyses for Ecological Network Analysis enable stronger inferences
2018, Environmental Modelling and SoftwareTowards a sounder interpretation of entropy-based indicators in ecological network analysis
2017, Ecological IndicatorsCitation Excerpt :A further effort is therefore needed in order to understand their behaviour, possibly through uncertainty and sensitivity analysis. Despite the lack of systematic investigations, uncertainty and sensitivity analyses have previously been performed on indicators based on network properties (e.g. Abarca-Arenas and Ulanowicz 2002, Allesina et al., 2005; Dame and Christian 2006; Borrett and Osidele 2007; Ayers and Scharler, 2011; Hines et al., 2015), highlighting its importance to facilitate a better interpretation of ecosystem state. Nowadays, various tools are available to explore the uncertainty, including the increasingly popular Linear Inverse Modelling (LIM, van Oevelen et al., 2010), or ENAtool (Guesnet et al., 2015) for such an approach.
An Ecosystem Approach for understanding status and changes of Nador lagoon (Morocco): Application for of food web models and ecosystem indices
2016, Estuarine, Coastal and Shelf Science