MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol 512 141ndash154 2014doi 103354meps10885

Published October 9

INTRODUCTION

R L Lindemanrsquos (1942) scientific paper entitledlsquoThe Trophic-Dynamic Aspect of Ecologyrsquo provided aframework for future research on energy flow andnutrient budgets in ecosystems and set the foun -dations for the development of systems ecologyLinde man established a theoretical model for nutri-ent cycling expressed explicitly in terms of energyflow described by mathematical equations anddeveloped the concept of the trophic pyramid His

adviser G E Hutchinson acknowledged Linde-manrsquos seminal work in an addendum to Lindeman(1942) writing that lsquothe final statement of the struc-ture of a biocoenosis consists of a pair of numbersone an integer determining the level one a fractiondetermining the efficiency [which] may even givesome hint of an undiscovered type of mathematicaltreatment of biological communitiesrsquo (p 418) Laterin 1979 Hutchinson added the statement lsquoLinde-manrsquos (1942) paper was the first one to indicate howbiological communities could be expressed as net-

copy Inter-Research 2014 middot wwwint-rescomCorresponding author nathalieniquilunicaenfr

Reaction of an estuarine food web to disturbanceLindemanrsquos perspective

Nathalie Niquil1 Alexandra Baeta2 Joatildeo Carlos Marques2 Aureacutelie Chaalali1 Jeacutereacutemy Lobry3 Joana Patriacutecio4

1CNRS UMR 7208 BOREA FRE 3484 BIOMEA Universiteacute de Caen Basse-Normandie IBFA Esplanade de la Paix CS 14032 14032 Caen Cedex 5 France

2IMAR minus Institute of Marine Research Marine and Environmental Research Centre (IMAR-CMA) co Department of Life Sciences Faculty of Sciences and Technology University of Coimbra 3004-517 Coimbra Portugal

3Irstea groupement de Bordeaux 50 avenue de Verdun-Gazinet 33612 Cestas France4European Commission Joint Research Centre Institute for Environment and Sustainability 21027 Ispra Italy

ABSTRACT Currently a central demand of coastal and marine food web researchers is the searchfor holistic and functional ecosystem health indicators Based on concepts directly derived fromR L Lindemanrsquos work (Lindeman 1942 Ecology 23399ndash418) 6 estuarine food web models wereanalysed to assess the potential impact of multiple disturbances on the trophic state of food websThe models described a Zostera noltii meadow and a bare sediment site in the Mondego estuaryPortugal during 3 distinct time periods corresponding to (1) nutrient enrichment (2) the imple-mentation of mitigation measures and (3) after a centenary flood We tested 4 candidate metricsdirectly derived from the trophic-dynamic concepts proposed by Lindeman (1942) food chainlength the detritivoryherbivory ratio trophic efficiency per trophic level and mean trophic effi-ciency Based on this case study none of the metrics are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of disturbances on the ecosystemrsquos trophic stateTrophic efficiency appeared to be sensitive to external pressures However the mean trophic effi-ciency of the system might not be sufficient to draw conclusions about the trophic efficiency of theecosystem trophic efficiency should be observed at the trophic level in order to better understandcascading effects within food webs

KEY WORDS Ecosystem functionmiddot Food web models middot Ecological network analysis middot Trophic efficiency middot Ecological indicator middot Mondego estuary

Resale or republication not permitted without written consent of the publisher

Contribution to the Theme Section lsquoTrophodynamics in marine ecologyrsquo FREEREE ACCESSCCESS

Mar Ecol Prog Ser 512 141ndash154 2014

works or channels through which energy is flowingand being dissipated just as would be the case withelectricity flowing through a network of conductorsrsquo(Hutchinson 1979 p 246)

In the same period Cook (1977) pointed out thatLindemanrsquos contribution stressed the major role oftrophic functions particularly quantitative relationsin the determination of community patterns throughsuccession and established the validity of a theoreticalorientation in ecology by creating a theoretical modelfor trophic interactions quantitatively represented bymathematical relations This allowed Linde man to de-velop a number of predictions with which the validityof the model could be assessed Finally the trophic-dynamic approach identified a fundamental dynamicprocess mdash energy flow mdash with which the seasonaltrophic relations of organisms could be integrated intothe long-term process of community change

Lindemanrsquos work was also criticised For instanceCousins (1985) found Lindemanrsquos description of eco-systems to be an oversimplified description of eco -systems since relationships are rarely linear andbecause species often consume at different trophiclevels In response to Cousinsrsquo remarks Ulanowicz(1995) defended the Lindeman scheme as a quantita-tive description of behaviour at the community levelAccording to Ulanowicz (1995) it was only necessaryto abandon the notion of discrete trophic levels andthen most ambiguities concerning trophic statuswould vanish That is a given taxon does not need tobe wholly assigned to a single trophic level and viceversa This idea built upon the proposition of Odum ampHeald (1975) that species could be positioned at somenon-integer trophic levels (ie the weighted averageof the number of links in the various pathways alongwhich it consumes) Ulanowicz proposed the use ofwhat he called the lsquoLindeman spinersquo a linear chainthat sums up the information of a complex trophicnetwork by dividing the compartments and theiractivity into discrete levels the first level being com-posed of both primary producers and detritus Thistool maps the complicated networks of the eco -systemrsquos flows into a common topological form thatallows distinct ecosystems to be compared

Complementary to these studies are the conceptsdeveloped by Odum (1969 1985) concerning foodweb analysis in the context of the development ofecosystems and in particular in situations of stressdefined as lsquounusual external perturbationsrsquo (Odum1985) Today the key question in ecology remainsHow does an ecosystem react to disturbance Cur-rently one of the major challenges is to find indica-tors that can adequately assess the ecological condi-

tion of a natural system considering it as a functionalunit (Rombouts et al 2013) Interestingly a candidatemetric is lsquoefficiencyrsquo first defined by Lindeman(1942) for each trophic level Working on Lake Men-dota he produced the first measures of this indexdescribed as the percentage of the production of onetrophic level converted to production by the nexttrophic level Since then this index has been and isstill widely used to assess the trophic state of an eco-system and is either known as lsquotrophic efficiencyrsquo orlsquotransfer efficiencyrsquo (eg Pauly amp Christensen 1995Jennings et al 2002 Libralato et al 2008 Coll et al2009 Fiscus 2009 Pranovi amp Link 2009 Baird et al2012 Heymans et al 2012) Reflecting on differentscenarios of fishing pressure Gascuel (2005) pro-posed that trophic efficiency should be considered asa key characteristic of the functioning of ecosystemsassociated with its sensitivity to fishing pressure Pranovi amp Link (2009) also showed different trophicefficiency values among fishery-exploited systemsarguing the need for between-site comparisons tomake the generalisation that is necessary for success-ful ecosystem-based management An emergingspecialised literature proposed the use of trophic effi-ciency as a descriptor of ecosystem health especiallyin lakes (Xu amp Mage 2001 Hecky 2006) and in eco-system-based fishery management (Rochet amp Tren -kel 2003 Libralato et al 2008 Coll et al 2009)

Another candidate metric which was also directlyinspired by Lindeman (1942) when he referred to therole of saprophageous organisms and heterotrophicbacteria and which was formalised by Ulanowicz(1992) is the detritivoryherbivory (DH) ratio (orherbivorydetritivory ratio) This ratio has alreadybeen tested as a candidate for defining functionalindicators of the food web but the observed trendvaried according to the situation described Forexample Ulanowicz (1992) observed a lower DHratio in disturbed situations whereas Dame amp Chris-tian (2007) observed exactly the opposite trend

Based on Lindemanrsquos original concepts the scien-tific community developed and evaluated new indi-cators to depict the trophic status of ecosystems andto link them to system stability and resilience (egUlanowicz amp Puccia 1990 Christensen amp Pauly 1993Libralato et al 2006 2008 Heymans et al 2007)With the ultimate goal of providing functional indica-tors for ecosystem-based management and follow-ing European directives (eg Good Ecological StatusWater Framework Directive) specific indicators wereproposed For instance regarding fishery pressuredifferent indices were commonly adopted and usedin the literature such as the primary production

142

Mar Ecol Prog Ser 512 141ndash154 2014

works or channels through which energy is flowingand being dissipated just as would be the case withelectricity flowing through a network of conductorsrsquo(Hutchinson 1979 p 246)

In the same period Cook (1977) pointed out thatLindemanrsquos contribution stressed the major role oftrophic functions particularly quantitative relationsin the determination of community patterns throughsuccession and established the validity of a theoreticalorientation in ecology by creating a theoretical modelfor trophic interactions quantitatively represented bymathematical relations This allowed Linde man to de-velop a number of predictions with which the validityof the model could be assessed Finally the trophic-dynamic approach identified a fundamental dynamicprocess mdash energy flow mdash with which the seasonaltrophic relations of organisms could be integrated intothe long-term process of community change

Lindemanrsquos work was also criticised For instanceCousins (1985) found Lindemanrsquos description of eco-systems to be an oversimplified description of eco -systems since relationships are rarely linear andbecause species often consume at different trophiclevels In response to Cousinsrsquo remarks Ulanowicz(1995) defended the Lindeman scheme as a quantita-tive description of behaviour at the community levelAccording to Ulanowicz (1995) it was only necessaryto abandon the notion of discrete trophic levels andthen most ambiguities concerning trophic statuswould vanish That is a given taxon does not need tobe wholly assigned to a single trophic level and viceversa This idea built upon the proposition of Odum ampHeald (1975) that species could be positioned at somenon-integer trophic levels (ie the weighted averageof the number of links in the various pathways alongwhich it consumes) Ulanowicz proposed the use ofwhat he called the lsquoLindeman spinersquo a linear chainthat sums up the information of a complex trophicnetwork by dividing the compartments and theiractivity into discrete levels the first level being com-posed of both primary producers and detritus Thistool maps the complicated networks of the eco -systemrsquos flows into a common topological form thatallows distinct ecosystems to be compared

Complementary to these studies are the conceptsdeveloped by Odum (1969 1985) concerning foodweb analysis in the context of the development ofecosystems and in particular in situations of stressdefined as lsquounusual external perturbationsrsquo (Odum1985) Today the key question in ecology remainsHow does an ecosystem react to disturbance Cur-rently one of the major challenges is to find indica-tors that can adequately assess the ecological condi-

tion of a natural system considering it as a functionalunit (Rombouts et al 2013) Interestingly a candidatemetric is lsquoefficiencyrsquo first defined by Lindeman(1942) for each trophic level Working on Lake Men-dota he produced the first measures of this indexdescribed as the percentage of the production of onetrophic level converted to production by the nexttrophic level Since then this index has been and isstill widely used to assess the trophic state of an eco-system and is either known as lsquotrophic efficiencyrsquo orlsquotransfer efficiencyrsquo (eg Pauly amp Christensen 1995Jennings et al 2002 Libralato et al 2008 Coll et al2009 Fiscus 2009 Pranovi amp Link 2009 Baird et al2012 Heymans et al 2012) Reflecting on differentscenarios of fishing pressure Gascuel (2005) pro-posed that trophic efficiency should be considered asa key characteristic of the functioning of ecosystemsassociated with its sensitivity to fishing pressure Pranovi amp Link (2009) also showed different trophicefficiency values among fishery-exploited systemsarguing the need for between-site comparisons tomake the generalisation that is necessary for success-ful ecosystem-based management An emergingspecialised literature proposed the use of trophic effi-ciency as a descriptor of ecosystem health especiallyin lakes (Xu amp Mage 2001 Hecky 2006) and in eco-system-based fishery management (Rochet amp Tren -kel 2003 Libralato et al 2008 Coll et al 2009)

Another candidate metric which was also directlyinspired by Lindeman (1942) when he referred to therole of saprophageous organisms and heterotrophicbacteria and which was formalised by Ulanowicz(1992) is the detritivoryherbivory (DH) ratio (orherbivorydetritivory ratio) This ratio has alreadybeen tested as a candidate for defining functionalindicators of the food web but the observed trendvaried according to the situation described Forexample Ulanowicz (1992) observed a lower DHratio in disturbed situations whereas Dame amp Chris-tian (2007) observed exactly the opposite trend

Based on Lindemanrsquos original concepts the scien-tific community developed and evaluated new indi-cators to depict the trophic status of ecosystems andto link them to system stability and resilience (egUlanowicz amp Puccia 1990 Christensen amp Pauly 1993Libralato et al 2006 2008 Heymans et al 2007)With the ultimate goal of providing functional indica-tors for ecosystem-based management and follow-ing European directives (eg Good Ecological StatusWater Framework Directive) specific indicators wereproposed For instance regarding fishery pressuredifferent indices were commonly adopted and usedin the literature such as the primary production

142

Niquil et al Estuarine food web and disturbances

required to sustain fisheries (Christensen amp Pauly1993) or the L index which provides a basis for esti-mating the maximum number of lsquoacceptablersquo catches(Libralato et al 2008) From a more functional per-spective the lsquokeystonenessrsquo index was developed byLibralato et al (2006) in order to assess the structur-ing role of species (or functional groups) in foodwebs and the Mixed Trophic Impact index toaddress the effect that an increase in the biomass ofone impacting group would have on the biomass ofanother (Ulanowicz amp Puccia 1990)

Because they attract a large human populationand because of their position at the intersection of themarine and terrestrial realms estuaries are dynamicsystems subjected to multiple and diverse pressuresThese include pollution excess nutrient loadinghabitat destruction and biodiversity changes (Hal -pern et al 2008) Estuaries are also naturally stressedsystems and Elliott amp Quintino (2007) emphasisedhow difficult it is to differentiate natural from humanstressors They are the object of numerous studies onthe functioning of ecosystems and authors often aimto describe how these ecosystems evolve in this multiple-pressure context Some of these studieshave been based on methodologies derived fromtrophodynamic concepts and have used inter-systemcomparisons of either different periods or sites(Christian et al 2005 2009 Patriacutecio et al 2006Leguerrier et al 2007 Heymans et al 2012 Niquil etal 2012 Chrystal amp Scharler 2014) Inthese studies the set of indices com-pared is either wide or focused onascendency-derived indices The origi-nality of the present paper lies in itsfocus (for an inter-system comparison)on the indices directly derived fromLindemanrsquos (1942) original paper

In this paper we tested whetherLinde manrsquos trophic-dynamic conceptscould be used to derive holistic andfunctional indicators of ecosystemhealth For this we assessed the im pactof disturbances affecting estuarine foodwebs on properties directly derivedfrom Lindeman (1942) We analysedthe effects of eutrophication changesin the hydro dynamic regime during theimplementation of mitigation measuresand an extreme flood event on severaltrophic-dynamic at tributes mean tro -phic level trophic efficiencies (trophicefficiency per tro phic level and meantrophic efficiency from level II to IV)

the DH ratio and Lindemanrsquos spine length Due tothe large amount of information on the structure andfunctioning of the intertidal food webs in the Mon-dego estuary (Portugal) (Patriacutecio amp Marques 2006Patriacutecio et al 2004 2009 Baeta et al 2009ab 2011)this system was chosen as a case study

MATERIALS AND METHODS

Study site

The Mondego estuary is a relatively small (860 ha)warm-temperate poly haline intertidal system loca -ted on the western Atlantic coast of Portugal It con-sists of 2 arms North arm and South arm (Fig 1) Thesouthern arm is characterised by intertidal mudflats(almost 75 of the area) which are exposed at lowtide The tidal range varies between 035 and 33 mdepending on the site and the tide coefficient whilethe water residence time varies between 1 or 2 d(northern arm) and 3 d (southern arm)

From 1991 to 1997 the communication betweenthe 2 arms of the estuary became totally interruptedin the upstream area which caused the river dis-charge to flow essentially through the northern armConsequently water circulation in the southern armbecame mainly dependent on tides and on the smallfreshwater input from a tributary the Pranto River

143

Fig 1 Locations of the sampling sites in the Mondego estuary Zostera mead-ows and bare sediment sites (grey circles) Left panel shows the change inarea covered by Zostera noltii in the southern arm of Mondego estuary be-tween 1986 and 2004 Mapping of benthic vegetation is based on field obser-vations aerial photographs and Geographical Information System analysis

(ArcView GIS v82)

Mar Ecol Prog Ser 512 141ndash154 2014

artificially controlled by a sluice (Marques et al2009) This led to clear eutrophication symptoms inthe southern arm (eg green macroalgal blooms)(Leston et al 2008 Patriacutecio et al 2009) In 1997 todecrease these eutrophication symptoms and to testways of improving the systemrsquos conditions the fresh-water discharge from the Pranto River sluice into thesouthern arm was reduced to a minimum in order todecrease nutrient inputs and was diverted to thenorthern arm by another sluice located furtherupstream Moreover the communication betweenthe northern and southern arms was re-establishedto a very limited extent (periods of only 15 to 2 hbefore and after each high-tide peak through a sec-tion of only 1 m2) to improve water circulation (Mar-ques et al 2009)

A long-term study of the Mondego estuary ecosys-tem carried out since the mid-1980s made it possibleto determine the systemrsquos responses to these modifi-cations to its physical conditions such as changes inthe Zostera noltii beds and in the green macroalgaeUlva spp coverage and biomass (eg Patriacutecio et al2009) Following the interruption of the upstreamcommunication between the 2 arms ecological condi-tions in the southern arm rapidly deteriorated Thecombined effect of increased water residence timeand higher nutrient concentrations was the main driv-ing force behind the seasonal Ulva spp blooms andthe consequent severe reduction in the area occupiedby Z noltii beds due to competition with the macro-algae (Patriacutecio et al 2009) The shift in benthic primaryproducers changed the benthic macrofauna trophicstructure (eg Marques et al 1997 2003 Patriacutecio ampMarques 2006) According to Patriacutecio et al (2009) after experimental mitigation measures were appliedin 1998 this trend appeared to reverse to a certain extent as the area occupied by Z noltii was partiallyregained the green Ulva spp blooms decreased mar -kedly and the macrofauna assemblages gave signs ofrecovering their former condition of the late 1980s

The winter of 20002001 was characterised byunprecedented high precipitation values comparedto the average long-term precipitation (200020011802 mm 1940 to 1997 1031 mm) which caused oneof the largest flood events of the century in the Mondego catchment area

Construction of mass-balanced models

Six mass-balanced models were constructed usingEcopath software (Christensen amp Pauly 1992) Theinput information came from an extensive field-

sampling program was expanded with stable isotopeanalysis (SIA) and completed with published refer-ences Fieldwork was conducted in 2 areas located inthe southern arm of the Mondego estuary (Z noltiimeadows and bare sediment area) during 3 periods19931994 19992000 and 20012002 Detailed de -scriptions of the study sites sampling program andprotocols network construction dietary matrices andSIA can be found in Baeta et al (2011) The quantita-tive contribution of the food sources of consumerswas estimated using mixing models as well as thequantitative information on sources of nutrition of theestuarine invertebrates inhabiting the 2 study areasAlthough information from the literature was used inthis study (equations based on local data on individ-ual size for metabolic parameters of all and diets of afew compartments) the majority of the inputs camefrom data directly collected from the sites during theperiods represented by the models (ie communitystructure biomass of all and diets of several compart-ments PB ratios) The food webs were built usingdata for benthic macrofauna species representingthe benthic food web and no compartments regard-ing higher trophic levels mdash fish and birds mdash wereincluded in the models as no information preciselycovered the periods of interest (for further details seePatriacutecio et al 2004)

Trophic analysis

For the 6 food webs built the trophic analysis wasderived from the lsquoNetwork Analysisrsquo routine includedin the Ecopath with Ecosim software (v6386480)

The lsquomean trophic levelrsquo was calculated for eachconsumer compartment as the length of the differentconsumption pathways averaged with a weight cor-responding to the quantity of carbon that passesalong each pathway (Odum amp Heald 1975 Ulano -wicz 1995) Level I corresponds to the primary pro-ducers and the non-living detrital compartmentsThe strict herbivores or detritivores consequentlyoccupy a position of II (one plus one) Other con-sumers are then allocated to several discrete trophicpositions according to the amounts that reach themalong feeding pathways of various lengths and themean trophic position is calculated as the mean path-way with a weight corresponding to the flow valuein g AFDW mminus2 yrminus1 This indicator was analysedtogether with the total consumption (sum of all flowsfrom the different preys) of the consumer

The same distribution of compartments within inte-ger trophic levels was used to build the lsquoLindeman

144

Mar Ecol Prog Ser 512 141ndash154 2014

artificially controlled by a sluice (Marques et al2009) This led to clear eutrophication symptoms inthe southern arm (eg green macroalgal blooms)(Leston et al 2008 Patriacutecio et al 2009) In 1997 todecrease these eutrophication symptoms and to testways of improving the systemrsquos conditions the fresh-water discharge from the Pranto River sluice into thesouthern arm was reduced to a minimum in order todecrease nutrient inputs and was diverted to thenorthern arm by another sluice located furtherupstream Moreover the communication betweenthe northern and southern arms was re-establishedto a very limited extent (periods of only 15 to 2 hbefore and after each high-tide peak through a sec-tion of only 1 m2) to improve water circulation (Mar-ques et al 2009)

A long-term study of the Mondego estuary ecosys-tem carried out since the mid-1980s made it possibleto determine the systemrsquos responses to these modifi-cations to its physical conditions such as changes inthe Zostera noltii beds and in the green macroalgaeUlva spp coverage and biomass (eg Patriacutecio et al2009) Following the interruption of the upstreamcommunication between the 2 arms ecological condi-tions in the southern arm rapidly deteriorated Thecombined effect of increased water residence timeand higher nutrient concentrations was the main driv-ing force behind the seasonal Ulva spp blooms andthe consequent severe reduction in the area occupiedby Z noltii beds due to competition with the macro-algae (Patriacutecio et al 2009) The shift in benthic primaryproducers changed the benthic macrofauna trophicstructure (eg Marques et al 1997 2003 Patriacutecio ampMarques 2006) According to Patriacutecio et al (2009) after experimental mitigation measures were appliedin 1998 this trend appeared to reverse to a certain extent as the area occupied by Z noltii was partiallyregained the green Ulva spp blooms decreased mar -kedly and the macrofauna assemblages gave signs ofrecovering their former condition of the late 1980s

The winter of 20002001 was characterised byunprecedented high precipitation values comparedto the average long-term precipitation (200020011802 mm 1940 to 1997 1031 mm) which caused oneof the largest flood events of the century in the Mondego catchment area

Construction of mass-balanced models

Six mass-balanced models were constructed usingEcopath software (Christensen amp Pauly 1992) Theinput information came from an extensive field-

sampling program was expanded with stable isotopeanalysis (SIA) and completed with published refer-ences Fieldwork was conducted in 2 areas located inthe southern arm of the Mondego estuary (Z noltiimeadows and bare sediment area) during 3 periods19931994 19992000 and 20012002 Detailed de -scriptions of the study sites sampling program andprotocols network construction dietary matrices andSIA can be found in Baeta et al (2011) The quantita-tive contribution of the food sources of consumerswas estimated using mixing models as well as thequantitative information on sources of nutrition of theestuarine invertebrates inhabiting the 2 study areasAlthough information from the literature was used inthis study (equations based on local data on individ-ual size for metabolic parameters of all and diets of afew compartments) the majority of the inputs camefrom data directly collected from the sites during theperiods represented by the models (ie communitystructure biomass of all and diets of several compart-ments PB ratios) The food webs were built usingdata for benthic macrofauna species representingthe benthic food web and no compartments regard-ing higher trophic levels mdash fish and birds mdash wereincluded in the models as no information preciselycovered the periods of interest (for further details seePatriacutecio et al 2004)

Trophic analysis

For the 6 food webs built the trophic analysis wasderived from the lsquoNetwork Analysisrsquo routine includedin the Ecopath with Ecosim software (v6386480)

The lsquomean trophic levelrsquo was calculated for eachconsumer compartment as the length of the differentconsumption pathways averaged with a weight cor-responding to the quantity of carbon that passesalong each pathway (Odum amp Heald 1975 Ulano -wicz 1995) Level I corresponds to the primary pro-ducers and the non-living detrital compartmentsThe strict herbivores or detritivores consequentlyoccupy a position of II (one plus one) Other con-sumers are then allocated to several discrete trophicpositions according to the amounts that reach themalong feeding pathways of various lengths and themean trophic position is calculated as the mean path-way with a weight corresponding to the flow valuein g AFDW mminus2 yrminus1 This indicator was analysedtogether with the total consumption (sum of all flowsfrom the different preys) of the consumer

The same distribution of compartments within inte-ger trophic levels was used to build the lsquoLindeman

144

Niquil et al Estuarine food web and disturbances

spinersquo of each food web This mapping summarisesthe complicated food web in terms of a single linearchain with discrete trophic levels (Ulanowicz 1995)Once the compartment is decomposed within integertrophic levels the flows are decomposed accordinglyand the contributions of each trophic level to respira-tion imports and exports are also evaluated (Baird ampUlanowicz 1989) The building of this linear chainenables the efficiency of the transfer from one aggre-gated trophic level to the next to be calculated Thisindex is calculated as the fraction of the total carboninput to a given level that is transmitted to the nexthigher level (Baird amp Ulanowicz 1989) The lsquomeantrophic efficiencyrsquo of the system is then derived fromthe geometric mean of the efficiencies of trophic levels II to IV (Baird amp Ulanowicz 1989 Heymans etal 2012) In the Ecopath software trophic efficiencyis calculated from level II up so the mean trophicefficiency does not consider level I nor detritus

The DH ratio (ie the values of the detritivory flowfrom detritus to level II divided by the value of theherbivory flow from primary producers to level II)was calculated for each of the 6 food webs

In order to compare these indicators directly de rivedfrom Lindeman (1942) with an indicator based on in-formation theory the lsquorelative internal ascendencyrsquo(internal ascendencyinternal development capacityAiCi) was calculated (Monaco amp Ulano wicz 1997)

RESULTS

The mean trophic level for each consumer com-partment is presented for the 3 food webs of theZostera bed and for the bare sediment food webs(Fig 2) over time The compartments are ranked indecreasing order of mean trophic levels (left axis)The species whose consumption exceeds the maxi-

145

Fig 2 Mean trophic level (TL no unit left axis) and consumption (g AFDW mminus2 yrminus1 right axis) for Zostera bed and bare sedi-ment food webs The scale of the right axis was truncated at 50 g AFDW mminus2 yrminus1 in order to keep it readable Values above thisthreshold are described in lsquoResultsrsquo Z = Zostera bed bs = bare sediment 93 = 19931994 99 = 19992000 01 = 20012002

Mar Ecol Prog Ser 512 141ndash154 2014

mum value of the graph (50 g AFDW mminus2 yrminus1) are thefollowing (1) Hydrobia ulvae (478 g AFDW mminus2 yrminus1

in 1993 1994 319 in 19992000 453 in 20012002)Scrobicularia plana (84 g AFDW mminus2 yrminus1 in 19992000 79 in 20012002) and Hediste diversicolor (70 gAFDW mminus2 yrminus1 in 20012002) at the Zostera sites(Fig 2) and (2) H ulvae (161 g AFDW mminus2 yrminus1 in19931994) S plana (67 g AFDW mminus2 yrminus1 in 19931994 157 in 19992000 123 in 20012002) andCyathura carinata (128 g AFDW mminus2 yrminus1 in 19931994 72 in 19992000 102 in 20012002) (Fig 2) atthe bare sediment sites

These high consumption values are measured fortaxa with a trophic level around 2 For instance thegastropod H ulvae is a grazer that feeds mainly onbenthic diatoms The same was observed for the sus-pension feeder S plana which also consumes phyto-plankton and detritus The trophic level value forthese highly active compartments (ie with a highconsumption) remains low with maximal valuesobserved for H diversicolor (max 205) and C cari-nata (max 217) Both are omnivores feeding on awide variety of different animals algae and detritusbut their diet remains dominated by detritus Themain animal prey for H diversicolor is H ulvae(Zostera bed in 20012002) and for C carinata it is Hdiversicolor (bare sediment in 20012002)

Among less active compartments (ie those with thelowest value of total consumption) those belonging toa trophic level above II correspond to the omnivorousand carnivorous annelids Eteone flava Glicera tri-dactyla Nephtys spp Mista picta Lumbri nereis Dio -patra neapolitana and Nemertini and to the decapodsCarcinus maenas and Crangon crangon (Fig 2) Thefood sources of these species in clude animals algaeand detritus (see dietary matrices in Baeta et al 2011supplementary material) The compartment that mostoften is the highest predator at the mean trophic levelis C maenas Its diet in Zostera beds as well as inbare sediment models is composed of 10 to 15 de-tritus (including bacteria) and 4 primary producersthe remaining 81 to 86 being a great variety of het-erotrophic prey items Its main prey consists of H di-versicolor C crangon and S plana Correlated withthe highest diversity in the Zostera bed models Cmaenas is positioned at a higher trophic level (298 to305) than in bare sediment models (287 to 295) Inone of the bare sediment food webs (19931994) 2taxa are positioned at a higher level than C maenasNemertini and G tridac tyla Their mean trophic lev-els however remain lower than III (291 and 290 re-spectively) In one case of the Zostera bed model(19992000) 2 species are positioned at a higher level

than C maenas E flava and M picta both positionedat a mean trophic level of 304

The carbon flow networks of the 6 models weretransformed into Lindeman spines (sensu Ulanowicz1995) The models for the 3 periods of the Zostera bedare illustrated in Fig 3 and those of the bare sedi-ment area in Fig 4

In order to make a distinction between detritivoryand herbivory the detritus pool (D) was separatedfrom the primary producers (P) Together both poolsform level I This allowed the DH ratio to be calcu-lated considering a functional emergent property offood web organisation (Table 1) Detritivory wasalways superior to herbivory ranging from 103 to225 In each food web the DH value of the Z noltiimeadows food web was higher than that of the baresediment area showing that detrital pathways areimportant for Zostera bed functioning Although thefood webs temporally showed differences in the DHratio it is worth noting that in both areas there wasan increase in the DH value after the century flood

The DH ratio was always higher in the Zosterameadows mainly due to the high biomass of H ulvaefound in this area H ulvae feeds not only on micro-phytobenthos (MPB) but also on detritus (in highproportion) In Zostera bed Site Z01 the DH ratiowas slightly higher than in Site Z93 due to theincrease in S plana biomass in 1999 and 2001 how-ever in 1999 the DH ratio was lower because thebiomass of H ulvae was much lower

In the bare sediment the DH ratio was lower in1993 and higher in 2001 In 1993 the consumption ofdetritus by H ulvae was significantly lower since theavailability of MPB was significantly higher The DHratio was close to 1 since the consumption of produc-ers by H ulvae was very similar to that of C carinataon detritus (in fact consumption by H ulvae + Splana of producers = consumption by C carinata + Hulvae of detritus) In 2001 the biomass of H ulvaewas very low and consequently its consumption wasalso very low C carinata remained the species withthe highest consumption of detritus

Examining the detritus pool inputs it was also pos-sible to differentiate between locally produced detri-tus and inputs originating outside Compared to alldetritus flows (inputs from outside + locally produceddetritus) 59 to 66 of the detritus was producedlocally The values for the Zostera bed food webs (59to 66) were similar to those for the bare sedimentmodels (60 to 63) In both systems this percentageincreased from 1993 to 1999 (after the implementa-tion of the mitigation measures) followed by a de -crease after the flood event (from 1999 to 2001)

146

Niquil et al Estuarine food web and disturbances 147

P2940

1152

1900II3717

6915

0000

3711

197800320 III

1245

1955

0000

1270

186400900 IV

0118

0193

0000

1235

017000870 V

000946

00157

0000

00976

0010200651 VI

0000570

0000884

0000

000587

000071300752 VII

0000039

0000065

0000

0000413

000005900898 VIII

0000

0000

0000

0000

D3206

5273

2990

2276 6147 0553 00496 000291 0000185 00005334

4279

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P3472

2009

1948II2843

7180

0000

3468

130400226 III

0691

1144

0000

8780

17060121 IV

00941

0176

0000

1232

013100682 V

000704

00134

0000

00924

00083800586 VI

0000440

0000810

0000

000576

000055200617 VII

0000023

0000045

0000

0000307

000002000417 VIII

0000

0000

0000

0000

D3606

5113

5110

2180 3565 0551 00423 000263 0000144 00007332

3825 Z93

P3104

6847

1894II3801

4333

0000

2741

107400235 III

0933

1087

0000

6895

14770132 IV

0128

0139

0000

0954

013500883 V

00109

00123

0000

00813

0010700816 VI

0000869

0000949

0000

000649

000084100805 VII

0000062

0000068

0000

0000461

000005600756 VIII

0000000

0000001

0000

0000004

D2987

5113

1837

1720 2840 0443 00393 000311 0000223 00000013590

2671

Z01

Z99

P

3368

6845

1005II

3141

3501

0000

1604

892000334 III

1104

0709

0000

5761

058500624 IV

00751

00423

0000

0403

00093400146 V

000121

0000549

0000

000674

000015300149 VI

0000014

0000006

0000

0000077

0000001000504 VII

0000

0000

0000

0000

D

3373

3050

1857

9758 3034 0226 000338 0000038 00002866

1662

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P

3674

7070

1914II

2907

2421

0000

2337

491900127 III

0406

0359

0000

3398

05870108 IV

00482

00436

0000

0418

0045100700 V

000304

000277

0000

00256

00014100348 VI

0000051

0000044

0000

0000420

000002200321 VII

0000

0000

0000

0000

D

3373

3058

2992

1495 1430 0180 00136 0000234 00004503

1964

P

3454

8528

1332II

3300

4725

0000

1690

586400209 III

0783

0447

0000

4341

052700790 IV

00736

00478

0000

0424

00092000147 V

000130

0000859

0000

000761

0000018000161 VI

0000001

0000000

0000

0000004

D

3161

3058

1611

1062 1803 0194 000346 00000022693

1475

bs93

bs01

bs99

Fig 3 Lindeman spines for each of the 3 Zostera meadow food webs (Z = Zostera bed 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Fig 4 Lindeman spines for each of the bare sediment area food webs (bs = bare sediment 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Mar Ecol Prog Ser 512 141ndash154 2014

The Lindeman spines also illustrate the flows ofrespiration detritus production and export leavingeach integer trophic level In the present case onlyrespiration and detritus production were taken intoaccount export was estimated as null in the model(Baeta et al 2011) Because most of the heterotrophicbiomass was located at level II (97 to 99) respira-tion flows followed the biomass distribution closelywith higher values at level II

The Zostera meadows food webs have longerchains than the bare sediment food webs This obser-vation results from the higher number of compart-ments in the Zostera bed food webs which in turnreflects the higher species richness of this area com-pared with the bare sediment area The highesttrophic efficiency value was always found at level III(Fig 5) ranging between 6 (bare sediment in 2001)and 13 (Zostera bed in 1999)

In terms of mean trophic efficiency (Fig 6) the values presented by Zostera bed models were higher(57 to 65) than those of bare sediment area mod-els (29 to 46) The mitigation measures in 1999corresponded to an increase in mean trophic effi-ciency in the Zostera site compared with the eutroph-ication period (57 to 65) followed by a minutedecrease after the flood event In the food webs of thebare sediment area the mean trophic efficiency val-ues were higher in the eutrophic period (46) thanin the other 2 periods (29 to 31) Our observationssuggest that the implementation of mitigation meas-ures had a stronger impact on the mean trophic effi-ciencies than the extreme flood event Compared toAiCi which describes the specialisation of the flowswe observed that the trend of variation over time wascomparable with the mean trophic efficiency for thebare sediment site but both were in contradiction tothe Zostera site In both sites AiCi was lower in19992000 and returned to a higher value in20012002

DISCUSSION

The Mondego estuary monitoring programmebegan in the mid-1980s This ecological monitoringprogramme was developed because it is well knownthat many ecological processes occur through or aredriven by factors acting on long temporal and largespatial scales Many scientists have collected valu-able information about the estuary over the last 30 yrincluding data on physical conditions primary pro-ducers zooplankton benthic macrofauna fish andbirds (eg Marques et al 2003 2013 Patriacutecio et al2004 2009 Lopes et al 2005 Dolbeth et al 2007Martinho et al 2007 Cardoso et al 2010 Falcatildeo et al2012 Veriacutessimo et al 2012 Neto et al 2013) The

148

02

022

024

026

028

03

032

034

036

0

001

002

003

004

005

006

007

Z93 Z99 Z01 bs93 bs99 bs01

AiC

i

TEm

TEm (left axis)AiCi (right axis)

Fig 5 Trophic efficiencies per trophic level for the 6 models(bs = bare sediment Z = Zostera bed 93 = 19931994 99 =19992000 01 = 20012002) Trophic level efficiency of levelIII = predation on level III by level IV consumption of level

II by level III

Fig 6 Mean trophic efficiency TEm (geometric mean of val-ues for integer levels calculated for trophic levels II to IV)and relative internal ascendency AiCi for the 6 models (bs= bare sediment Z = Zostera bed 93 = 19931994 99 =

19992000 01 = 20012002)

Site No comp TEm () DH ratio

Z93 36 57 196Z99 31 65 141Z01 24 63 225bs93 25 46 103bs99 20 29 111bs01 20 31 165

Table 1 Characteristics of the 6 original Ecopath models(Baeta et al 2011) No comp = number of compartmentsTEm = mean trophic efficiency DH = DetritivoryHer-bivory Z = Zostera bed site bs = bare sediment site 93 =

19931994 99 = 19992000 01 = 20012002

Niquil et al Estuarine food web and disturbances

analysis of these data was the basis for a betterunderstanding of the structure and functioning ofthis estuarine ecosystem

In the early and mid-1990s nutrient enrichment inthe system led to macroalgal blooms that decreasedthe previously dominant macrophyte communities(ie Zostera noltii) In 1998 mitigation measureswere implemented in order to restore the Z noltiibeds and the overall quality of the system A few spe-cies responded positively (eg higher biomass ofScrobicularia plana and Hediste diversicolor) othersnegatively (eg lower biomass of Hydrobia ulvae)which led to more structured and stable populationscloser to the less disturbed conditions observed in the1980s (Patriacutecio et al 2009) According to Baeta et al(2011) the Z noltii site showed a higher number oftrophic compartments and a higher level of systemactivity (ie sum of consumptions respiration flowsto detritus production total system throughput netprimary production and system omnivory index) thanthe bare sediment site The differences in total sys-tem throughput between the 2 sites for the 3 periodswere mainly due to differences in the biomass of pri-mary producers (higher primary production at theZostera site) In both habitats after recovery meas-ures were implemented there was an increase in Splana and H diversicolor biomass consumption res-piration and flows to detritus and a decrease in Hulvae biomass and associated flows which increasedagain after the centenary flood (Baeta et al 2011)This study showed that the estuarine food webs wereaffected differently by each disturbance event andsuggested that a high system throughput seems to beassociated with higher stress levels contradicting theidea that higher system activity is a sign of healthierconditions

Lindemanrsquos (1942) trophic-dynamic concepts andUlanowiczrsquos (1995) developments bring about an -other way to analyse the 6 case studies presented inBaeta et al (2011)

Advantages and limitations of the chosen numerical approach

This study was based on high quality models asrevealed by the large amount of local informationused Because the 6 models shared comparablestructures (differences among considered trophiccompartments were directly linked with differencesin biodiversity) the comparison could be based onindicators from network analysis shown to be sensi-tive to the structure of the food web model (Johnson

et al 2009) However to maintain the high reliabilityin the estimated flows we made the choice not toconsider birds and fish Even if these compartmentshad been described previously (eg Lopes et al 2005Martinho et al 2007) the information was not suffi-cient to exactly determine their role during eachstudied period and at each study site This has to betaken into account in our comparison In particular itreduces the possibility of adding other metrics to ourcomparison or comparing index values with thosederived from models of other estuarine ecosystemsthat usually consider a wider area and include birdsand fish (Christian et al 2005 Leguerrier et al 2007Heymans et al 2012)

Lindeman spine topology

The length of the Lindeman spines varied betweenhabitats (ie inter-habitat changes) and among yearswithin habitats (ie inter-annual differences) TheZostera bed chains were longer than those of thebare sediment site At first it may seem possible togeneralise that the trophic chain length is propor-tional to the number of compartments present in thefood web ie that more compartments will result in alonger chain However this trend is only verified insome case studies For example the food web of theZostera bed site in 2001 has 24 compartments but itschain is 1 step longer (number of trophic levels 8)than that of the bare sediment site in 1993 with 25compartments (number of trophic levels 7) Othervariables besides number of compartments probablycontribute to determining the length of the energychains Ulanowicz (1992) based on data taken from atidal marsh tributary creek of the Crystal River(Florida USA) and from a similar creek that was sub-ject to an average 6degC rise in ambient temperaturebecause of exposure to the effluent from an adjacentpower-generating station proposed that the lengthof the Lindeman spine is related to the level of stressie that shorter chains would reflect a higher degreeof stress According to Ulanowiczrsquos (1992) proposalthe bare sediment food webs with shorter chainswould reflect a higher level of stress in this habitatthan in the Zostera bed site If fact Z noltii hasadapted to the challenging conditions affecting inter-tidal habitats including environmental heterogene-ity and disturbance (Hemminga amp Duarte 2000) Thismacrophyte can act as a buffer to disruption essen-tially as a function of shelter offered by foliage andbecause this speciesrsquo root networks stabilise the sed-iment in which they grow

149

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Mar Ecol Prog Ser 512 141ndash154 2014

mum value of the graph (50 g AFDW mminus2 yrminus1) are thefollowing (1) Hydrobia ulvae (478 g AFDW mminus2 yrminus1

in 1993 1994 319 in 19992000 453 in 20012002)Scrobicularia plana (84 g AFDW mminus2 yrminus1 in 19992000 79 in 20012002) and Hediste diversicolor (70 gAFDW mminus2 yrminus1 in 20012002) at the Zostera sites(Fig 2) and (2) H ulvae (161 g AFDW mminus2 yrminus1 in19931994) S plana (67 g AFDW mminus2 yrminus1 in 19931994 157 in 19992000 123 in 20012002) andCyathura carinata (128 g AFDW mminus2 yrminus1 in 19931994 72 in 19992000 102 in 20012002) (Fig 2) atthe bare sediment sites

These high consumption values are measured fortaxa with a trophic level around 2 For instance thegastropod H ulvae is a grazer that feeds mainly onbenthic diatoms The same was observed for the sus-pension feeder S plana which also consumes phyto-plankton and detritus The trophic level value forthese highly active compartments (ie with a highconsumption) remains low with maximal valuesobserved for H diversicolor (max 205) and C cari-nata (max 217) Both are omnivores feeding on awide variety of different animals algae and detritusbut their diet remains dominated by detritus Themain animal prey for H diversicolor is H ulvae(Zostera bed in 20012002) and for C carinata it is Hdiversicolor (bare sediment in 20012002)

Among less active compartments (ie those with thelowest value of total consumption) those belonging toa trophic level above II correspond to the omnivorousand carnivorous annelids Eteone flava Glicera tri-dactyla Nephtys spp Mista picta Lumbri nereis Dio -patra neapolitana and Nemertini and to the decapodsCarcinus maenas and Crangon crangon (Fig 2) Thefood sources of these species in clude animals algaeand detritus (see dietary matrices in Baeta et al 2011supplementary material) The compartment that mostoften is the highest predator at the mean trophic levelis C maenas Its diet in Zostera beds as well as inbare sediment models is composed of 10 to 15 de-tritus (including bacteria) and 4 primary producersthe remaining 81 to 86 being a great variety of het-erotrophic prey items Its main prey consists of H di-versicolor C crangon and S plana Correlated withthe highest diversity in the Zostera bed models Cmaenas is positioned at a higher trophic level (298 to305) than in bare sediment models (287 to 295) Inone of the bare sediment food webs (19931994) 2taxa are positioned at a higher level than C maenasNemertini and G tridac tyla Their mean trophic lev-els however remain lower than III (291 and 290 re-spectively) In one case of the Zostera bed model(19992000) 2 species are positioned at a higher level

than C maenas E flava and M picta both positionedat a mean trophic level of 304

The carbon flow networks of the 6 models weretransformed into Lindeman spines (sensu Ulanowicz1995) The models for the 3 periods of the Zostera bedare illustrated in Fig 3 and those of the bare sedi-ment area in Fig 4

In order to make a distinction between detritivoryand herbivory the detritus pool (D) was separatedfrom the primary producers (P) Together both poolsform level I This allowed the DH ratio to be calcu-lated considering a functional emergent property offood web organisation (Table 1) Detritivory wasalways superior to herbivory ranging from 103 to225 In each food web the DH value of the Z noltiimeadows food web was higher than that of the baresediment area showing that detrital pathways areimportant for Zostera bed functioning Although thefood webs temporally showed differences in the DHratio it is worth noting that in both areas there wasan increase in the DH value after the century flood

The DH ratio was always higher in the Zosterameadows mainly due to the high biomass of H ulvaefound in this area H ulvae feeds not only on micro-phytobenthos (MPB) but also on detritus (in highproportion) In Zostera bed Site Z01 the DH ratiowas slightly higher than in Site Z93 due to theincrease in S plana biomass in 1999 and 2001 how-ever in 1999 the DH ratio was lower because thebiomass of H ulvae was much lower

In the bare sediment the DH ratio was lower in1993 and higher in 2001 In 1993 the consumption ofdetritus by H ulvae was significantly lower since theavailability of MPB was significantly higher The DHratio was close to 1 since the consumption of produc-ers by H ulvae was very similar to that of C carinataon detritus (in fact consumption by H ulvae + Splana of producers = consumption by C carinata + Hulvae of detritus) In 2001 the biomass of H ulvaewas very low and consequently its consumption wasalso very low C carinata remained the species withthe highest consumption of detritus

Examining the detritus pool inputs it was also pos-sible to differentiate between locally produced detri-tus and inputs originating outside Compared to alldetritus flows (inputs from outside + locally produceddetritus) 59 to 66 of the detritus was producedlocally The values for the Zostera bed food webs (59to 66) were similar to those for the bare sedimentmodels (60 to 63) In both systems this percentageincreased from 1993 to 1999 (after the implementa-tion of the mitigation measures) followed by a de -crease after the flood event (from 1999 to 2001)

146

Niquil et al Estuarine food web and disturbances 147

P2940

1152

1900II3717

6915

0000

3711

197800320 III

1245

1955

0000

1270

186400900 IV

0118

0193

0000

1235

017000870 V

000946

00157

0000

00976

0010200651 VI

0000570

0000884

0000

000587

000071300752 VII

0000039

0000065

0000

0000413

000005900898 VIII

0000

0000

0000

0000

D3206

5273

2990

2276 6147 0553 00496 000291 0000185 00005334

4279

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P3472

2009

1948II2843

7180

0000

3468

130400226 III

0691

1144

0000

8780

17060121 IV

00941

0176

0000

1232

013100682 V

000704

00134

0000

00924

00083800586 VI

0000440

0000810

0000

000576

000055200617 VII

0000023

0000045

0000

0000307

000002000417 VIII

0000

0000

0000

0000

D3606

5113

5110

2180 3565 0551 00423 000263 0000144 00007332

3825 Z93

P3104

6847

1894II3801

4333

0000

2741

107400235 III

0933

1087

0000

6895

14770132 IV

0128

0139

0000

0954

013500883 V

00109

00123

0000

00813

0010700816 VI

0000869

0000949

0000

000649

000084100805 VII

0000062

0000068

0000

0000461

000005600756 VIII

0000000

0000001

0000

0000004

D2987

5113

1837

1720 2840 0443 00393 000311 0000223 00000013590

2671

Z01

Z99

P

3368

6845

1005II

3141

3501

0000

1604

892000334 III

1104

0709

0000

5761

058500624 IV

00751

00423

0000

0403

00093400146 V

000121

0000549

0000

000674

000015300149 VI

0000014

0000006

0000

0000077

0000001000504 VII

0000

0000

0000

0000

D

3373

3050

1857

9758 3034 0226 000338 0000038 00002866

1662

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P

3674

7070

1914II

2907

2421

0000

2337

491900127 III

0406

0359

0000

3398

05870108 IV

00482

00436

0000

0418

0045100700 V

000304

000277

0000

00256

00014100348 VI

0000051

0000044

0000

0000420

000002200321 VII

0000

0000

0000

0000

D

3373

3058

2992

1495 1430 0180 00136 0000234 00004503

1964

P

3454

8528

1332II

3300

4725

0000

1690

586400209 III

0783

0447

0000

4341

052700790 IV

00736

00478

0000

0424

00092000147 V

000130

0000859

0000

000761

0000018000161 VI

0000001

0000000

0000

0000004

D

3161

3058

1611

1062 1803 0194 000346 00000022693

1475

bs93

bs01

bs99

Fig 3 Lindeman spines for each of the 3 Zostera meadow food webs (Z = Zostera bed 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Fig 4 Lindeman spines for each of the bare sediment area food webs (bs = bare sediment 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Mar Ecol Prog Ser 512 141ndash154 2014

The Lindeman spines also illustrate the flows ofrespiration detritus production and export leavingeach integer trophic level In the present case onlyrespiration and detritus production were taken intoaccount export was estimated as null in the model(Baeta et al 2011) Because most of the heterotrophicbiomass was located at level II (97 to 99) respira-tion flows followed the biomass distribution closelywith higher values at level II

The Zostera meadows food webs have longerchains than the bare sediment food webs This obser-vation results from the higher number of compart-ments in the Zostera bed food webs which in turnreflects the higher species richness of this area com-pared with the bare sediment area The highesttrophic efficiency value was always found at level III(Fig 5) ranging between 6 (bare sediment in 2001)and 13 (Zostera bed in 1999)

In terms of mean trophic efficiency (Fig 6) the values presented by Zostera bed models were higher(57 to 65) than those of bare sediment area mod-els (29 to 46) The mitigation measures in 1999corresponded to an increase in mean trophic effi-ciency in the Zostera site compared with the eutroph-ication period (57 to 65) followed by a minutedecrease after the flood event In the food webs of thebare sediment area the mean trophic efficiency val-ues were higher in the eutrophic period (46) thanin the other 2 periods (29 to 31) Our observationssuggest that the implementation of mitigation meas-ures had a stronger impact on the mean trophic effi-ciencies than the extreme flood event Compared toAiCi which describes the specialisation of the flowswe observed that the trend of variation over time wascomparable with the mean trophic efficiency for thebare sediment site but both were in contradiction tothe Zostera site In both sites AiCi was lower in19992000 and returned to a higher value in20012002

DISCUSSION

The Mondego estuary monitoring programmebegan in the mid-1980s This ecological monitoringprogramme was developed because it is well knownthat many ecological processes occur through or aredriven by factors acting on long temporal and largespatial scales Many scientists have collected valu-able information about the estuary over the last 30 yrincluding data on physical conditions primary pro-ducers zooplankton benthic macrofauna fish andbirds (eg Marques et al 2003 2013 Patriacutecio et al2004 2009 Lopes et al 2005 Dolbeth et al 2007Martinho et al 2007 Cardoso et al 2010 Falcatildeo et al2012 Veriacutessimo et al 2012 Neto et al 2013) The

148

02

022

024

026

028

03

032

034

036

0

001

002

003

004

005

006

007

Z93 Z99 Z01 bs93 bs99 bs01

AiC

i

TEm

TEm (left axis)AiCi (right axis)

Fig 5 Trophic efficiencies per trophic level for the 6 models(bs = bare sediment Z = Zostera bed 93 = 19931994 99 =19992000 01 = 20012002) Trophic level efficiency of levelIII = predation on level III by level IV consumption of level

II by level III

Fig 6 Mean trophic efficiency TEm (geometric mean of val-ues for integer levels calculated for trophic levels II to IV)and relative internal ascendency AiCi for the 6 models (bs= bare sediment Z = Zostera bed 93 = 19931994 99 =

19992000 01 = 20012002)

Site No comp TEm () DH ratio

Z93 36 57 196Z99 31 65 141Z01 24 63 225bs93 25 46 103bs99 20 29 111bs01 20 31 165

Table 1 Characteristics of the 6 original Ecopath models(Baeta et al 2011) No comp = number of compartmentsTEm = mean trophic efficiency DH = DetritivoryHer-bivory Z = Zostera bed site bs = bare sediment site 93 =

19931994 99 = 19992000 01 = 20012002

Niquil et al Estuarine food web and disturbances

analysis of these data was the basis for a betterunderstanding of the structure and functioning ofthis estuarine ecosystem

In the early and mid-1990s nutrient enrichment inthe system led to macroalgal blooms that decreasedthe previously dominant macrophyte communities(ie Zostera noltii) In 1998 mitigation measureswere implemented in order to restore the Z noltiibeds and the overall quality of the system A few spe-cies responded positively (eg higher biomass ofScrobicularia plana and Hediste diversicolor) othersnegatively (eg lower biomass of Hydrobia ulvae)which led to more structured and stable populationscloser to the less disturbed conditions observed in the1980s (Patriacutecio et al 2009) According to Baeta et al(2011) the Z noltii site showed a higher number oftrophic compartments and a higher level of systemactivity (ie sum of consumptions respiration flowsto detritus production total system throughput netprimary production and system omnivory index) thanthe bare sediment site The differences in total sys-tem throughput between the 2 sites for the 3 periodswere mainly due to differences in the biomass of pri-mary producers (higher primary production at theZostera site) In both habitats after recovery meas-ures were implemented there was an increase in Splana and H diversicolor biomass consumption res-piration and flows to detritus and a decrease in Hulvae biomass and associated flows which increasedagain after the centenary flood (Baeta et al 2011)This study showed that the estuarine food webs wereaffected differently by each disturbance event andsuggested that a high system throughput seems to beassociated with higher stress levels contradicting theidea that higher system activity is a sign of healthierconditions

Lindemanrsquos (1942) trophic-dynamic concepts andUlanowiczrsquos (1995) developments bring about an -other way to analyse the 6 case studies presented inBaeta et al (2011)

Advantages and limitations of the chosen numerical approach

This study was based on high quality models asrevealed by the large amount of local informationused Because the 6 models shared comparablestructures (differences among considered trophiccompartments were directly linked with differencesin biodiversity) the comparison could be based onindicators from network analysis shown to be sensi-tive to the structure of the food web model (Johnson

et al 2009) However to maintain the high reliabilityin the estimated flows we made the choice not toconsider birds and fish Even if these compartmentshad been described previously (eg Lopes et al 2005Martinho et al 2007) the information was not suffi-cient to exactly determine their role during eachstudied period and at each study site This has to betaken into account in our comparison In particular itreduces the possibility of adding other metrics to ourcomparison or comparing index values with thosederived from models of other estuarine ecosystemsthat usually consider a wider area and include birdsand fish (Christian et al 2005 Leguerrier et al 2007Heymans et al 2012)

Lindeman spine topology

The length of the Lindeman spines varied betweenhabitats (ie inter-habitat changes) and among yearswithin habitats (ie inter-annual differences) TheZostera bed chains were longer than those of thebare sediment site At first it may seem possible togeneralise that the trophic chain length is propor-tional to the number of compartments present in thefood web ie that more compartments will result in alonger chain However this trend is only verified insome case studies For example the food web of theZostera bed site in 2001 has 24 compartments but itschain is 1 step longer (number of trophic levels 8)than that of the bare sediment site in 1993 with 25compartments (number of trophic levels 7) Othervariables besides number of compartments probablycontribute to determining the length of the energychains Ulanowicz (1992) based on data taken from atidal marsh tributary creek of the Crystal River(Florida USA) and from a similar creek that was sub-ject to an average 6degC rise in ambient temperaturebecause of exposure to the effluent from an adjacentpower-generating station proposed that the lengthof the Lindeman spine is related to the level of stressie that shorter chains would reflect a higher degreeof stress According to Ulanowiczrsquos (1992) proposalthe bare sediment food webs with shorter chainswould reflect a higher level of stress in this habitatthan in the Zostera bed site If fact Z noltii hasadapted to the challenging conditions affecting inter-tidal habitats including environmental heterogene-ity and disturbance (Hemminga amp Duarte 2000) Thismacrophyte can act as a buffer to disruption essen-tially as a function of shelter offered by foliage andbecause this speciesrsquo root networks stabilise the sed-iment in which they grow

149

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Niquil et al Estuarine food web and disturbances 147

P2940

1152

1900II3717

6915

0000

3711

197800320 III

1245

1955

0000

1270

186400900 IV

0118

0193

0000

1235

017000870 V

000946

00157

0000

00976

0010200651 VI

0000570

0000884

0000

000587

000071300752 VII

0000039

0000065

0000

0000413

000005900898 VIII

0000

0000

0000

0000

D3206

5273

2990

2276 6147 0553 00496 000291 0000185 00005334

4279

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P3472

2009

1948II2843

7180

0000

3468

130400226 III

0691

1144

0000

8780

17060121 IV

00941

0176

0000

1232

013100682 V

000704

00134

0000

00924

00083800586 VI

0000440

0000810

0000

000576

000055200617 VII

0000023

0000045

0000

0000307

000002000417 VIII

0000

0000

0000

0000

D3606

5113

5110

2180 3565 0551 00423 000263 0000144 00007332

3825 Z93

P3104

6847

1894II3801

4333

0000

2741

107400235 III

0933

1087

0000

6895

14770132 IV

0128

0139

0000

0954

013500883 V

00109

00123

0000

00813

0010700816 VI

0000869

0000949

0000

000649

000084100805 VII

0000062

0000068

0000

0000461

000005600756 VIII

0000000

0000001

0000

0000004

D2987

5113

1837

1720 2840 0443 00393 000311 0000223 00000013590

2671

Z01

Z99

P

3368

6845

1005II

3141

3501

0000

1604

892000334 III

1104

0709

0000

5761

058500624 IV

00751

00423

0000

0403

00093400146 V

000121

0000549

0000

000674

000015300149 VI

0000014

0000006

0000

0000077

0000001000504 VII

0000

0000

0000

0000

D

3373

3050

1857

9758 3034 0226 000338 0000038 00002866

1662

TL

TST()

biomass

exports and catches

respiration

consumption predationTE

flow to detritus

flow to detritus

P

3674

7070

1914II

2907

2421

0000

2337

491900127 III

0406

0359

0000

3398

05870108 IV

00482

00436

0000

0418

0045100700 V

000304

000277

0000

00256

00014100348 VI

0000051

0000044

0000

0000420

000002200321 VII

0000

0000

0000

0000

D

3373

3058

2992

1495 1430 0180 00136 0000234 00004503

1964

P

3454

8528

1332II

3300

4725

0000

1690

586400209 III

0783

0447

0000

4341

052700790 IV

00736

00478

0000

0424

00092000147 V

000130

0000859

0000

000761

0000018000161 VI

0000001

0000000

0000

0000004

D

3161

3058

1611

1062 1803 0194 000346 00000022693

1475

bs93

bs01

bs99

Fig 3 Lindeman spines for each of the 3 Zostera meadow food webs (Z = Zostera bed 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Fig 4 Lindeman spines for each of the bare sediment area food webs (bs = bare sediment 93 =19931994 99 = 19992000 01 = 20012002) All flow values are in g AFDW mminus2 yrminus1 and all bio-mass values in g AFDW mminus2 P = primary producers D = detritus pool TL = trophic level TE =

trophic efficiency TST = total system throughput Roman numerals indicate trophic levels

Mar Ecol Prog Ser 512 141ndash154 2014

The Lindeman spines also illustrate the flows ofrespiration detritus production and export leavingeach integer trophic level In the present case onlyrespiration and detritus production were taken intoaccount export was estimated as null in the model(Baeta et al 2011) Because most of the heterotrophicbiomass was located at level II (97 to 99) respira-tion flows followed the biomass distribution closelywith higher values at level II

The Zostera meadows food webs have longerchains than the bare sediment food webs This obser-vation results from the higher number of compart-ments in the Zostera bed food webs which in turnreflects the higher species richness of this area com-pared with the bare sediment area The highesttrophic efficiency value was always found at level III(Fig 5) ranging between 6 (bare sediment in 2001)and 13 (Zostera bed in 1999)

In terms of mean trophic efficiency (Fig 6) the values presented by Zostera bed models were higher(57 to 65) than those of bare sediment area mod-els (29 to 46) The mitigation measures in 1999corresponded to an increase in mean trophic effi-ciency in the Zostera site compared with the eutroph-ication period (57 to 65) followed by a minutedecrease after the flood event In the food webs of thebare sediment area the mean trophic efficiency val-ues were higher in the eutrophic period (46) thanin the other 2 periods (29 to 31) Our observationssuggest that the implementation of mitigation meas-ures had a stronger impact on the mean trophic effi-ciencies than the extreme flood event Compared toAiCi which describes the specialisation of the flowswe observed that the trend of variation over time wascomparable with the mean trophic efficiency for thebare sediment site but both were in contradiction tothe Zostera site In both sites AiCi was lower in19992000 and returned to a higher value in20012002

DISCUSSION

The Mondego estuary monitoring programmebegan in the mid-1980s This ecological monitoringprogramme was developed because it is well knownthat many ecological processes occur through or aredriven by factors acting on long temporal and largespatial scales Many scientists have collected valu-able information about the estuary over the last 30 yrincluding data on physical conditions primary pro-ducers zooplankton benthic macrofauna fish andbirds (eg Marques et al 2003 2013 Patriacutecio et al2004 2009 Lopes et al 2005 Dolbeth et al 2007Martinho et al 2007 Cardoso et al 2010 Falcatildeo et al2012 Veriacutessimo et al 2012 Neto et al 2013) The

148

02

022

024

026

028

03

032

034

036

0

001

002

003

004

005

006

007

Z93 Z99 Z01 bs93 bs99 bs01

AiC

i

TEm

TEm (left axis)AiCi (right axis)

Fig 5 Trophic efficiencies per trophic level for the 6 models(bs = bare sediment Z = Zostera bed 93 = 19931994 99 =19992000 01 = 20012002) Trophic level efficiency of levelIII = predation on level III by level IV consumption of level

II by level III

Fig 6 Mean trophic efficiency TEm (geometric mean of val-ues for integer levels calculated for trophic levels II to IV)and relative internal ascendency AiCi for the 6 models (bs= bare sediment Z = Zostera bed 93 = 19931994 99 =

19992000 01 = 20012002)

Site No comp TEm () DH ratio

Z93 36 57 196Z99 31 65 141Z01 24 63 225bs93 25 46 103bs99 20 29 111bs01 20 31 165

Table 1 Characteristics of the 6 original Ecopath models(Baeta et al 2011) No comp = number of compartmentsTEm = mean trophic efficiency DH = DetritivoryHer-bivory Z = Zostera bed site bs = bare sediment site 93 =

19931994 99 = 19992000 01 = 20012002

Niquil et al Estuarine food web and disturbances

analysis of these data was the basis for a betterunderstanding of the structure and functioning ofthis estuarine ecosystem

In the early and mid-1990s nutrient enrichment inthe system led to macroalgal blooms that decreasedthe previously dominant macrophyte communities(ie Zostera noltii) In 1998 mitigation measureswere implemented in order to restore the Z noltiibeds and the overall quality of the system A few spe-cies responded positively (eg higher biomass ofScrobicularia plana and Hediste diversicolor) othersnegatively (eg lower biomass of Hydrobia ulvae)which led to more structured and stable populationscloser to the less disturbed conditions observed in the1980s (Patriacutecio et al 2009) According to Baeta et al(2011) the Z noltii site showed a higher number oftrophic compartments and a higher level of systemactivity (ie sum of consumptions respiration flowsto detritus production total system throughput netprimary production and system omnivory index) thanthe bare sediment site The differences in total sys-tem throughput between the 2 sites for the 3 periodswere mainly due to differences in the biomass of pri-mary producers (higher primary production at theZostera site) In both habitats after recovery meas-ures were implemented there was an increase in Splana and H diversicolor biomass consumption res-piration and flows to detritus and a decrease in Hulvae biomass and associated flows which increasedagain after the centenary flood (Baeta et al 2011)This study showed that the estuarine food webs wereaffected differently by each disturbance event andsuggested that a high system throughput seems to beassociated with higher stress levels contradicting theidea that higher system activity is a sign of healthierconditions

Lindemanrsquos (1942) trophic-dynamic concepts andUlanowiczrsquos (1995) developments bring about an -other way to analyse the 6 case studies presented inBaeta et al (2011)

Advantages and limitations of the chosen numerical approach

This study was based on high quality models asrevealed by the large amount of local informationused Because the 6 models shared comparablestructures (differences among considered trophiccompartments were directly linked with differencesin biodiversity) the comparison could be based onindicators from network analysis shown to be sensi-tive to the structure of the food web model (Johnson

et al 2009) However to maintain the high reliabilityin the estimated flows we made the choice not toconsider birds and fish Even if these compartmentshad been described previously (eg Lopes et al 2005Martinho et al 2007) the information was not suffi-cient to exactly determine their role during eachstudied period and at each study site This has to betaken into account in our comparison In particular itreduces the possibility of adding other metrics to ourcomparison or comparing index values with thosederived from models of other estuarine ecosystemsthat usually consider a wider area and include birdsand fish (Christian et al 2005 Leguerrier et al 2007Heymans et al 2012)

Lindeman spine topology

The length of the Lindeman spines varied betweenhabitats (ie inter-habitat changes) and among yearswithin habitats (ie inter-annual differences) TheZostera bed chains were longer than those of thebare sediment site At first it may seem possible togeneralise that the trophic chain length is propor-tional to the number of compartments present in thefood web ie that more compartments will result in alonger chain However this trend is only verified insome case studies For example the food web of theZostera bed site in 2001 has 24 compartments but itschain is 1 step longer (number of trophic levels 8)than that of the bare sediment site in 1993 with 25compartments (number of trophic levels 7) Othervariables besides number of compartments probablycontribute to determining the length of the energychains Ulanowicz (1992) based on data taken from atidal marsh tributary creek of the Crystal River(Florida USA) and from a similar creek that was sub-ject to an average 6degC rise in ambient temperaturebecause of exposure to the effluent from an adjacentpower-generating station proposed that the lengthof the Lindeman spine is related to the level of stressie that shorter chains would reflect a higher degreeof stress According to Ulanowiczrsquos (1992) proposalthe bare sediment food webs with shorter chainswould reflect a higher level of stress in this habitatthan in the Zostera bed site If fact Z noltii hasadapted to the challenging conditions affecting inter-tidal habitats including environmental heterogene-ity and disturbance (Hemminga amp Duarte 2000) Thismacrophyte can act as a buffer to disruption essen-tially as a function of shelter offered by foliage andbecause this speciesrsquo root networks stabilise the sed-iment in which they grow

149

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Mar Ecol Prog Ser 512 141ndash154 2014

The Lindeman spines also illustrate the flows ofrespiration detritus production and export leavingeach integer trophic level In the present case onlyrespiration and detritus production were taken intoaccount export was estimated as null in the model(Baeta et al 2011) Because most of the heterotrophicbiomass was located at level II (97 to 99) respira-tion flows followed the biomass distribution closelywith higher values at level II

The Zostera meadows food webs have longerchains than the bare sediment food webs This obser-vation results from the higher number of compart-ments in the Zostera bed food webs which in turnreflects the higher species richness of this area com-pared with the bare sediment area The highesttrophic efficiency value was always found at level III(Fig 5) ranging between 6 (bare sediment in 2001)and 13 (Zostera bed in 1999)

In terms of mean trophic efficiency (Fig 6) the values presented by Zostera bed models were higher(57 to 65) than those of bare sediment area mod-els (29 to 46) The mitigation measures in 1999corresponded to an increase in mean trophic effi-ciency in the Zostera site compared with the eutroph-ication period (57 to 65) followed by a minutedecrease after the flood event In the food webs of thebare sediment area the mean trophic efficiency val-ues were higher in the eutrophic period (46) thanin the other 2 periods (29 to 31) Our observationssuggest that the implementation of mitigation meas-ures had a stronger impact on the mean trophic effi-ciencies than the extreme flood event Compared toAiCi which describes the specialisation of the flowswe observed that the trend of variation over time wascomparable with the mean trophic efficiency for thebare sediment site but both were in contradiction tothe Zostera site In both sites AiCi was lower in19992000 and returned to a higher value in20012002

DISCUSSION

The Mondego estuary monitoring programmebegan in the mid-1980s This ecological monitoringprogramme was developed because it is well knownthat many ecological processes occur through or aredriven by factors acting on long temporal and largespatial scales Many scientists have collected valu-able information about the estuary over the last 30 yrincluding data on physical conditions primary pro-ducers zooplankton benthic macrofauna fish andbirds (eg Marques et al 2003 2013 Patriacutecio et al2004 2009 Lopes et al 2005 Dolbeth et al 2007Martinho et al 2007 Cardoso et al 2010 Falcatildeo et al2012 Veriacutessimo et al 2012 Neto et al 2013) The

148

02

022

024

026

028

03

032

034

036

0

001

002

003

004

005

006

007

Z93 Z99 Z01 bs93 bs99 bs01

AiC

i

TEm

TEm (left axis)AiCi (right axis)

Fig 5 Trophic efficiencies per trophic level for the 6 models(bs = bare sediment Z = Zostera bed 93 = 19931994 99 =19992000 01 = 20012002) Trophic level efficiency of levelIII = predation on level III by level IV consumption of level

II by level III

Fig 6 Mean trophic efficiency TEm (geometric mean of val-ues for integer levels calculated for trophic levels II to IV)and relative internal ascendency AiCi for the 6 models (bs= bare sediment Z = Zostera bed 93 = 19931994 99 =

19992000 01 = 20012002)

Site No comp TEm () DH ratio

Z93 36 57 196Z99 31 65 141Z01 24 63 225bs93 25 46 103bs99 20 29 111bs01 20 31 165

Table 1 Characteristics of the 6 original Ecopath models(Baeta et al 2011) No comp = number of compartmentsTEm = mean trophic efficiency DH = DetritivoryHer-bivory Z = Zostera bed site bs = bare sediment site 93 =

19931994 99 = 19992000 01 = 20012002

Niquil et al Estuarine food web and disturbances

analysis of these data was the basis for a betterunderstanding of the structure and functioning ofthis estuarine ecosystem

In the early and mid-1990s nutrient enrichment inthe system led to macroalgal blooms that decreasedthe previously dominant macrophyte communities(ie Zostera noltii) In 1998 mitigation measureswere implemented in order to restore the Z noltiibeds and the overall quality of the system A few spe-cies responded positively (eg higher biomass ofScrobicularia plana and Hediste diversicolor) othersnegatively (eg lower biomass of Hydrobia ulvae)which led to more structured and stable populationscloser to the less disturbed conditions observed in the1980s (Patriacutecio et al 2009) According to Baeta et al(2011) the Z noltii site showed a higher number oftrophic compartments and a higher level of systemactivity (ie sum of consumptions respiration flowsto detritus production total system throughput netprimary production and system omnivory index) thanthe bare sediment site The differences in total sys-tem throughput between the 2 sites for the 3 periodswere mainly due to differences in the biomass of pri-mary producers (higher primary production at theZostera site) In both habitats after recovery meas-ures were implemented there was an increase in Splana and H diversicolor biomass consumption res-piration and flows to detritus and a decrease in Hulvae biomass and associated flows which increasedagain after the centenary flood (Baeta et al 2011)This study showed that the estuarine food webs wereaffected differently by each disturbance event andsuggested that a high system throughput seems to beassociated with higher stress levels contradicting theidea that higher system activity is a sign of healthierconditions

Lindemanrsquos (1942) trophic-dynamic concepts andUlanowiczrsquos (1995) developments bring about an -other way to analyse the 6 case studies presented inBaeta et al (2011)

Advantages and limitations of the chosen numerical approach

This study was based on high quality models asrevealed by the large amount of local informationused Because the 6 models shared comparablestructures (differences among considered trophiccompartments were directly linked with differencesin biodiversity) the comparison could be based onindicators from network analysis shown to be sensi-tive to the structure of the food web model (Johnson

et al 2009) However to maintain the high reliabilityin the estimated flows we made the choice not toconsider birds and fish Even if these compartmentshad been described previously (eg Lopes et al 2005Martinho et al 2007) the information was not suffi-cient to exactly determine their role during eachstudied period and at each study site This has to betaken into account in our comparison In particular itreduces the possibility of adding other metrics to ourcomparison or comparing index values with thosederived from models of other estuarine ecosystemsthat usually consider a wider area and include birdsand fish (Christian et al 2005 Leguerrier et al 2007Heymans et al 2012)

Lindeman spine topology

The length of the Lindeman spines varied betweenhabitats (ie inter-habitat changes) and among yearswithin habitats (ie inter-annual differences) TheZostera bed chains were longer than those of thebare sediment site At first it may seem possible togeneralise that the trophic chain length is propor-tional to the number of compartments present in thefood web ie that more compartments will result in alonger chain However this trend is only verified insome case studies For example the food web of theZostera bed site in 2001 has 24 compartments but itschain is 1 step longer (number of trophic levels 8)than that of the bare sediment site in 1993 with 25compartments (number of trophic levels 7) Othervariables besides number of compartments probablycontribute to determining the length of the energychains Ulanowicz (1992) based on data taken from atidal marsh tributary creek of the Crystal River(Florida USA) and from a similar creek that was sub-ject to an average 6degC rise in ambient temperaturebecause of exposure to the effluent from an adjacentpower-generating station proposed that the lengthof the Lindeman spine is related to the level of stressie that shorter chains would reflect a higher degreeof stress According to Ulanowiczrsquos (1992) proposalthe bare sediment food webs with shorter chainswould reflect a higher level of stress in this habitatthan in the Zostera bed site If fact Z noltii hasadapted to the challenging conditions affecting inter-tidal habitats including environmental heterogene-ity and disturbance (Hemminga amp Duarte 2000) Thismacrophyte can act as a buffer to disruption essen-tially as a function of shelter offered by foliage andbecause this speciesrsquo root networks stabilise the sed-iment in which they grow

149

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Niquil et al Estuarine food web and disturbances

analysis of these data was the basis for a betterunderstanding of the structure and functioning ofthis estuarine ecosystem

In the early and mid-1990s nutrient enrichment inthe system led to macroalgal blooms that decreasedthe previously dominant macrophyte communities(ie Zostera noltii) In 1998 mitigation measureswere implemented in order to restore the Z noltiibeds and the overall quality of the system A few spe-cies responded positively (eg higher biomass ofScrobicularia plana and Hediste diversicolor) othersnegatively (eg lower biomass of Hydrobia ulvae)which led to more structured and stable populationscloser to the less disturbed conditions observed in the1980s (Patriacutecio et al 2009) According to Baeta et al(2011) the Z noltii site showed a higher number oftrophic compartments and a higher level of systemactivity (ie sum of consumptions respiration flowsto detritus production total system throughput netprimary production and system omnivory index) thanthe bare sediment site The differences in total sys-tem throughput between the 2 sites for the 3 periodswere mainly due to differences in the biomass of pri-mary producers (higher primary production at theZostera site) In both habitats after recovery meas-ures were implemented there was an increase in Splana and H diversicolor biomass consumption res-piration and flows to detritus and a decrease in Hulvae biomass and associated flows which increasedagain after the centenary flood (Baeta et al 2011)This study showed that the estuarine food webs wereaffected differently by each disturbance event andsuggested that a high system throughput seems to beassociated with higher stress levels contradicting theidea that higher system activity is a sign of healthierconditions

Lindemanrsquos (1942) trophic-dynamic concepts andUlanowiczrsquos (1995) developments bring about an -other way to analyse the 6 case studies presented inBaeta et al (2011)

Advantages and limitations of the chosen numerical approach

This study was based on high quality models asrevealed by the large amount of local informationused Because the 6 models shared comparablestructures (differences among considered trophiccompartments were directly linked with differencesin biodiversity) the comparison could be based onindicators from network analysis shown to be sensi-tive to the structure of the food web model (Johnson

et al 2009) However to maintain the high reliabilityin the estimated flows we made the choice not toconsider birds and fish Even if these compartmentshad been described previously (eg Lopes et al 2005Martinho et al 2007) the information was not suffi-cient to exactly determine their role during eachstudied period and at each study site This has to betaken into account in our comparison In particular itreduces the possibility of adding other metrics to ourcomparison or comparing index values with thosederived from models of other estuarine ecosystemsthat usually consider a wider area and include birdsand fish (Christian et al 2005 Leguerrier et al 2007Heymans et al 2012)

Lindeman spine topology

The length of the Lindeman spines varied betweenhabitats (ie inter-habitat changes) and among yearswithin habitats (ie inter-annual differences) TheZostera bed chains were longer than those of thebare sediment site At first it may seem possible togeneralise that the trophic chain length is propor-tional to the number of compartments present in thefood web ie that more compartments will result in alonger chain However this trend is only verified insome case studies For example the food web of theZostera bed site in 2001 has 24 compartments but itschain is 1 step longer (number of trophic levels 8)than that of the bare sediment site in 1993 with 25compartments (number of trophic levels 7) Othervariables besides number of compartments probablycontribute to determining the length of the energychains Ulanowicz (1992) based on data taken from atidal marsh tributary creek of the Crystal River(Florida USA) and from a similar creek that was sub-ject to an average 6degC rise in ambient temperaturebecause of exposure to the effluent from an adjacentpower-generating station proposed that the lengthof the Lindeman spine is related to the level of stressie that shorter chains would reflect a higher degreeof stress According to Ulanowiczrsquos (1992) proposalthe bare sediment food webs with shorter chainswould reflect a higher level of stress in this habitatthan in the Zostera bed site If fact Z noltii hasadapted to the challenging conditions affecting inter-tidal habitats including environmental heterogene-ity and disturbance (Hemminga amp Duarte 2000) Thismacrophyte can act as a buffer to disruption essen-tially as a function of shelter offered by foliage andbecause this speciesrsquo root networks stabilise the sed-iment in which they grow

149

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Mar Ecol Prog Ser 512 141ndash154 2014

Over time no clear trend in chain length variationwas noticeable within each habitat In the Zosterabed area the 3 food webs had the same chain lengthregardless of period and disturbance type In thebare sediment area the chain length decreased dur-ing and after implementation of the mitigation meas-ures which somehow contradicts Ulanowicz (1992)Further investigation is needed to clearly understandthe connection between chain length and level ofdisturbance

Detritivoryherbivory ratio

In general the food webs of the Zostera meadowstended to have higher DH ratio values than the baresediment site food webs (except for Site bs01) Withineach period the same result was noticed Our find-ings are not unexpected since macrophyte habitatsare characterised by high availability of detrituseither resulting from the direct decay of plants orbecause their roots constitute traps for detrital par -ticles In both cases huge amounts of detritus (andbacteria) are accessible for consumption by benthicorganisms

According to Ulanowicz (1992) systems with high -er levels of stress usually show lower DH valuesBased on this observation we could conclude thatthe Zostera meadows area is at a lower stress levelthan the bare sediment area However the resultswere less clear when we analysed the differencesamong periods within each habitat In both areasthere was an increase in the DH value after the cen-tury flood mainly driven by changes in the abun-dance and biomass of H ulvae (feeding on microphy-tobenthos and detritus) S plana Cyathura carinataand H diversicolor In this case after the extremeflood event in 2001 (a high-level disturbance) detri-tivory increased largely exceeding herbivory inwhichever habitat was considered Our findings arein line with those of Dame amp Christian (2007) whostudied 4 salt-marsh ponds during periods of lowstress high stress and post-disturbance respectivelyand observed an increase in detritivory during thehigh stress conditions These findings are in contrastto Ulanowiczrsquos (1992) proposition Therefore furthertests are necessary before being able to use the DHratio as an operational metric to quantify and inter-pret the impact of disturbances on the trophic stateand functioning of an ecosystem

Moreover as DH is mainly related to the diet ofspecies with high consumption (in this case studyspecies with consumption values above 50 g AFDW

mminus2 yrminus1) changes in community composition speciesbiomass and diet composition are particularly rele-vant The diet matrices used to build the 6 food websof the Mondego estuary were built (for several species) using isotopic ratios assessed from localsamples and analysed with a stable isotope mixingmodel (ISOSOURCE software Phillips amp Gregg2003) This analysis confirmed that the dominantspecies have dietary plasticity over both time andspace For instance the diet of H ulvae shifted froma main consumption of microphytobenthos in thebare sediment food web in 1993 (75 microphyto-benthos 25 detritus) to a diet dominated by detri-tus in the Zostera bed area (68 55 and 68 in 19931999 and 2001 respectively) These findings empha-sise the importance of building food webs usinginformation about local diets and it is interesting forthe robustness of the numerical approach to use mix-ing model results for constraining food-web models(Baeta et al 2011 Pacella et al 2013)

Trophic efficiency

This broad term which designates a measure thatcan be estimated from Lindeman spines has differentuses in the scientific literature Thus the lack of aunique and clear definition might raise serious diffi-culties when comparing trophic efficiencies amongdifferent studies Caution is therefore essential Themean trophic efficiency of an ecosystem is generallyconsidered to be the geometric average of the trophicefficiencies of all trophic levels (Baird amp Ulanowicz1989) For instance trophic level I considers only pri-mary producers (Miehls et al 2009) or becomes levelI+D when concatenating primary producers and de-tritus (eg Baird et al 2012) The primary productionincluded in the Ecopath routine is considered as netprimary production (the gross primary production mi-nus the respiration of the primary producer consid-ered) Thus the total value is considered to be avail-able for herbivory and the value of the respiration byprimary producers is not estimated For this reason itis not possible to calculate the trophic efficiency atlevel I when using Ecopath without having informa-tion about gross primary production or respiration atlevel I and the mean trophic efficiency has been con-sidered for all levels from II to the top (Coll et al2009) or from II to IV (Heymans et al 2012) Unfortu-nately this prevents any comparison between thetrophic efficiency values of our 6 models and thetrophic efficiency values of models that consideredtrophic level I or I+D (Baird et al 2004 2007 Duan et

150

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Niquil et al Estuarine food web and disturbances

al 2009 Miehls et al 2009) This is disappointingsince this first level seems to be highly sensitive topressures such as eutrophication (Christian et al2009) or invasion by non-indigenous species (Baird etal 2012) For the future it would be important to in-clude in Ecopath the possibility of estimating the res-piration of primary producers This would allow thetrophic efficiency at level I+D to be estimated andwould ensure a better comparability of trophic effi-ciencies at the system level

In general the lsquotrophic efficiencies per trophic levelrsquo(see Fig 6) at the Zostera bed site were higher thanthose of the bare sediment area For both habitats andfor the 3 periods the trophic efficiency at level II waslow and its value varied within a minor interval (13 to33) Interestingly the highest trophic efficiencyvalues were found at level III The trophic efficiencyper trophic level found for the 6 food webs of the Mon-dego estuary was different from the results found forthe Sylt-Roslashmoslash Bight ecosystem in the northern Wad-den Sea (Baird et al 2004 2007 2012) and for theNeuse River estuary in North Carolina USA (Christianet al 2009) These 2 ecosystems were characterisedby maximum values at level I (primary producers +detritus) and a monotonous decrease in trophic effi-ciency values with increasing trophic levels Further-more Duan et al (2009) studied the Pearl River est -uary China a coastal ecosystem that has beenover fished and has received a high level of combinedpollution since the 1980s The authors built 2 foodwebs one for 1981 and one for 1998 when the estuarywas heavily ex ploited In 1981 the highest trophic efficiency value was found at level III (118) like inthe Mondego estuary (ie with a similar pattern) butthat was not the case for the model of the year 1998where the lowest trophic efficiency was found at levelIII (56) and the other trophic levels showed valuesabove 11 This drop in the trophic efficiency of levelIII was linked to the strong increase in jellyfish bio-mass that was poorly consumed These studies showtherefore that the pattern of variation in trophic effi-ciency per trophic level varies between ecosystemsand disturbance levels

It is also worth noting that each habitat reacted dif-ferently to the type of disturbance acting upon thesystem In the Zostera area the mean trophic effi-ciency was higher after the implementation of themitigation measures while in the bare sediment areathe trophic efficiency decreased This might berelated to the changes in primary production thatoccurred in the area In fact in 1993 the bare sedi-ment area had an unusual input of energy as a conse-quence of the macroalgal blooms since the mitiga-

tion measures such an input has not been observedNo major changes in mean trophic efficiency wereseen after the centenary flood Analysing the scien-tific literature on the topic stress is usually associatedwith low system efficiency For example between1974 and 1989 the Venetian Lagoon experienced ahigh level of eutrophication (Libralato et al 2004)During this period the mean trophic efficiency of thelagoon was extremely low (ie 05 Libralato et al2004) Christian et al (2009) described a de crease inthe mean trophic efficiency of the Neuse River estu-ary after a severe episode of hypoxia Baird et al(2012) reported a decrease in mean trophic efficiencyin the Sylt-Roslashmoslash Bight ecosystem However this con-nection between higher stress levels and a decreasein mean trophic efficiency cannot be generalisedThere are cases where stress lowers the values of thetrophic efficiencies of some trophic levels but this iscompensated by an increase in the trophic efficiencyof another level resulting in an overall value that ishigher in disturbed situations (see for instance Duanet al 2009 Chrystal amp Scharler 2014) In ecosystemsaffected by fisheries similar trends were observed byColl et al (2009) who made the following hypothesislsquothe food web became more efficient with time andmay be due to less biomass and production in theecosystemrsquo (p 2098) The literature also pointed outan increase in the mean trophic efficiency in accor-dance with the meta-analysis (based on 53 models)provided by Heymans et al (2012) According toHeymans et al (2012) mean trophic efficiency in -creases with fishing pressure (no fishing low andhigh fishing pressure scenarios distinguished) andmore precisely the relationship between meantrophic efficiency and the L index seems very close tolinearity (Libralato et al 2008) Heymans et al (2012)also noticed that mean trophic efficiency depends onthe size longitude type and depth of the ecosystemHowever all these authors agree with Rochet ampTrenkel (2003) who present trophic efficiency as agood indicator for informing on ecosystem functionchanges and quantifying potential structural chan -ges Our findings analysing other disturbances (ieeutrophication hydrodynamic alterations and an ex -treme climatic event mdash a centenary flood) suggestthat the mean trophic efficiency of the system mightnot be sufficient to draw conclusions about the tro -phic efficiency of the ecosystem Rather it is essentialto take into account 2 levels of integration the overallefficiency and the efficiency of each trophic level Inorder to obtain functional indicators of ecosystemconditions it is worth comparing the properties ofmany of the coastal and marine food webs already

151

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Mar Ecol Prog Ser 512 141ndash154 2014

published However extra care should be taken tocope with differences in the type of ecosystemhabitat acting pressures levels of integration (egmean trophic efficiency and efficiencies per trophiclevel) and also the lack of standardised methodolo-gies between studies (eg number of compartmentsconsidered and aggregation level) It is essential tobe aware that the lack of straightforward compara-bility might be a severe constraint to extracting pat-terns and proposing generalisations

Finally comparison of the mean trophic efficiencywith the AiCi shows that the relationship betweenthe specialisation of internal flows and trophic effi-ciency (Monaco amp Ulanowicz 1997) is not direct Al -though the trend of variation was the same over timefor the bare sediment site the opposite trend wasobserved for the Zostera site According to theorieslinking ascendency to maturity (Ulanowicz 1997Ulanowicz et al 2009) the most stressful situationsfor the 2 habitats would be during the mitigationperiod This may be associated with the recent chan -ges in nutrient loading followed by a short period ofadaptation for the ecosystem

CONCLUSIONS

Our results indicated that none of the tested meas-ures (ie Lindeman spine length DH ratio meantrophic efficiency and trophic efficiency per trophiclevel) are yet ready to be used per se as operationalmetrics to quantify and interpret the impact of distur-bances on the trophic state of an ecosystem It seemsneither easy nor linear to establish a clear linkbetween structural characteristics such as biodiver-sity and trophic attributes of ecosystems Howeveralthough the comparability of the models was note-worthy the fact that only benthos was taken intoaccount precludes comparison with whole ecosystemmodels Such research needs to be pursued andwould require the development and application of aconceptual framework (with common and useful setsof indicators ie a multi-indices approach) integrat-ing a holistic approach based on as many observa-tions as possible (Cury et al 2005 Shin et al 2010 Fuet al 2012) According to Christian et al (2009) aframework combining ecological network analysisand functional aspects would result in an improvedapproach to ecosystem-based management even ifthere is no clear consensus on its implementation(Peterson et al 2000 Dame amp Christian 2006) How-ever despite the vast literature on new trophic attrib-utes derived from ecological network analysis (eg

Coll et al 2009 Pranovi amp Link 2009 Heymans et al2012) this issue needs further investigations and pro-cesses as recent works (such as Fu et al 2012) havepointed out the need for improvements to be carriedout for instance taking into account the sitesrsquo speci-ficity and associated local anthropogenic pressuresIn this framework indicators derived from networkanalysis are promising for defining ecosystem healthindicators They allow the whole ecosystem structureto be considered and to function in a context ofmatura tion and stress (Christensen 1995 Ulanowicz1997 Bondavalli et al 2006) They can also be re -lated to resilience (Christensen 1995 Heymans et al2007 Lobry et al 2008) and they allow the detectionof non-linear major changes (ie abrupt shifts) at theecosystem level (Savenkoff et al 2007 Tomczak etal 2013)

Acknowledgements This study is an introduction to theGIP-Seine Aval project ANTROPOSEINE It was partially(bibliographic analysis) funded by DEVOTES (DEVelop-ment Of innovative Tools for understanding marine bio -diversity and assessing good Environmental Status) fundedby the European Union under the 7th Framework Pro-gramme lsquoThe Ocean for Tomorrowrsquo Theme (grant agree-ment no 308392) wwwdevotes-projecteu This work wasalso supported by the Portuguese National Board of Scien-tific Research (FCT) with a grant (SFRHBPD744472010)We thank Benoit Lebreton (Lienss La Rochelle) for helpfuldiscussions Axel Rossberg and 2 anonymous referees whoprovided valuable criticisms and helpful comments in thefinal stages of the paper

LITERATURE CITED

Baeta A Pinto R Valiela I Richard P Niquil N Marques JC(2009a) δ15N and δ13C in the Mondego estuary food web seasonal variation in producers and consumers MarEnviron Res 67 109minus116

Baeta A Valiela I Rossi F Pinto R Richard P Niquil N Mar-ques JC (2009b) Eutrophication and trophic structure inresponse to the presence of the eelgrass Zostera noltiiMar Biol 156 2107minus2120

Baeta A Niquil N Marques JC Patriacutecio J (2011) Modellingthe effects of eutrophication mitigation measures and anextreme flood event on estuarine benthic food webs EcolModell 222 1209minus1221

Baird D Ulanowicz RE (1989) The seasonal dynamics of theChesapeake Bay ecosystem Ecol Monogr 59 329minus364

Baird D Asmus H Asmus R (2004) Energy flow of a borealintertidal ecosystem the Sylt-Roslashmoslash Bight Mar Ecol ProgSer 279 45minus61

Baird D Asmus H Asmus R (2007) Trophic dynamics ofeight intertidal communities of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea Mar Ecol Prog Ser 351 25minus41

Baird D Asmus H Asmus R (2012) Effect of invasive specieson the structure and function of the Sylt-Roslashmoslash Bight eco-system northern Wadden Sea over three time periods

152

Niquil et al Estuarine food web and disturbances

Mar Ecol Prog Ser 462 143minus161Bondavalli C Bodini A Rossetti G Allesina S (2006) Detect-

ing stress at the whole-ecosystem level the case of amountain lake (Lake Santo Italy) Ecosystems 9 768minus787

Cardoso PG Leston S Grilo TF Bordalo MD Crespo D Raf-faelli D Pardal MA (2010) Implications of nutrient de -cline in the seagrass ecosystem success Mar Pollut Bull60 601minus608

Christensen V (1995) Ecosystem maturity mdash towards quan-tification Ecol Modell 77 3minus32

Christensen V Pauly D (1992) A guide to the Ecopath II software system (version 21) ICLARM Software Manila

Christensen V Pauly D (1993) Flow characteristics of aqua -tic ecosystems In Christensen V Pauly D (eds) Trophicmodels of aquatic ecosystems ICLARM Conf Proc 26 338minus352

Christensen V Walters CJ (2004) Ecopath with Ecosim methods capabilities and limitations Ecol Modell 172 109minus139

Christian RR Baird D Luczkovich J Johnson JC ScharlerUM Ulanowicz RE (2005) Role of network analysis incomparative ecosystem ecology of estuaries In BelgranoA Scharler UM Dunne J Ulanowicz RE (eds) Aquaticfood webs Oxford University Press Oxford p 25minus40

Christian RR Brinson MM Dame JK Johnson G PetersonCH Baird D (2009) Ecological network analyses andtheir use for establishing reference domain in functionalassessment of an estuary Ecol Modell 220 3113minus3122

Chrystal RA Scharler UM (2014) Network analysis indicesreflect extreme hydrodynamic conditions during anextended dry period in a shallow estuarine lake (Lake StLucia) South Africa Ecol Indic 38 130minus140

Coll M Palomera I Tudela S (2009) Decadal changes in aNW Mediterranean Sea food web in relation to fishingexploitation Ecol Modell 220 2088minus2102

Cook RE (1977) Raymond Lindeman and the trophic-dynamic concept in ecology Science 198 22minus26

Cousins SH (1985) Ecologists build pyramids again New Sci106 50minus54

Cury PM Shannon LJ Roux JP Daskalov GM Jarre AMoloney CL Pauly D (2005) Trophodynamic indicatorsfor an ecosystem approach to fisheries ICES J Mar Sci62 430minus442

Dame JK Christian RR (2006) Uncertainty and the use ofnetwork analysis for ecosystem-based fishery manage-ment Fisheries 31 331minus341

Dame JK Christian RR (2007) A statistical test of networkanalysis Can it detect differences in food web proper-ties Ecosystems 10 906minus923

Dolbeth M Cardoso PG Ferreira SM Verdelhos T RaffaelliD Pardal MA (2007) Anthropogenic estuarine commu-nity over a 10-year period Mar Pollut Bull 54 576minus585

Duan LJ Li SY Liu Y Moreau J Christensen V (2009) Mod-eling changes in the coastal ecosystem of the Pearl RiverEstuary from 1981 to 1998 Ecol Modell 220 2802minus2818

Elliott M Quintino V (2007) The estuarine quality paradoxenvironmental homeostasis and the difficulty of detect-ing anthropogenic stress in naturally stressed areas MarPollut Bull 54 640minus645

Falcatildeo J Marque SC Pardal MA Marques JC Primo ALAzeiteiro UM (2012) Mesozooplankton structural res -ponses in a shallow temperate estuary following restora-tion measures Estuar Coast Shelf Sci 112 23minus30

Fiscus DA (2009) Comparative network analysis towardcharacterization of systemic organization for human-

environmental sustainability Ecol Modell 220 3123minus3132

Fu C Gaichas S Link JS Bundy A and others (2012) Rela-tive importance of fisheries trophodynamic and environ-mental drivers in a series of marine ecosystems Mar EcolProg Ser 459 169minus184

Gascuel D (2005) The trophic-level based model a theore -tical approach of fishing effects on marine ecosystemsEcol Modell 189 315minus332

Halpern BS Walbridge S Selkoe KA Kappel CV and others(2008) A global map of human impact on marine eco -systems Science 319 948minus952

Hecky RE (2006) Ecosystem health of large lakes identify-ing utility and metrics A proposal for a new theme areafor the fishery research program of the Great Lakes Fish-ery Commission University of Waterloo Waterloo

Hemminga MA Duarte CM (2000) Seagrass ecology 1stedn Cambridge University Press Cambridge

Heymans JJ Gueacutenette S Christensen V (2007) Evaluatingnetwork analysis indicators of ecosystem status in theGulf of Alaska Ecosystems 10 488minus502

Heymans JJ Coll M Christensen V Libralato S (2012) Eco-path theory modelling and application to coastal ecosys-tems In Baird D Mehta A (eds) Treatise on estuarineand coastal science Vol 9 estuarine and coastal ecosys-tem modelling Elsevier Oxford p 93minus113

Hutchinson GE (1979) The kindly fruits of the earth recol-lections of an embryo-ecologist Yale University PressNew Haven CT

Jennings S Greenstreet SPR Hill L Piet GJ Pinnegar JKWarr KL (2002) Long-term trends in the trophic structureof the North Sea fish community evidence from stable-isotope analysis size-spectra and community metricsMar Biol 141 1085minus1097

Johnson GA Niquil N Asmus H Bacher C Asmus R BairdD (2009) The effects of aggregation on the performanceof the inverse method and indicators of network analysisEcol Modell 220 3448minus3464

Leguerrier D Degreacute D Niquil N (2007) Network analysisand inter-ecosystem comparison of two intertidal mudflatfood webs (Brouage Mudflat and Aiguillon Cove SWFrance) Estuar Coast Shelf Sci 74 403minus418

Leston S Lilleboslash AI Pardal MA (2008) The response of pri-mary producer assemblages to mitigation measures toreduce eutrophication in a temperate estuary EstuarCoast Shelf Sci 77 688minus696

Libralato S Pranovi F Raicevich S Da Ponte F and others(2004) Ecological stages of the Venice Lagoon analysedusing landing time series data J Mar Syst 51 331minus344

Libralato S Christensen V Pauly D (2006) A method foridentifying keystone species in food web models EcolModell 195 153minus171

Libralato S Coll M Tudela S Palomera I Pranovi F (2008)Novel index for quantification of ecosystem effects offishing as removal of secondary production Mar EcolProg Ser 355 107minus129

Lindeman RL (1942) The trophic-dynamic aspect of ecologyEcology 23 399minus418

Lobry J David V Pasquaud S Lepage M Sautour BRochard E (2008) Diversity and stability of an estuarinetrophic network Mar Ecol Prog Ser 358 13minus25

Lopes RJ Muacuterias T Cabral JA Marques JC (2005) A tenyear study of variation trends and seasonality of a shore-bird community in the Mondego estuary PortugalWaterbirds 28 8minus18

153

Mar Ecol Prog Ser 512 141ndash154 2014

Marques JC Pardal MA Nielsen SN Joslashrgensen SE (1997)Analysis of the properties of exergy and biodiversityalong an estuarine gradient of eutrophication Ecol Mod-ell 102 155minus167

Marques JC Nielsen SN Pardal MA Joslashrgensen SE (2003)Impact of eutrophication and river management within aframework of ecosystem theories Ecol Modell 166 147minus168

Marques JC Salas F Patriacutecio J Neto J Teixeira H (2009)Ecological indicators for coastal and estuarine environ-mental assessment a user guide WIT Press Southampton

Marques L Carriccedilo A Bessa F Gaspar R Neto J Patriacutecio J(2013) Response of intertidal macrobenthic communitiesand primary producers to mitigation measures in a tem-perate estuary Ecol Indic 25 10minus22

Martinho F Leitatildeo R Cabral H Neto JM Marques JCPardal MA (2007) The use of nursery areas by fish in atemperate estuary Portugal Hydrobiologia 587 281minus290

Miehls ALJ Mason DM Frank KA Krause AE Peacor SDTaylor WW (2009) Invasive species impacts on ecosystemstructure and function a comparison of Oneida LakeNew York USA before and after zebra mussel invasionEcol Modell 220 3194minus3209

Monaco ME Ulanowicz RE (1997) Comparative ecosystemtrophic structure of three US mid-Atlantic estuaries MarEcol Prog Ser 161 239minus254

Neto JM Barroso DV Barriacutea P (2013) Seagrass QualityIndex (SQI) a Water Framework Directive compliant toolfor the assessment of transitional and coastal intertidalareas Ecol Indic 30 130minus137

Niquil N Chaumillon E Johnson GA Bertin X and others(2012) The effect of physical drivers on ecosystem indicesderived from ecological network analysis comparisonacross estuarine ecosystems Estuar Coast Shelf Sci 108 132minus143

Odum EP (1969) The strategy of ecosystem developmentScience 164 262minus270

Odum EP (1985) Trends expected in stressed ecosystemsBioScience 35 419minus422

Odum WE Heald EJ (1975) The detritus-based food web ofan estuarine mangrove community In Cronin LE (ed)Estuarine research Academic Press New York NYp 265minus286

Pacella SR Lebreton L Richard P Phillips D DeWitt THNiquil N (2013) Incorporation of diet information derivedfrom Bayesian stable isotope mixing models into mass-balanced marine ecosystem models a case study fromthe Marennes-Oleacuteron Estuary France Ecol Modell 267 127minus137

Patriacutecio J Marques JC (2006) Mass balanced models of thefood web in three areas along a gradient of eutrophica-tion symptoms in the south arm of the Mondego estuary(Portugal) Ecol Modell 197 21minus34

Patriacutecio J Ulanowicz R Pardal MA Marques JC (2004)Ascendency as ecological indicator a case study on estu-arine pulse eutrophication Estuar Coast Shelf Sci 60 23minus35

Patriacutecio J Ulanowicz R Pardal MA Marques JC (2006)Ascendency as ecological indicator for environmentalquality assessment at the ecosystem level a case studyHydrobiologia 555 19minus30

Patriacutecio J Neto JM Teixeira H Salas F Marques JC (2009)The robustness of ecological indicators to detect long-term changes in the macrobenthos of estuarine systemsMar Environ Res 68 25minus36

Pauly D Christensen V (1995) Primary production requiredto sustain global fisheries Nature 374 255minus257

Peterson CH Summerson HC Thomson E Lenihan HS andothers (2000) Synthesis of linkages between benthic andfish communities as a key to protecting essential fishhabitat Bull Mar Sci 66 759minus774

Phillips DL Gregg JW (2003) Source partitioning using stable isotopes coping with too many sources Oecologia136 261minus269

Pranovi F Link JS (2009) Ecosystem exploitation andtrophodynamic indicators a comparison between theNorthern Adriatic Sea and Southern New England ProgOceanogr 81 149minus164

Rochet MJ Trenkel VM (2003) Which community indicatorscan measure the impact of fishing A review and propos-als Can J Fish Aquat Sci 60 86minus99

Rombouts I Beaugrand G Fizzala X Gaill F and others(2013) Food web indicators under the Marine StrategyFramework Directive From complexity to simplicityEcol Indic 29 246minus254

Savenkoff C Castonguay M Chabot D Hammill MOBourdages H Morissette L (2007) Changes in the north-ern Gulf of St Lawrence ecosystem estimated by inversemodelling Evidence of a fishery-induced regime shiftEstuar Coast Shelf Sci 73 711minus724

Shin YJ Bundy A Shannon LJ Simier M and others (2010)Can simple be useful and reliable Using ecological indi-cators to represent and compare the states of marine eco-systems ICES J Mar Sci 67 717minus731

Tomczak MT Heymans JJ Yletyinen J Niiranen S Otto SABlenckner T (2013) Ecological network indicators of eco-system status and change in the Baltic Sea PLoS ONE 8 e75439

Ulanowicz RE (1992) Ecosystem health and trophic flow net-works In Costanza R Norton BG Haskell BD (eds) Eco-system health new goals for environmental manage-ment Island Press Washington DC p 190minus225

Ulanowicz RE (1995) Ecosystem trophic foundations Linde-man exonerata In Patten BC Jorgensen S (eds) Com-plex ecology the part-whole relation in ecosystemsPrentice-Hall Upper Saddle River NJ p 549minus560

Ulanowicz RE (1997) Ecology the ascendent perspectiveColombia University Press New York NY

Ulanowicz RE Puccia CJ (1990) Mixed trophic impacts inecosystems Coenoses 5 7minus16

Ulanowicz RE Goerner SJ Lietaer B Gomez R (2009) Quan-tifying sustainability resilience efficiency and the returnof information theory Ecol Complex 6 27minus36

Veriacutessimo H Bremner J Garcia C Patriacutecio J van der LindenP Marques JC (2012) Assessment of the subtidal macro-benthic community functioning of a temperate estuaryfollowing environmental restoration Ecol Indic 23 312minus322

Xu W Mage JA (2001) A review of concepts and criteria forassessing agroecosystem health including a preliminarycase study of southern Ontario Agric Ecosyst Environ 83 215minus233

154

Submitted November 12 2013 Accepted May 22 2014 Proofs received from author(s) August 19 2014