The role of autotrophic production in different coastal habitats in the production of fish in estuaries is an important consideration in coastal management and conservation. In the estuarine waters of the Australian east coast, many economically important fish species occur over mudflats lacking conspicuous vegetation. I used stable isotope analysis to examine where such fish ultimately derived their nutrition, in the subtropical waters of southern Moreton Bay, Queensland, Australia. I first tested traditional processing methodologies of autotroph samples, in this case of mangrove leaves, and examined variability in mangrove isotope values at different spatial scales. Mangrove leaves processed using time-consuming grinding showed no significant difference in isotope values than coarsely broken leaf fragments. Isotope values of green leaves were not meaningfully different from yellow or brown leaves that would normally be the leaves that actually dropped on to the sediment. Future analyses therefore can use green leaves, since they are more abundant and therefore more easily collected, and can simply be processed as whole leaf fragments rather than being ground to a powder. Carbon and nitrogen isotope values varied at several spatial scales. The proportion of variability partitioned at different scales varied depending on the species of mangrove and element (C or N) analysed. To properly represent a geographic area, isotope analysis should be done on leaves collected at different locations and, especially, from different trees within locations. The autotrophic source(s) supporting food webs leading to fish production on mudflats might be either in situ microphytobenthos or material transported from adjacent habitats dominated by macrophytes. I tested the importance of these sources by measuring ?13C values of 22 fish species and six autotroph taxa (microphytobenthos on mudflats, and seagrass, seagrass epiphytic algae, mangroves, saltmarsh succulents and saltmarsh grass in adjacent habitats) in Moreton Bay. I calculated the distribution of feasible contributions of each autotroph to fishes. All fish ?13C values lay in the enriched half of the range for autotrophs. For over 90% of fishes, the top three contributing autotrophs were seagrass, epiphytes and saltmarsh grass, with median estimates of approximately 60-90% from these sources combined. Seagrass was typically ranked as the main contributor based on medians, while epiphytic algae stood out based on 75th percentile contributions. The other three sources, including MPB, were ranked in the top three contributors for only a single fish. Organic matter from seagrass meadows is clearly important at the base of food webs for fish on adjacent unvegetated mudflats, either through outwelling of particular organic matter or via a series of predator-prey interactions (trophic relay). Modelling results indicate that saltmarsh grass (Sporobolus) also had high contributions for many fish species, but this is probably a spurious result, reflecting the similarity in isotope values of this autotroph to seagrass. Carbon from adjacent habitats and not in situ microphytobenthos dominates the nutrition for this suite of 22 fishes caught over mudflats. The ultimate autotrophic sources supporting production of three commercially important fish species from Moreton Bay were re-examined by further analysing carbon and nitrogen stable isotope data. Mean isotope values over the whole estuary for fish and autotroph sources were again modelled to indicate feasible combinations of sources. Variability in isotope values among nine locations (separated by 3-10 km) was then used as a further test of the likelihood that sources were involved in fish nutrition. A positive spatial correlation between isotope values of a fish species and an autotroph indicates a substantial contribution from the autotroph. Spatial correlations were tested with a newly developed randomisation procedure using differences between fish and autotroph values at each location, based on carbon and nitrogen isotopes combined in two-dimensional space. Both whole estuary modelling and spatial analysis showed that seagrass, epiphytic algae and particulate organic matter in the water column, potentially including phytoplankton, are likely contributors to bream (Acanthopagrus australis) nutrition. However, spatial analysis also showed that mangroves were involved (up to 33% contribution), despite a very low contribution based on whole estuary modelling. Spatial analysis for sand whiting (Sillago ciliata) demonstrated the importance of two sources, mangroves and microalgae on the mudflats, considered unimportant based on whole estuary modelling. No spatial correlations were found between winter whiting (Sillago maculata) and autotrophs, either because fish moved among locations or relied on different autotrophs at different locations. Spatial correlations between consumer and source isotope values provide a useful analytical tool for identifying the role of autotrophs in foodwebs, and were used here to demonstrate that organic matter from adjacent habitats, and in some cases also in situ production of microalgae, were important to fish over mudflats. Whilst recognising that production from several habitats is implicated in the nutrition of fishes over mudflats in Moreton Bay, clearly the major source is from seagrass meadows. Organic matter deriving from seagrass itself and/or algae epiphytic on seagrass is the most important source at the base of fisheries food webs in Moreton Bay. The importance of seagrass and its epiphytic algae to production of fisheries species in Moreton Bay reinforces the need to conserve and protect seagrass meadows from adverse anthropogenic influences.

The following four broad aims were addressed in this study. (1) To ascertain whether the characteristics of the benthic macroinvertebrate assemblages within the different nearshore marine habitat types identified by Valesini et al. (2003) on the lower west coast of Australia differ significantly, and whether the pattern of those spatial differences matches those among the environmental characteristics that were used to distinguish those habitat types; (2) To develop a quantitative approach for classifying nearshore habitats in estuarine waters that employs readily-available data for a range of enduring environmental characteristics, and to use that approach to classify the various habitat types present in nearshore waters of the Swan-Canning Estuary on the lower west coast of Australia; (3) To test the hypothesis that the characteristics of the benthic macroinvertebrate assemblages in the in the Swan-Canning Estuary differ significantly among nearshore habitat types, and that the pattern of those differences matches that among the environmental characteristics used to distinguish those habitat types and (4) To test the hypothesis that, as a result of environmental changes in the Swan-Canning Estuary, the characteristics of the benthic macroinvertebrate assemblages at various habitats in this estuary in 1986/7 differ from those in 2003/4. To address the first aim, benthic macroinvertebrates were sampled seasonally for one year in the subtidal waters and intertidal zone (upper and lower swash zones) at the six nearshore habitat types that were identified by Valesini et al. (2003) on the lower west coast of Australia. The habitat types, which differed mainly in the extent of their exposure to wave activity and whether seagrass and/or nearshore reefs were present, had been distinguished quantitatively using values for a suite of seven statistically-selected enduring environmental characteristics. The faunal samples yielded a total of 121 species representing eight phyla, among which the Polychaeta, Malacostraca and Bivalvia were the most speciose classes and contributed ~ 38, 23 and 10%, respectively, to the total number of individuals. The total number of species and mean density of macroinvertebrates was far greater at the most protected habitat type (1), which also contained dense beds of seagrass, than at any other habitat type, i.e. 70 species and 209.2 individuals 0.1 m-2, compared to 32 species and 36.9 individuals 0.1 m-2 at the most exposed habitat type (6), which had a substrate comprised only of sand. Differences among habitat type influenced the benthic macroinvertebrate species composition to a greater extent than differences among either zones or seasons. Significantly different faunal compositions were detected among those latter two factors only at the most protected habitat type. The faunal assemblage at habitat type 1 was clearly the most distinct from those at the other five habitat types, particularly in the subtidal zone (R-statistics=0.642-0.831, p=0.1%), and was typified by five abundant polychaete species that were adapted to deposit-feeding. In contrast, the fauna at habitat type 6 was typified by four crustacean species and a species of bivalve and polychaete, whose mobility and tough external surface facilitated their survival and feeding in those turbulent waters. The extents of the differences in species composition among the six habitat types was significantly matched with that among the suite of enduring environmental characteristics that distinguished those habitat types, particularly in the case of the subtidal zone (Rho=0.676). Such results indicated that the environmental variables used to distinguish the nearshore habitat types could be used to reliably predict the types of benthic macroinvertebrate species likely to occur at any site along the lower west coast of Australia. The above biological validation of the nearshore marine habitat classification scheme developed by Valesini et al. (2003) provided the justification for the approach to the second broad aim of this study, namely to develop a quantitative scheme for classifying habitat types in the Swan-Canning Estuary. This approach was similar to that employed by Valesini et al. (2003) in that it considers that differences among habitat types are well reflected by differences in a suite of enduring environmental variables. However, it improves on that earlier method by employing a completely objective and quantitative approach. Thus, a large number of environmentally-diverse nearshore sites (102) were initially selected throughout the Swan-Canning Estuary and a suite of 13 enduring environmental variables quantified at each using remotely-sensed images of the estuary in a Geographic Information System. Such variables were chosen to reflect either (i) the type of substrate and submerged vegetation present, (ii) the extent of exposure to wave action or (iii) the location of the site within the estuary with respect to its vicinity to marine and fresh water sources. These data were then subjected to the CLUSTER routine and associated SIMPROF procedure in the PRIMER v6 multivariate statistical package to quantitatively identify those groups of sites that did not differ significantly in their environmental characteristics, and thus represented habitat types. Eighteen habitat types were identified, which were shown to well reflect spatial differences in a suite of non-enduring water quality and sediment characteristics that were measured in situ at a range of estuarine sites during both summer and winter in 2005 (Rho=0.683 and 0.740, respectively, p=0.1%). However, those latter environmental characteristics required far more time in the field and laboratory to quantify than the enduring variables used to identify the habitat types. Benthic macroinvertebrates were sampled during summer and winter in 2005 in the shallow subtidal regions (~1 m depth) at sites representing eight of the habitat types identified in the Swan-Canning Estuary. These samples contained a total of 51 and 36 species during summer and winter, respectively, and, in both seasons, represented nine phyla, namely Annelida, Crustacea, Mollusca, Sipuncula, Nematoda, Platyhelminthes, Cnidaria, Uniramia and Nemertea. The compositions of the benthic macroinvertebrate assemblages differed significantly among habitat types and, to a similar extent, between seasons (Global R-statistic=0.408 and 0.409, respectively, p=0.1%). However, the spatial differences were considerable greater in winter than in summer (Global R-statistic=0.536 vs 0.280, p=0.1%), presumably due to the greater spatial variation in particular non-enduring in situ environmental characteristics, such as redox depth and salinity. While the number of species, overall density and taxonomic distinctness of benthic macroinvertebrates also differed significantly among habitats, those variables differed to a greater extent between seasons, being greater in winter than in summer. While the measures of taxonomic distinctness tended to be greater at habitat types located in the lower to middle reaches, i.e. habitat types 6, 7, 9, 10, 13 and 18, than the upper reaches i.e. habitat types 1 and 3, the number of species and overall density reflected this trend only during winter. During summer, the mean numbers of species at habitat types 1, 3, 6 and 10 (3.4-6.0) were significantly lower than those at habitat types 7, 13, and 18 (8.8-10.9), whereas the overall density of benthic macroinvertebrates was far greater at habitat type 7 (32260 individuals 0.1 m-2)than at any other habitat type in this season (3135-18552 individuals 0.1 m-2). Overall, the greatest differences in assemblage composition occurred between those at habitat types 1 and 18 (R-statistic=0.669, p=0.1%), which were located in the uppermost region of the estuary and the lower reaches of the basin, respectively, and differed to the greatest extent in their enduring environmental characteristics. The assemblage at habitat type 1, and also that at habitat type 3, located just downstream, were relatively distinct from those at all other habitat types, particularly during winter (R-statistics=0.666-0.993, p=0.1%). The fauna at the first of these habitat types was relatively depauperate, containing low numbers of species and densities, and was characterised by the polychaetes Leitoscoloplos normalis and Ceratonereis aequisetis and the bivalve Arthritica semen. The assemblage at habitat type 3 was also characterised by those three species and the amphipod Paracorophium minor and the polychaete Boccardiella limnicola. In contrast, the assemblage at habitat type 18 was characterised by a more diverse assemblage, i.e. the polychaetes Capitella capitata, C. aequisetis, L. normalis and Pseudopolydora kempi, the amphipods, Grandidierella propodentata and Corophium minor and the bivalve Sanguinolaria biradiata. The number of species was among the highest at this habitat type during both seasons, which was also reflected in the high taxonomic diversity, and the overall density was the highest in winter and second highest in summer. Despite the above faunal differences, those between assemblages at habitat types 7 and 9, which were both located in the basin of the Swan-Canning Estuary, were similar in magnitude to those that occurred between pairs of habitat types located in two different regions of the estuary. Although both habitat types 7 and 9 were characterised by a similar suite of species, i.e. Oligochaete spp., C. aequisetis, C. capitata, C. minor, G. propodentata, L. normalis, and S. biradiata, the substantial differences in assemblage composition between these habitat types in both summer and winter (R-statistics=0.570 and 0.725, respectively) was due to marked differences in the relative contributions of each of these species. Significant and strong correlations were shown to exist in both summer and winter between the pattern of differences in the benthic macroinvertebrate assemblages among habitat types and that among the enduring environmental characteristics used to identify those habitat types (Rho=0.625 and 0.825, respectively, p=0.1%). Furthermore, these correlations were greater than those obtained between the benthic macroinvertebrate fauna and any combination of the non-enduring environmental characteristics (i.e. water quality and sediment parameters) recorded in situ at each habitat type (Rho=0.508 and 0.824, in summer and winter, respectively, p=o.1%). This demonstrates the greater capacity of surrogate enduring environmental characteristics to account for differences in the range of variables that may influence the distribution of benthic invertebrate fauna. Thus, the lists of characteristic benthic macroinvertebrate taxa produced for each of the eight habitat types studied in the Swan-Canning Estuary provide a reliable benchmark by which to gauge any future changes in those fauna. Moreover, these results indicate that the above habitat classification scheme can be used to reliably predict the types of benthic macroinvertebrate fauna that are likely to occur at any nearshore site of interest in this estuarine system. The final component of this study showed that the benthic macroinvertebrate assemblages at four sites in the middle reaches of the Swan-Canning Estuary in 2003/4 differed significantly from those recorded at the same sites in 1986/7. Such differences were reflected in (1) changes in the relative densities of a suite of ten species that were responsible for distinguishing the faunas in these two periods, (2) the absence of 22 rare species in 2003/4 (i.e. 42% of the number of species recorded in 1986/7), (3) the presence of 17 new species in 2003/4, including an abundant polychaete that is likely to have been introduced and (4) a far greater extent of seasonal variation in the number of species and densities of benthic macroinvertebrates in 2003/4. Such changes are likely to be related to lower sediment oxygen levels in certain seasons in 2003/4, as well as an altered hydrological regime due to increased temperatures and decreased rainfall in that more recent period. The fact that these changes have occurred within the Swan-Canning Estuary highlights the need for effective management tools, such as the habitat classification scheme and associated faunal survey undertaken in this study. Such data will provide a sound basis by which to examine the ways in which fauna vary spatially within the system, and allow for the establishment of comprehensive benchmarks for detecting future changes.

Dégradés par plusieurs siècles d’endiguement et menacés par les effets du changement climatique, les marais intertidaux européens font aujourd’hui l’objet d’un nombre croissant de restaurations par rétablissement des marées sur d’anciennes zones endiguées, suivant une tendance générale appelée dépoldérisation. Cette thèse décrit la trajectoire de restauration des marais intertidaux à partir des communautés de necton (i.e., poissons et crustacés décapodes), sur la base de deux sites dépoldérisés accidentellement dans l’estuaire de la Gironde : le marais de Mortagne et l’île Nouvelle. L’évolution du necton dans les marais dépoldérisés est caractérisée à travers une approche synchronique à l’échelle de la mosaïque des habitats estuariens. La composition du necton dans les marais dépoldérisés est en grande partie similaire à celles des milieux intertidaux naturels et se différencie nette-ment des marais endigués, où les espèces exotiques d’eau douce prospèrent. Les marais restaurés apparaissent particulièrement attractifs pour le mulet porc, Liza ramada. La distribution des traits fonctionnels au sein des communautés de poissons révèle un changement de structure fonctionnelle au cours de la dépoldérisation, le temps de restauration étant inféré à partir d’un gradient de naturalité des habitats estuariens. En particulier, la spécialisation, la dispersion et la richesse fonctionnelle augmentent au cours de la restauration. Concernant l’évolution des règles d’assemblage, aucune tendance claire ne se dégage : les patrons diffèrent selon les traits fonctionnels.La dynamique temporelle de la composition des communautés de necton dans l’un des sites dépoldérisés montre une succession de trois phases au cours des quatre années suivant la restauration. Un changement plus abrupt concernant l’abondance de certaines espèces et la structure fonctionnelle des communautés est identifié au milieude la seconde phase, moins de deux ans après la dépoldérisation. La cinétique de transformation des communautés est probablement ralentie par le drainage incomplet du site dans les premiers temps de la dépoldérisation. Les résultats de cette thèse permettent de mieux comprendre la trajectoire de restauration des marais intertidaux, sans intervention humaine, tout en soulignant les atouts et les limites de certains modèles théoriques utilisés en écologie de la restauration. / European intertidal marshes have greatly declined during the past centuries because of land claim andhave recently been threatened by the effects of climate change. In recent years, an increasing number of tidal resto-ration projects have been implemented, following a global trend termed de-embankment. This thesis describes therestoration trajectory of tidally restored marshes from the response of nekton (i.e., fish and decapod crustaceans)based on two storm-breached study sites in the Gironde estuary : the Mortagne marsh and the Nouvelle island.Nekton assemblage composition of tidally restored marshes was described through a whole-estuary approach en-compassing natural, restored and dyked habitats. Nekton assemblages of tidally restored marshes showed highcompositional similarity with natural intertidal habitats and greatly departed from dyked marshes. Dyked marsheswere characterized by high occurrence of freshwater exotic taxa. Tidally restored marshes seemed particularly at-tractive for the thin-lipped grey mullet, Liza ramada. The functional traits’ distribution within fish communities revealeda change in functional structure in the course of restoration using the naturalness state of habitats as a proxy forrestoration time. Most notably, functional specialization, dispersion and richness increased with restoration time.No clear pattern of change was observed in community assembly over time ; trends chiefly varied with respect totraits. Temporal trends in one of the restored sites revealed three phases in the first four years of tide restoration.The abundance of several species and the community functional structure changed more steeply in the midst ofthe second phase, less than two years following tide restoration. Community turnover was probably delayed dueto incomplete drainage of the site in the early stages of restoration. The results of this thesis lead to a better un-derstanding of the restoration trajectory of tidal marshes without human intervention and highlight the strengths andweaknesses of some theoretical models used in restoration ecology.