Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
1. Estuarine wetlands are important nurseries for fish and decapod crustaceans. Flood mitigation structures (such as levees, culverts and floodgates) that fragment wetland habitat can reduce fish and crustacean passage and subsequently impact biodiversity.
2. Remediating structures to enhance connectivity, tidal flushing and fish and crustacean passage are assumed to be important ways to rehabilitate estuarine wetlands, but they are rarely evaluated with a robust sampling protocol. Furthermore, studies are inconsistently applied across different barrier types, and success is variable. Consequently, those rehabilitating wetlands are left with an incomplete understanding of what trajectories of change (if any) may be expected from barrier remediation.
3. In collaboration with landholders and managers, ‘floodgate remediation’ (structural and operational changes to increase tidal flushing and connectivity) was undertaken in three tidal creeks in two coastal river systems in northern New South Wales, Australia. Changes in fish and crustacean passage were measured for two different techniques (flap gates built into larger gates and the intermittent opening of gates with manual winching) using a BACI design over 2 years. Temporal changes in assemblages and species richness in managed creeks were compared to those in reference creeks (i.e. without floodgates) and control creeks (with closed floodgates).
4. Both types of floodgate remediation enhanced the passage of fish and crustaceans and had significant impacts on assemblages in managed creeks when compared to control and reference creeks. This shift was sustained for the duration of our study in two of the three creeks and was driven primarily by an increase in the number of estuarine–marine-dependent species.
5.Synthesis and applications. Our study demonstrates that floodgate remediation can facilitate fish and crustacean passage and rehabilitate aquatic assemblages in defaunated, tidally restricted wetlands. Given that the vast majority of floodgates throughout south-eastern Australia can be altered to promote connectivity, such remediation may play an important part in guarding against future declines in estuarine connectivity arising from climate change.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Estuarine wetlands can act as important nurseries for juvenile fishes and invertebrates (Beck et al. 2001; Sheaves 2005). Not only do wetland habitats enhance survival of juveniles (Boesch & Turner 1984; Kneib 1997), their roles in primary and secondary production, sediment retention and groundwater recharge are thought to have cascading effects that support greater fisheries production (Jordan, Smith & Nestlerode 2009). As such, a loss of connectivity between wetlands and the rest of an estuary can compromise ecological function (Meynecke 2009).
It is argued that restoring wetland connectivity is a valuable tool for addressing biodiversity losses in coastal environments (Thayer & Kentula 2005). This view is supported by studies where the removal of tidal restriction has been met with rapid and sustained recovery in wetland assemblages (e.g. Able et al. 2008). In other instances, however, the response can be less pronounced, possibly related to the extent of connectivity loss prior to rehabilitation vs. the extent improvement achieved (Raposa & Roman 2003; Eberhardt, Burdick & Dionne 2011). These aforementioned studies all focused on the remediation of culverts, and given the variability of response, further difference may be expected for other types of tidal restrictions, such as floodgates. Floodgates (i.e. top-hinged structures that open seaward on the ebb tide and shut against a culvert on a flooding tide) installed for flood mitigation are a predominant form of tidal restriction in coastal wetlands of south-eastern Australia (Williams & Watford 1997). The detrimental impact of floodgates on connectivity and fish and crustacean diversity has been acknowledged (Pollard & Hannan 1994; Kroon & Ansell 2006), but their complete removal is typically not an option because of competing land use and flood mitigation interests. However, it is believed that the vast majority of floodgates in south-eastern Australia could be structurally or operationally changed to enhance connectivity and possibly fauna passage (Williams & Watford 1997). The different ways of achieving this remain largely untested and warrant further investigation.
In this study, we examined responses of aquatic assemblages in tidal creeks in two coastal river systems in northern New South Wales (NSW) Australia to ‘floodgate remediation’ (structural or operational changes made to floodgates) intended to improve connectivity and tidal flushing. The relative responses of two different methods of floodgate remediation (small flap gates built into larger gates and the intermittent opening of gates with manual winching) were compared to determine whether these different approaches restored connectivity for fish and crustacean passage. If floodgate remediation was effective in this respect, we predicted the following. First, prior to floodgate remediation, assemblages in gated creeks would differ from those in reference (i.e. un-gated) creeks, with fewer juvenile fishes and crustaceans of species requiring access to estuarine–marine environments. Secondly, after opening floodgates, assemblages in managed (i.e. opened) creeks would come to resemble those in un-gated reference creeks, with an increase in juveniles of estuarine–marine species, whereas assemblages in gated control creeks will remain distinct. To test these predictions, we utilised an asymmetrical beyond BACI design (Underwood 1991) in the two river systems over a 2-year period and compared assemblages in managed creeks with those in multiple un-gated reference and gated control creeks before and after floodgate remediation began. As such, this study applied a robust sampling protocol, including relevant treatments and sufficient replicates across space and time, to test the efficacy of the management solution of interest (Memmott et al. 2010).
Materials and methods
The Macleay and Clarence River systems are two wave-dominated estuaries in NSW, Australia, with catchment areas of 11 385 and 22 400 km2, respectively (Roy et al. 2001). Both rivers contribute substantially to commercial and recreational fisheries in NSW (Saintilan 2004). Since European settlement in the 19th century, floodplains in northern NSW have been drained extensively through the construction of drainage and flood mitigation works, primarily for agricultural development (Tulau 1999). This includes the installation of floodgates to control surface run-off and flooding, and restrict saltwater inflow and tidal exchange to make surrounding land suitable for agriculture. In both rivers, the effects of floodgate opening were tested with the cooperation and assistance of landholders and managers.
The experimental design involved a comparison between one remediated or managed-gated (MG1) and three nearby tidal reference (R1-3) creeks with no floodgates (Fig. 1a; Table S1, Supporting information). Given the proximity of sampling creeks and their similar sizes and habitats, we assumed similar species assemblages at managed and reference creeks in the absence of floodgates. No control creeks where gates remained closed throughout our study were found; hence, our hypotheses for the Macleay River related only to changes from before to after opening in the MG creek relative to the reference creeks.
Samples were collected in each of 20 months from May 2007 until December 2008, on 10 occasions prior to and 14 occasions after floodgate remediation commenced (Fig. 2a; details below). The complete suite of temporal samples was used to plot trajectories of change in each creek through time (MG or R). For the statistical comparisons, we created a balanced design by excluding the four samples in the after period that were collected over summer (times 11–14, Fig. 2a), as no comparable summer samples were collected in the before period. The resultant four-factor model was (i) Before vs. After (BA) (fixed, two levels); (ii) Time (Ti) (random, 10 levels, nested within BA); (iii) Treatment (Tr) (fixed, two levels, MG vs. R); and (iv) Creek (Ck) (random, with one level nested in MG and three levels nested in R) (see below for detail).
An improved design was possible for the Clarence River, which involved a comparison between two managed creeks (MG2 and MG3), three tidal reference creeks (R4-R6) without any floodgates and two control creeks with gates that remained closed (C1 and C2) (Fig. 1b; Table S1, Supporting information). We hypothesised that managed creeks would come to resemble reference creeks following floodgate opening, whereas no such change would occur in closed creeks, and compared fish and crustacean samples taken from MGs, Rs and Cs at multiple times before and after floodgate openings.
Samples were collected every 2 months over a 21-month period (September 2000–May 2002), on six occasions prior to and five occasions after floodgate remediation commenced at both MGs (Fig. 2b). As in the Macleay, the complete suite of temporal samples was used to examine trajectories of change at each creek or treatment through time (MG, C or R), but only those collected in corresponding months (times 1–5 and 7–11, Fig. 2b) were used for statistical comparisons of before vs. after periods. The resultant asymmetrical, four-factor model was (i) BA (fixed, two levels); (ii) Ti (random, five levels, nested within BA); (iii) Tr (fixed, three levels, MGs vs. Rs vs. Cs); and (iv) Ck (random, two levels nested within each of MGs and Cs, three levels nested with Rs) (see below for detail).
Fish and crustaceans were collected during daylight hours using a fine mesh seine net (10 m headline × 2 m drop × 6 mm stretch mesh) pursed onto the shore [see Kroon & Ansell (2006) for details]. It was necessary to sample the tidal reference locations at different stages in the tidal cycle to ensure effective seine netting. In general, R1, R2, R4 and R6 were sampled around high slack time and R3 and R5 were sampled around slack low tide. Whilst tidal changes in fish assemblages have been recorded (e.g. Morrison et al. 2002), habitat effects on species assemblages generally tend to override any tidal effects (e.g. Ribeiro et al. 2006). Therefore, it was assumed that any impact of floodgate remediation on assemblages would override any potential differences attributed to sampling at different times of the tide. Three seine hauls (c. 10 m apart) were collected from two areas (50 or 150 m upstream of the floodgate or creek mouth) within each creek (Fig. 1). A pilot study indicated that three seine hauls in an area captured 86% of species present. The two areas were considered far enough apart that sampling at one would not influence sampling at the other. Data from the three seines per area were summed, resulting in two composite replicates per creek at each sampling occasion. Fish and crustaceans were identified to the lowest taxonomic level. For sampling in the Macleay River, not all crustacean taxa were kept, consequently only the commercial prawn species and all fish were used to test for effects of floodgate remediation in this system.
Measurements of pH, salinity and dissolved oxygen were taken at all areas according to the methods outlined in Kroon & Ansell (2006). In brief, this involved averaging the values across three surfaces (<30 cm) and three 1 m depth measurements taken adjacent to where seines were collected, using a hand held Horiba water quality meter.
Nonparametric permutational analyses of variance (Type III sum-of-squares) (permanova in the primer v6 software package, Anderson 2001; McArdle & Anderson 2001) were used to test for changes in assemblages in MGs over time relative to Rs or Cs. permanova is especially useful for mixed-model asymmetrical analyses because pseudo-F tests (analogous to traditional quasi-F ratios) can be created for both main effects and interactions in even the most complex designs using linear combinations of appropriate mean squares (Terlizzi et al. 2005). Because P-values are calculated using permutation tests, the P-values for all pseudo-F tests are perfectly correct (unlike quasi-F ratios; Anderson, Gorley & Clarke 2008).
Assemblage data were fourth-root transformed to reduce the influence of very abundant species (Clarke & Green 1988) before calculating Bray–Curtis similarities (Bray & Curtis 1957) using a dummy variable (Clarke, Somerfield & Chapman 2006). Nonmetric multidimensional scaling (nMDS; Kruskal & Wish 1978) ordinations were used to present multivariate patterns in the combined fish and crustacean assemblage data; here, we present plots of centroids for each MG creek along with centroids for the combined control or reference creeks at each time of sampling. Agglomerative hierarchical clustering was used to help interpret ordinations and to identify significant natural grouping of samples using similarity profile permutations (SIMPROF; Clarke, Somerfield & Gorley 2008) (with dendrograms presented in Fig. S1 of the Supporting Information).
To examine changes in the types of taxa based on their requirement to move between different parts of the estuary, the following functional groups were used: Estuarine–marine (E-M) are saltwater species that require access to either estuarine or ocean waters; Freshwater-estuarine (F-E) are euryhaline species that can occupy both freshwater or saltwater; Freshwater (F) species are those typically confined to freshwater tributaries of estuaries. The number of taxa in these different groups (and the total number of species) were analysed univariately using the permanova framework mentioned previously and Euclidean resemblance measures. This produces a test statistic equivalent to the F value of traditional anova, but the null distribution of the test statistic in permanova is produced by permutation, thus avoiding the usual normality assumptions of anova.
The SIMPER procedure (Clarke 1993) was used on presence/absence transformed data to identify those species that were most important (accounting for ≥3% dissimilarity; Terlizzi et al. 2005) in driving changes in species richness among treatments and times. For MG3, which showed only a short-term response to floodgate opening, the after period was analysed as two discrete periods (times 7–9 and 10–11) to better describe early vs. later changes. The consistency ratio (Dissimilarity/SD), calculated for all important species, indicated whether a species consistently contributed (values >1) to the average dissimilarity between managed and reference creeks in the majority of times in the before or after period, or only at certain times (Clarke 1993). Full SIMPER results are given in (Table S3, Supporting information).
Changes in water quality parameters were analysed univariately within the permanova design outlined previously. Dissolved oxygen was the only parameter excluded from analyses in the Clarence owing to the frequency of missing data. The BIOENV procedure (Clarke & Ainsworth 1993) was used to determine how well the patterns in water quality parameters correlated (Pearson’s) with the changing assemblage patterns.
A total of 1038 seines were hauled, yielding 75 taxa (61 fish and 14 decapod crustacean taxa) and 580 086 individuals (130 669 fish and 449 417 crustaceans) (Table S2, Supporting information). One-third of the taxa and 6% of all individuals were of economic importance, with Metapenaeus macleayi school prawn (Haswell, 1879), Liza argentea (Quoy and Gaimard, 1825) and Mugil cephalus (Linnaeus, 1758) (mullets), and Acanthopagrus australis yellow-finned bream (Owen, 1853) being the most abundant. In the Clarence, shrimp species comprised 97% of the total crustacean catch. The vast majority of individuals (86%) and species (65%) were classified as primarily estuarine–marine, followed by freshwater (15%, 12%), then freshwater-estuarine (12%, 2%).
In the Macleay River, the nMDS ordination shows similar species assemblages in the three reference creeks with small temporal changes in assemblage composition throughout our study (e.g. Fig. 3a). In contrast, MG1 showed a significant shift in assemblage composition from before to after the opening of the floodgates, which was not observed in the reference creeks over the same time period (Figs 3a and S1, Supporting Information). There was temporal variability in the magnitude of the change from before to after floodgate opening in MG1 [significant Ti(BA) × Ck(Tr) interaction, Table 1], due largely to assemblages sampled at one of the times after opening in MG1 (T16) being similar to some of the times before opening. Overall, however, assemblages at MG1 came to resemble those in the reference creeks after the floodgates were opened (Figs 3a and S1, Supporting Information), with the average dissimilarity between MG1 and Rs falling from 76·4 before opening to 37·6 in the after opening (SIMPER, Table S3, Supporting information).
Table 1. Multivariate permanova comparisons of fish and crustacean assemblages in the Macleay (44 taxa) and Clarence (66 taxa) river systems at various times (Ti) from before to after (BA) floodgate opening
Source of variation
Treatments in the Macleay are 1 managed vs. 3 reference creeks and the Clarence are 2 managed vs. 3 reference vs. 2 control creeks. P-values presented only for those terms that test for an effect of opening floodgates. A posteriori pairwise comparisons were used to investigate significant interactions (results in text).
†Pperm values were obtained using 999 permutations under a reduced model, with those in bold indicating significant sources of variation at α = 0·05.
BA × Tr
BA × Ck(Tr)
Ti(BA) × Tr
Ti(BA) × Ck(Tr)
BA × Tr
BA × Ck(Tr)
Ti(BA) × Tr
Ti(BA) × Ck(Tr)
The total number of species varied significantly between MG1 and the three reference creeks from before to after the opening of the floodgates (BA × Tr Pseudo-F =12·96, P =0·001). Specifically, in MG1, the number of species increased after the floodgates were opened to become similar to the numbers in the reference creeks, whilst there was no equivalent temporal change in the reference creeks (Fig. 4a). This pattern was driven primarily by estuarine–marine species (Fig. 4b, BA × Tr Pseudo-F =20·55, P =0·002), with a doubling of species within the 1-month sampling interval following floodgate opening and a further similar increase over the next 8 months (Fig. 5a). The estuarine–marine species responsible primarily for the change were M.macleayi and the fish Redigobius macrostoma (Günther, 1861), Ambassis spp., Favonigobius exquisitus (Whitley, 1950), Pandaka lidwilli (McCulloch, 1917), Gobiopterus semivestitus (Munro, 1949), L. argentea and A. australis (Table S2, SIMPER: Table S3, Supporting information).
In contrast, smaller changes in freshwater-estuarine and freshwater species were observed at MG1 in response to floodgate opening (Fig. 4c,d), and these were significant only at certain times in some creeks [for freshwater-estuarine species, Ti(BA) × Ck(Tr) Pseudo-F =1·54, P =0·047). The freshwater-estuarine fish species Pseudomugil signifer (Kner, 1865) and Philypnodon grandiceps (Krefft, 1864) increased in presence and abundance, whilst that of the freshwater fish Philypnodon macrostoma decreased (Table S2, SIMPER Table S3, Supporting information). The mean consistency ratio (Diss/SD) for those species that discriminated between MG1 and Rs prior to opening decreased from 1·4 to 0·8 after opening, indicating that species differences were consistent for the majority of times before floodgate opening, but less so following floodgate opening.
Salinity, dissolved oxygen and pH, collectively, were correlated with the assemblage patterns observed in the Macleay River (ρ = 0·282, P =0·01). There was no evidence for floodgate remediation influencing salinity or dissolved oxygen (Table S1, Supporting information; Salinity: BA × Tr Pseudo-F =1·52, P =0·24; DO: Pseudo-F =1·79, P =0·21). There was a small mean increase in pH (7·0–7·3) from before to after at MG1 compared to a decline in the Rs (Table S1, Supporting information; BA × Tr Pseudo-F =4·19, P =0·034), which is consistent with an effect of the floodgate remediation on pH.
In the Clarence, the nMDS ordinations show that species assemblages in the reference creeks changed over time, but, on average, did not differ from the before to after period. (Figs 3b and S1b, Supporting Information). Species assemblages did not differ among the reference creeks, except at three sampling occasions that were preceded by a large flood event in the Clarence River (Fig. 2; Kroon & Ludwig 2010). Similarly, species assemblages in the control creeks (C) did not change from before to after and were consistently different from assemblages in the reference creeks (Fig. 3b), except at Time 4 when one reference was similar to the controls (Fig. S1b, Supporting information). Prior to floodgate opening, species assemblages in the managed-gate creeks (MG2, 3) differed significantly from those in reference creeks, and MG3 also differed from control creeks [Ti(BA) × Ck(Tr) interaction, Table 1]. Species assemblages in both the managed creeks showed a clear shift from before to after, coming to resemble the reference (R) creeks soon after gate opening (Fig. 3b, Table 1). However, this response was more pronounced and sustained in MG2 than in MG3, with the assemblage in MG3 becoming even more dissimilar to Rs in the last two times (10–11) than before opening [average dissimilarity before (times 7–9 = 58·24, after times 10–11 = 63·42)] (Figs 3b and S1b, Supporting information). Thus, significant differences in species assemblages were detected only at particular times and creeks [Ti(BA) × Ck(Tr) interaction, Table 1].
Patterns in species numbers demonstrated conclusively that closed floodgates lead to fewer species upstream (Fig. 4a). Species richness in MG2 and MG3 increased significantly following floodgate opening and became more similar to the three reference creeks, with no such temporal change occurring in the two control creeks (Pseudo-F =4·58, P = 0·005 for BA × Tr). As in the Macleay, this pattern was driven by estuarine–marine species (Figs 4b and 5b), with crustacean species such as M. macleayi, Acetes sibogae australis (Colefax, 1940) and Macrobrachium cf. novaehollandiae (De Man, 1908), and fish species such as A. sibogae australis, Afurcagobius tamarensis (Johnston, 1883), Ambassis spp., L. argentea, G. semivestitus and M. cephalus and P. signifer increasing shortly following floodgate opening (Fig. 5b, Tables S4 and S5, Supporting information). The duration of this response by estuarine–marine species varied between MG2 and MG3 [Pseudo-F =4·12, P =0·001 for Ti(BA) × MG(Tr)], with the number of most estuarine–marine species at MG3 not being sustained and falling to levels equivalent to those at the two control creeks in the last two samples (times 10 and 11) (Fig. 5b). There was a nonsignificant trend for numbers of freshwater-estuarine species to increase from before to after floodgate opening relative to Cs and Rs, but no such pattern for freshwater species (Fig. 4c,d).
Salinity and pH collectively were weakly correlated with the assemblage patterns observed in the Clarence River (ρ = 0·083, P =0·05), but showed no significant response to floodgate remediation (Table S1, Supporting information). There was an overall increase in both salinity and pH from before to after across all creeks, but the pattern of change was not different between managed, reference and control creeks (Salinity: BA × Tr Pseudo-F =0·79, P =0·615; pH: Tr Pseudo-F =1·38, P =0·273).
Improved Connectivity and Reponses in Passage
Our results showed that restoring tidal connectivity triggered measurable shifts in assemblages in three managed creeks across two river systems. As predicted, assemblages in creeks with closed floodgates were significantly different from those in un-gated reference creeks, with 70–80% fewer estuarine–marine species than reference creeks. Following floodgate remediation, assemblages in managed creeks quickly changed to closely resemble those in reference creeks. In some instances, the difference in species richness was reduced to only 15%, showing significant and rapid recolonisation of estuarine–marine species, including juveniles of economically significant species such as A. sibogae australis, L. argentea, M. macleayi and M. cephalus. This shift in the assemblage was sustained for the duration of our study in two of the three creeks, whilst in the third managed creek, the shift was sustained only for three species of fish (possible explanations discussed below). The changes were clearly a result of floodgate remediation, as similar changes were not observed in (un-gated) reference or (gated) control creeks. This supports the notion that if a floodgate cannot be removed, it can be structurally or operationally modified to restore biotic passage (Kroon & Ansell 2006) and the nature of this response can be both rapid and sustained.
Species rapidly recolonised upstream areas following floodgate opening as a result of the restoration of physical passage. There was no evidence that vegetative habitats upstream of the floodgates changed over the 12-month period, and there were no ecologically significant changes in water quality in managed creeks relative to the controls or references. Although this latter result is contrary to what has been observed in other creeks upstream of floodgates (Pollard & Hannan 1994), it suggests that a certain amount of tidal exchange may have been occurring when the gates were closed. This ‘leakiness’ was observed at MG1 through the permeable levee that separates the creek from the main channel (K. Wilkinson, unpublished data). This degree of tidal flushing may have contributed to the speed with which fish and crustaceans could move into rehabilitated creeks once passage was enhanced at the floodgates.
Fish are extremely mobile, and it has been shown that they will quickly move into defaunated habitats as they become available (Peterson & Bayley 1993; Sheldon & Meffe 1995). Rapid responses to improved connectivity are most likely to be seen in those species that are strong dispersers and early colonisers (Hohausova, Lavoy & Allen 2010). For example, sparids are habitat opportunists (Miller & Skilleter 2006) and are key colonising species owing to their ability to cross relatively large expanses of sand where little protection from predation is found (Fernandez et al. 2007). In the present study, the sparid A. sibogae australis responded rapidly to floodgate opening. After early colonisation by some species, further development of the assemblage over time can be affected by postsettlement processes such as foraging, predator–prey interactions and competition (Poizat & Crivelli 1997). Predatory fish species are more likely to respond to larger aggregations of prey species (Stewart & Jones 2001), and in our study, the early colonisation of rehabilitated tidal creeks by prey species such as A. sibogae australis (shrimp) may have led to successional changes as predators that feed on these prey capitalise on the more abundant food source. This may partly explain the increase in species such as H. castelnaui, L. argentea, M. cephalus and A. sibogae australis observed in this study.
Spatial Variation in Response and the Effectiveness of Different Floodgate Remediation Approaches
In this study, the nature of response differed between the different remediation methods. Two of the three creeks showed rapid and sustained changes in assemblages once floodgate remediation began to increase connectivity. In the third creek (MG3), improved connectivity led to a rapid but less-pronounced and less-sustained response. Varied responses to coastal wetland rehabilitation are common (Simenstad, Reed & Ford 2006). In some instances, the responses to removing tidal restrictions can be rapid and sustained as observed at MG1 and MG2 (e.g. Morgan & Short 2002; Thom, Zeigler & Borde 2002), gradual and sustained (Able et al. 2008) or follow totally unexpected trajectories (Simenstad, Reed & Ford 2006). Such variability is likely to be a consequence of differences in the type of floodgate remediation used and differences among the creeks within the landscape. At MG1, a true test of the performance of automated, tide-actuated gates was not achieved as the automated flaps were removed by vandals for a significant part of the study. Nevertheless, passage at this creek was enhanced by the 500 × 500 mm openings in the middle portion of the floodgates, as it also was for intermittent manual winching (MG2).
The response to the manual operation of entire gates via a winch system in the Clarence differed between creeks, and responses in species richness were not as pronounced in MG3. One reason for this may be that the relative proximity of the remediated creeks to nearby habitats differed between creeks. Unlike MG1 and MG2, MG3 was not directly connected to the main river channel, but was located at the end of a large lake. Dispersal pathways therefore likely differed, being shallower and farther from the supply of recruits in the instance of MG3, two factors that have been shown to affect recolonisation rates (Beck et al. 2001; Hohausova, Lavoy & Allen 2010). Furthermore, the habitat upstream of MG3 appeared more degraded than in other gated creeks. Clearly, there will be a lag in any habitat improvement (if indeed it occurs) to improved water connectivity.
Structural or operational differences of the floodgates cannot be discounted as possible reasons for differing responses. The floodgates at MG3 consisted of pipe culverts in comparison with MG1 and MG2, which were box culverts. Fish passage is inversely related to flow velocity (Castro-Santos 2005), and pipe culverts with higher velocities than box culverts have been shown to have lower passage rates for a large range of species and size classes (Warren & Pardew 1998). There are an increasing number of floodgate styles and modifications available, and further research is required into their eco-hydraulic performance to better inform operational criteria (e.g. flap position and opening times around diurnal, lunar and seasonal cycles and in relation to river flooding) to maximise rehabilitation outcomes.
Temporal Variation of Responses
In this study, we observed temporal changes in the utilisation of estuarine habitats by fish and crustaceans and assemblage changes because of natural disturbances. To ensure that floodgate remediation is effective, processes relating to recolonisation and succession (previously discussed) need to be considered, as well as spatio-temporal variation in fish and crustacean distribution over larger scales. We observed that once opened, managed-gate creeks most resembled reference creeks during the spring and early summer months, with this similarity reducing in autumn and winter. Juvenile A. sibogae australis, L. argentea, M. cephalus and M. macleayi, and adult Ambassis spp. and A. sibogae australis entered managed creeks following opening of floodgates during spring and early summer. This period follows the peak offshore spawning periods (between May and August) for some of these species (e.g. A. sibogae australis and M. cephalus) in south-eastern Australia (Pollock, Weng & Morton 1983; Miller & Skilleter 2006). Floodgates are typically closed during these periods and when adults of ubiquitous species such as Ambassis spp. and A. sibogae australis reproduce (Henry 1977; Mitlon & Arthington 1985). The management implications are that floodgate opening should coincide with critical settlement periods to allow movement of species into coastal floodplain habitats; in northern NSW, this period is the low-rainfall winter and spring seasons.
It would appear from our study that floodgate remediation can induce improvements in assemblages that can withstand transient disturbances. The major floods, hypoxic conditions and associated fish kills experienced in both river systems during this study (Johnston et al. 2003; Walsh, Copeland & Waestlake 2004) are cases in point. Despite these floods leading to a rapid loss of species across all tidal creeks, assemblages recovered quickly and within a few months, the differences owing to creek connectivity and floodgate remediation were re-established. This supports observations from these floods previously by Kroon & Ludwig (2010) and demonstrates that estuarine assemblages can be resilient to episodic flooding and that positive responses to floodgate remediation can withstand such disturbances.
Implications for Coastal Floodplain Management and Further Research
Although the ecological impacts of tidal restrictions are well documented (Pollard & Hannan 1994; Raposa & Roman 2001) and different management approaches are recommended and sometimes implemented (e.g. Simenstad & Warren 2002 and references therein), often this is not in collaboration with applied practitioners nor accompanied by a sophisticated or substantial sampling protocol (Memmott et al. 2010). This results in lost learning and future management opportunities (Palmer et al. 2005). In our study, we showed that through collaborative efforts between landholders, conservation managers and local community members, the opening of floodgates can be managed to achieve improved connectivity and fish and crustacean passage. The replication of this across two river systems suggests that these outcomes may be achieved by managers in other river systems (Holl 2010).
Whilst improved access to habitats has been demonstrated in this study, it is the quality of reconnected habitat that will drive improvements in ecosystem function. Future efforts to quantify the contribution of tidal wetland rehabilitation to fisheries production will assist resource managers to compare environmental, social and economic interests in coastal environments. The link will partly depend on the nursery value (including measures of predation, growth, return movements and survival) of the reconnected habitat (Beck et al. 2001). Moreover, increased production will also depend on whether nursery habitats are limiting recruitment at larger scales. Because commercial fish catches are often correlated with estuarine habitat availability and connectivity (Blaber 2009; Turner 1992; Manson et al. 2005; Meynecke, Lee & Duke 2008), including those in NSW (Saintilan 2004), there is reason to believe that improved connectivity as demonstrated in this study may be beneficial for at least some species. Given that estuarine habitats are likely to become more fragmented in many parts of the world owing to climate change (Vinagre et al. 2011) and that over 30% of NSW tidal barriers (including 99% of all floodgates) are considered modifiable to enhance connectivity (Williams & Watford 1997), floodgate remediation may play an important role in guarding against future coastal habitat loss. To what extent this is the case warrants further study, with the findings having direct relevance to how managers prioritise rehabilitation projects in coastal environments.
We are grateful to all technicians and volunteers who assisted with field and laboratory work, the Clarence Floodplain Project, and all the landholders who allowed us access to their properties. The Clarence fieldwork was conducted when F.J.K. was employed by the former NSW Fisheries, and was funded by grant 1998/215 from the Australian Fisheries Research and Development Corporation. The Macleay component was funded by a Federal Government grant administered through the Southern Rivers CMA as part of the Bringing Back the Fish Project. Michael Lowry provided invaluable comments on the draft manuscript. Animals were collected in accordance with the appropriate animal care and ethics research authority (98/11) and a Section 37 research permit. The comments of two anonymous reviewers significantly improved the quality of this manuscript.