INTRODUCTION

Roads, dikes, levees, and other structures that block or restrict tidal flow to salt marshes are common throughout coastal New England and elsewhere. These tidal restrictions result in altered hydrology and biogeochemistry, and degraded vegetation, nekton, and bird communities compared to unrestricted salt marshes (Burdick et al., 1997; Eertman et al., 2002; Gray et al., 2002; Llanso, Bell, and Vose, 1998; Myshrall et al., 2000; Portnoy, 1991, 1999; Portnoy and Giblin, 1997; Raposa, 2002; Roman, Garvine, and Portnoy, 1995; Raposa and Roman, 2003; Roman, Niering, and Warren, 1984; Roman et al., 2002; Sinicrope et al., 1990; Vose and Bell, 1994; Warren et al., 2002). Salt marshes are integral to the proper functioning of estuarine systems, and in recognition of the negative effects of tidal restrictions, efforts are now underway to restore tidal flow and ecological function to many of these degraded systems. Unfortunately, salt marsh restoration projects are not always accompanied by comprehensive ecological monitoring, and monitoring programs that compare pre- and postrestoration conditions are even less frequent. For example, in a review of Gulf of Maine marsh restoration projects, it was found that basic environmental parameters (e.g., salinity and vegetation) were monitored at more than 78% of the sites, but more involved biological parameters such as nekton and birds were monitored at only about half of all sites (Konisky et al., 2006). Ecological monitoring is critical to understand the effects of each individual restoration project and improve our understanding of how these systems respond to restoration in general in order to better guide future restoration efforts. Indeed, a growing number of case studies demonstrate that restoring tidal flow to restricted marshes can return ecological functionality by improving water quality conditions, vegetation, and nekton and bird communities (Barrett and Niering, 1993; Burdick et al., 1997; Jivoff and Able, 2003; Raposa and Roman, 2003; Roman et al., 2002; Teo and Able, 2003; Thom, Zeigler, and Borde, 2002; Warren et al., 2002).

Potter Pond is a tide-restricted estuarine pond and salt marsh complex located on Prudence Island, Rhode Island, in the Narragansett Bay National Estuarine Research Reserve. Under tide-restricting conditions, tidal exchange with Narragansett Bay was negligible, low oxygen conditions were frequent, and the pond was virtually covered by excessive macroalgal growth. Tidal exchange was returned to this site by replacing a pair of crushed culverts with a new, larger culvert in April 2003. A second crushed culvert at the back end of the pond was also replaced to return estuarine tidal flow to what had become an essentially freshwater impoundment dominated by a monostand of common reed, Phragmites australis. Therefore, this restoration effort was essentially composed of two individual, yet connected, restorations.

The purpose of this study was to monitor ecologically relevant parameters including water quality, vegetation, nekton, and birds before and after restoration to document the effects of tidal restoration at Potter Pond. This study also uses a nearby natural marsh, Coggeshall Marsh, as an experimental control for Potter Pond. Results from this study add to the knowledge base of the effects of salt marsh restoration. Comparisons are made between results from this study and those from previous studies in New England. This study adds a unique perspective because the tide-restricted Potter Pond was consistently impounded with excessively high water levels, whereas other restricted marshes described in the literature are instead typically deprived of adequate estuarine water (e.g., Burdick et al., 1997; Jivoff and Able, 2003; Myshrall et al., 2000; Roman et al., 2002).

STUDY SITES

The Potter Pond restoration site is located in the Narragansett Bay National Estuarine Research Reserve on Prudence Island, Rhode Island (Figure 1). Potter Pond is a small (2.3 ha) estuarine pond fringed by a narrow band of salt marsh vegetation. It has been almost completely tide-restricted by a farm access road since at least the 1930s (the date of the earliest aerial photographs of the area), but the exact duration of the restriction is unknown. Before restoration, Potter Pond was connected to Narragansett Bay under the road by two crushed concrete 38-cm-diameter culverts. A second crushed plastic culvert further restricted the back portion of the pond under another road, resulting in an impounded fresh/brackish area dominated by P. australis with minimal open water. This 0.24-ha Phragmites-dominated area is hereafter referred to as the Upper Impoundment, whereas the estuarine pond and fringing marsh area (2.04 ha) is hereafter referred to as the Lower Impoundment. Potter Pond (or Pond) refers to the entire complex, consisting of both impoundments. All crushed culverts were replaced in April 2003 to restore proper tidal exchange. A 1.5-m × 1.2-m, 15.25-m-long aluminum arch culvert was installed to connect the Lower Impoundment with Narragansett Bay, and a 61-cm-diameter round pipe with flared ends was installed to reconnect the back end of the Lower Impoundment to the Upper Impoundment.

The control site, Coggeshall Marsh, is a 25.5-ha salt marsh located 1.5 km to the north of Potter Pond on Prudence Island. Coggeshall is completely unrestricted and is an excellent example of a mature southern New England meadow marsh, interspersed with tidal creeks, some remnant marsh pools, and unmaintained mosquito ditches. It is dominated by low marsh Spartina alterniflora and salt meadow species (e.g., Spartina patens, Distichlis spicata, and Juncus gerardii), with the shrubs Iva frutescens and Baccharis halimifolia bordering the upland.

MATERIALS AND METHODS

Monitoring was conducted before and after culvert replacement to quantify the ecological responses of Potter Pond while using Coggeshall Marsh as an experimental control. In this way, a before-after-control-impact ([BACI] Stewart-Oaten, Murdoch, and Parker, 1986; Underwood, 1992) experimental design was utilized. The distribution and area of habitats, along with water quality, emergent and submersed vegetation, nekton, and birds were all monitored before and after restoration.

Vegetation types and habitats in Potter Pond were mapped to submeter accuracy using a Trimble Pro-XR GPS unit in 2000 and again in 2004. The outlines of all major habitat types were delineated by walking with the GPS and then importing into ArcView 3.2a. Mapped habitats included estuarine open water (subtidal), mud flats, Salicornia spp., S. alterniflora, salt meadow (composed primarily of S. patens and D. spicata), I. frutescens, and P. australis. In addition, individual tufts of S. alterniflora were counted and marked on the 2004 map to record the colonization of this species onto the new mud flats.

Basic water quality parameters were monitored before and after tidal restoration using YSI model 6000 sondes programmed to collect data every 30 minutes. Two sondes were deployed in Potter Pond: one was deployed in the Lower Impoundment, near the connection with Narragansett Bay, and the second was deployed in the Upper Impoundment. Control data were obtained with a third YSI sonde that is permanently deployed nearby in Potter Cove (part of Narragansett Bay) as part of the National Estuarine Research Reserve's System-Wide Monitoring Program. Prerestoration data were collected between July 12 and July 25, 2000, and for approximately 3 months immediately before the culvert replacement in April 2003. Monitoring continued for approximately 3 weeks immediately after restoration and then again from July through early September 2003. However, because of problems with the sondes, the only period when high quality data were simultaneously collected from all three sites in summer after restoration was from late August to early September. Every 30 minutes, each sonde collected data for temperature, salinity, dissolved oxygen, water depth, and pH. Daily changes in water levels (a proxy for tide range) were calculated for each area by subtracting the lowest depth reading from the highest reading each day.

Emergent vegetation at Potter Pond and Coggeshall was monitored using 1-m² plots located at intervals along randomly selected transects in the marshes. All transects ran from the upland edge to a primary open water body (either a main tidal creek or open water pond, but not marsh pools). Twenty-two plots along seven transects were sampled in the Lower Impoundment, 12 plots along two transects were sampled in the Upper Impoundment, and 21 plots along three transects were sampled in Coggeshall Marsh. In addition, three transects were sampled in Coggeshall Marsh. Vegetation was monitored in each plot using the point-intercept method (Elzinga, Salzer, and Willoughby, 1998) at the end of the growing seasons (late August to September) in 2000, 2003, and 2004. In each quadrat, the percent cover of all species was quantified, and, when present, the heights of up to 12 S. alterniflora and P. australis stems were measured. Percent cover of macroalgae was sampled in the Lower Impoundment in September 2000 and 2003 using 30 1-m² plots that were randomly established throughout the open water. Macroalgae had disappeared by 2004, and sampling did not occur during that year.

Nekton was sampled in Potter Pond and Coggeshall Marsh using a throw trap according to protocols described in Raposa, Roman, and Heltshe (2003). The trap measured 1 m² × 0.5 m high and was constructed of an aluminum frame surrounded by 3-mm mesh hardware cloth on the four sides. Captured nekton was removed from the trap using a 1-m × 0.5-m dip net with 3-mm mesh. Each site was sampled in July and September of 2000, 2003, and 2004. On each date, 25 stations were sampled from each site. All samples were collected from aquatic habitats adjacent to the vegetated marsh surface (e.g., creeks, pools, ponds) on ebbing tides, beginning after the marsh surface had drained of tidal water. A station was sampled by slowly approaching it from the marsh surface and quickly throwing the trap into the water. After the trap was secured into the sediment, dip netting was conducted from all four sides of the trap until nekton was absent from three consecutive swipes. Each individual animal was identified, counted, and returned live to the field.

Birds were monitored in Potter Pond and Coggeshall Marsh in 2000, 2003, and 2004 using visual point counts (e.g., Howe et al., 1997). One monitoring point was located in the Lower Impoundment, one was in the Upper Impoundment, and four were located at Coggeshall Marsh. All birds observed using the marsh within 10 minutes at each point were identified and counted. Fly-overs were not counted unless they were actively feeding above the marsh. Each point was sampled approximately biweekly from late June through September (seven dates each year). To facilitate comparisons, all data were converted to the mean number of birds observed per viewing effort.

Changes in ecological parameters were analyzed using an array of statistical tests. Water quality comparisons were made using analysis of variance (ANOVA) and Student-Newman Keuls (SNK) pairwise comparisons. Changes in emergent vegetation, nekton, and bird community composition were analyzed among the 3 years of the study using analysis of similarity (ANOSIM). If any statistical differences in composition were detected, the species most responsible for the differences, based on percent contribution, were determined using similarity percentages (SIMPER). Both ANOSIM and SIMPER are part of the PRIMER 6 statistical package (PRIMER-E Ltd.). Changes in the total abundance of all nekton and of all birds were analyzed using ANOVA followed by SNK pairwise comparisons after transforming the data to address the assumptions of normality and equal variance. Richness of nekton and of birds was estimated using the jackknife technique described by Heltshe and Forrester (1983) and compared among years using Student's t test.

For all water quality parameters, statistical comparisons were only made among sites during the same year; interannual comparisons were not made within each site because monitoring was not conducted during the exact same time each year. For all biological parameters, all analyses were performed within the same marsh among years, and no attempt was made to directly compare parameters between marshes (e.g., between Potter Pond and Coggeshall Marsh) within the same year. Such analyses can be useful to track the degree to which restoration results in increased similarity between the restoring marsh and a reference marsh. However, in this study, it was not felt that Potter Pond would necessarily become more similar (in terms of ecological structure) to Coggeshall Marsh with time. Although they are both salt marshes, the two sites remain very different in terms of size and habitat characteristics. Instead, Coggeshall Marsh was simply used as an experimental control to account for interannual variability in the parameters that were monitored.

For all biotic parameters except nekton, the Upper and Lower Impoundments were treated as separate experimental units. For nekton, however, data from the two impoundments were pooled before any statistical analyses because nekton freely moved between the two sites with the tides on a daily basis after restoration, and the two areas essentially functioned as one contiguous dynamic tidal pond for nekton.

RESULTS

DISCUSSION

Initial results from monitoring clearly demonstrate that restoration of the Potter Pond marsh complex improved ecological conditions in this system. The increased tidal exchange with Narragansett Bay improved tidal range in Potter Pond, which flushed out stagnant and decaying mats of macroalgae. Phragmites height and cover were significantly reduced, and pioneer species such as Salicornia spp. and S. alterniflora began encroaching on newly exposed mud flats in the Lower Impoundment. Bird abundance (especially shore-bird abundance) increased significantly after restoration in the Lower Impoundment. However, the responses of nekton and the emergent marsh vegetation community did not proceed as expected. In contrast to what was observed at other salt marsh restoration sites in Narragansett Bay (Raposa, 2002; Roman et al., 2002), and despite improved marsh access and water quality, nekton density actually decreased after restoration at Potter Pond. In addition, the composition of the emergent marsh vegetation community in the Lower Impoundment did not change even after 2 years of restoration; this was because most vegetation species typically found in New England salt marshes were already present in this area before restoration. At the very least, these somewhat unexpected findings further illustrate that ecological responses to salt marsh restoration are to some degree site-specific and depend heavily on prerestoration or tide-restricted conditions, making it more difficult to identify generalized ecological responses to restoration and demonstrating that ecological monitoring should be conducted before and after any salt marsh restoration project. In addition, generalized patterns will be all the more difficult to identify unless monitoring at restricted, restoring, reference, and control sites uses standardized or consistent methodologies and is conducted for multiyear times scales (Konisky et al., 2006).

Some of the more conspicuous ecological changes in Potter Pond were surely in response to the newly formed mudflats that were exposed after tidal restoration. An abundant, mixed-species assemblage of shorebirds arrived during the first year after restoration to feed on the new tidal flats. Although an increase in these birds was expected, the magnitude of the increase was notable. Overall, bird abundance increased by 1400% during the first year of restoration. Abundance decreased somewhat in year two, but was still 1000% higher than it was under tide-restricted conditions. At a marsh restoration site in New Jersey, mudflat area increased after restoration, and this increased the richness and abundance of mudflat foragers, mostly due to large flocks of sandpipers (Seigel, Hatfield, and Hartman, 2005). This illustrates that vegetated salt marshes are not the only habitats that are adversely affected by tide-restricting structures; other important estuarine habitats, such as exposed mud flats, should also be considered as targets of tidal restoration.

The new mudflats also facilitated the expansion and colonization of S. alterniflora and Salicornia spp. By the second year of postrestoration, these species had already firmly established themselves on the flats. In the short term, the Lower Impoundment of Potter Pond should continue to provide foraging habitat for shorebirds, but over the long term it is expected that emergent vegetation will continue to colonize the open mud flats. This should eventually result in a small, well-developed salt marsh with a single major tidal creek bisecting and draining the marsh. The loss of mudflat habitat to vegetation colonization has been linked to a decrease in avian use (Eertman et al., 2002), and the increased value of Potter Pond for shorebirds, although impressive initially, should diminish if marsh development progresses as expected.

The restored tidal patterns in Potter Pond elicited additional ecological changes. Conspicuous among them was the disappearance of large decaying mats of green marcoalgae and the decrease or loss of nekton species that are intimately associated with macroalgae. Cyprinodon variegatus and L. parva are common in submerged aquatic vegetation (SAV) such as eelgrass (Zostera marina), widgeongrass (Ruppia maritima), and macroalgae (Able and Fahay, 1998; Raposa, 2002; Raposa, personal observation; Raposa and Oviatt, 2000; Sogard and Able, 1991); Cyprinodon variegatus density decreased dramatically after restoration of Potter Pond, and L. parva disappeared entirely from the pond after only 1 year. Although it was not quantified in this study, some of the decrease in these species may be attributable to increased predation by wading birds. Deeper water and SAV (including macroalgae) both provide a refuge from wading bird predation for small nekton (Gawlik, 2002; Powell, 1987; Rozas and Odum, 1988). After tidal flow was restored to Potter Pond, macroalgal habitat was lost and shallower water was much more prevalent during ebbing and low tide stages. Both of these conditions are conducive to increased predation by wading birds and may explain the increased abundance of these species (e.g., E. thula and Tringa melanoleuca) and the decrease in the density of certain nekton species after restoration. Indeed, direct predation by these birds on nekton in shallow water was frequently observed in Potter Pond once macroalgae was lost after restoration (Raposa, personal observation).

The overall decrease in nekton density after restoration was not expected and could be misinterpreted as a detrimental ecological response to restoration. However, the decrease in nekton density in this case may have been due to abnormally high numbers of small juveniles of hardy marsh species (e.g., F. heteroclitus, C. variegatus, and L. parva) that took refuge in macroalgae throughout the Lower Impoundment before restoration; when algae was lost because of restored tidal flushing, densities of these species and of nekton overall were significantly reduced. The issue here is not that nekton density decreased, but rather that density alone can sometimes prove to be a poor indicator of ecological function. In the case of Potter Pond under tide-restricted conditions, piscivorous birds and fish were not able to forage extensively on small nekton, resulting in artificially high nekton densities. Larger fish could not enter the pond through the crushed culverts, and wading birds could not effectively forage in the deep, algae-ridden water. The restoration of tidal flow resulted in a decrease in nekton density, but this was surely in part due to increased predation pressure, at least from wading birds. This indicates that density should not be the only indicator of nekton condition in salt marshes, which supports the contention of Van Horn (1983) that density can be a misleading indicator of habitat quality in general (e.g., high densities are not always better or natural). Other indicators of nekton function, such as richness, biomass, growth rates, and predation rates by larger fish and birds, should also be measured to fully understand the response of nekton to restoration.

A common goal of salt marsh restoration in New England is the reduction in cover and height of P. australis, and this goal was achieved in each of the 2 years immediately after restoration of Potter Pond. Both of these parameters decrease sharply as salinities exceed 26 ppt (Warren et al., 2002), and this level was achieved on average in the Upper Impoundment of Potter Pond after tidal restoration. As live P. australis was lost in the Upper Impoundment, it was replaced by an increasing area of open water habitat that was rapidly used by highly mobile nekton such as F. heteroclitus and Palaemonetes spp. These and other nekton species rely heavily on habitats in the upper reaches of salt marshes that retain water at low tide, such as marsh pools and distal ends of small creeks and ditches (Adamowicz, 2002; Raposa and Roman, 2001; Rozas, McIvor, and Odum, 1988), and the creation of this habitat within the former P. australis monoculture was an important result of restoration in this study. Nekton was documented utilizing this new habitat during the summer in this study, and it is expected that species such as F. heteroclitus and C. variegatus will also use it for overwintering in the marsh (Able and Fahay, 1998; Raposa, 2003; Smith and Able, 1994).

In contrast, bird richness, abundance, and community composition remained unchanged in the Upper Impoundment in both years after restoration. This is partly because the large increase in overall bird abundance that was seen in this area in 2004 was mostly due to one flock of I. bicolor that was resting on the reeds on a single date. Although there were no overall statistical changes, some subtle effects of restoration on bird use of the Upper Impoundment were apparent. For example, the new edge that formed when some P. australis was converted into open water attracted new edge-associated songbirds (e.g., Dumetella carolinensis, Geothlypis trichas, D. petechia, and Quiscalus quiscula) to the area after restoration. A nesting pair of Turdus migratorius with chicks was even observed in the edge of the P. australis. This supports earlier studies showing that low diversity P. australis monocultures with little open water are typically used by a bird community dominated by a single species (A. phoeniceus) and that increasing habitat heterogeneity after restoration increases marsh use by multiple avian guilds (Craig and Beal, 1992; Seigel, Hatfield, and Hartman, 2005).

Phragmites australis has been shown to be important habitat for the songbird A. phoeniceus and nesting habitat for wading bird species such as Egretta caerulea, E. thula, and Plegadis falcinellus (Parsons, 2003; Seigel, Hatfield, and Hartman, 2005). The ecological responses to the loss of P. australis at Potter Pond further suggest that complete eradication of this species may not be necessary to improve habitat for birds and nekton. Instead, merely increasing habitat heterogeneity by fragmenting a P. australis monoculture with open water or tidal creeks can benefit edge-associated bird species and nekton species that are adapted to extreme upstream marsh habitats. Thus, in the case of Potter Pond, if P. australis eventually begins to stabilize or even rebound from the initial shock of tidal restoration (e.g., Roman et al. [2002] observed 2 years of decreasing P. australis height after restoration, followed by an increase in year three), the increased habitat benefit can still be maintained as long as the P. australis/water mosaic is also maintained.

This study provides an excellent example of the difference between using a nearby or contiguous marsh as a reference site and an experimental control. Coggeshall Marsh and Potter Pond are very different in terms of overall marsh size, habitat composition and configuration, and hydrology. Therefore, in this study, Coggeshall was never intended to function as a reference site to which Potter Pond would become more ecologically similar with time. Indeed, there are no marshes in the immediate vicinity of the study area that are structurally similar enough to Potter Pond to be able to serve as a proper reference site. This problem is not unique to this study; in New Jersey, Seigel, Hatfield, and Hartman (2005) were unable to identify a suitable nearby reference site because of issues related to working in an urban environment and because most other marshes in the study area were themselves heavily invaded by P. australis. For the Potter Pond study, Coggeshall Marsh was used as the experimental control in a BACI design. Using this approach, this study was able to control for and extract the effects that were due to natural interannual variability, making it possible to clearly illustrate which ecological responses were due to tidal restoration (e.g., the increase in F. majalis density in Potter Pond in 2004), and which changes were not (e.g., the decrease in F. heteroclitus and Palaemonetes spp. densities in Potter Pond in 2004). This demonstrates that even if an appropriate reference site cannot be identified, a nearby marsh can easily serve a useful function by helping to control for natural variability. Without this type of control, it can never be unequivocally determined whether an observed change in a restoration site was actually due to the restoration (e.g., because a control marsh was not used, Seigel, Hatfield, and Hartman [2005] were not able to determine whether decreasing sandpiper numbers over time were due to changes in the restoration marsh or to larger regional trends).

In summary, restoration of the degraded Potter Pond system quickly resulted in improved ecological conditions after only 1 year of restored tidal flow, and these improvements persisted into the second year. Some responses occurred immediately after restoration (e.g., nekton use of the new water in the Upper Impoundment), whereas others were more gradual (e.g., S. alterniflora and Salicornia spp. colonization of the newly exposed mud flats). Continued monitoring over the long term will quantify the long-term rates and magnitudes of these changes and determine if these restoration-induced improvements persist. The unique ecological responses that were observed at Potter Pond further emphasize that high-quality pre-and postrestoration monitoring at experimental as well as control sites should be undertaken with every salt marsh restoration. Despite its small size (2.3 ha), Potter Pond provides an excellent example of the benefits that can be derived from restoring tide-restricted habitats in New England and will hopefully encourage coastal resource managers to further seek to restore degraded sites regardless of their size. This is especially important in Rhode Island where the mean size of all marshes in Narragansett Bay is 2 ha (Trocki, 2003). Larger restorations may result in more publicity, and larger restoring marshes may more effectively provide some functions (e.g., storm buffering) than smaller marshes, but restoration of multiple smaller sites can also lead to improved estuarine conditions in the aggregate.