New England Salt Marsh Pools: Analysis of Geomorphic and Geographic Parameters, Macrophyte Distribution and Nekton Use

Salt marsh pools are shallow, steep-sided depressions that remain flooded throughout a tidal cycle and provide important habitat for nekton (fish and decapod crustaceans). Although New England pools are relatively unstudied, efforts are underway to expand pool habitat through an open marsh water management technique known as ditch plugging. The purpose of this study was to quantify pool geomorphic and geographic characteristics, to determine distribution of vegetative cover types and nekton species, and to evaluate effects of ditch plugging on nekton use of pools. Over 30 ditched and unditched marshes were surveyed from the Connecticut shore of Long Island Sound to southern Maine, USA for pool physical traits. Pools from ditched and unditched marshes had similar sizes, depths, and distances to tidal flow, but pool density (#/ha marsh) and coverage (m pool/ha marsh) were over twice as great at unditched versus ditched marshes. Ditch intensity (m ditch/ ha marsh) was negatively correlated to pool density and coverage. Pool vegetative cover was surveyed at 12 New England marshes using 0.25 m quadrats and the Braun-Blanquet scale for coverage estimation. Eight cover types were identified with the most common being "bare" (pool bottom visible), filamentous green algae, algal flocculation, and Ruppia maritima. Pool cover types varied by region, with bare more prevalent at northern sites. Canonical correlation analysis indicated that environmental conditions supporting R. maritima and algal floe coverage were distinguished by water depth, soft sediment depth, and water column nitrogen concentrations. Pool nekton sampling using 1 m throw traps in 1999 and 2000 documented 12 fish and 4 decapod species with Fundulus heteroclitus comprising 80% of the fish catch. Species richness and density were greater in southern than in northern New England sites. In contrast to results from other studies, fish densities were not affected by the presence of submerged aquatic vegetation (R. maritima). Data suggested that F. heteroclitus use pools as nursery habitat and as refuges from predation, selecting among pools with different environmental conditions as the season progressed and juvenile size increased. Restoration efforts, therefore, should maintain a range of pool conditions to support a diverse nekton community and its changing needs. The effects of ditch plugging at 3 salt marshes in the Rachel Carson National Wildlife Refuge, Maine, USA, were evaluated using a BACI (Before-After-ControlImpact) study design. Throw-trap sampling produced a total of 7 fish and 3 decapod species in salt marsh pools with F. heteroclitus accounting for 89% of the fish catch. After controlling for natural variation using the BACI design, nekton community and total fish density remained unaltered at Moody and Granite Point. Decreased species richness at Granite Point marsh was thought due to physical barriers created by ditch plugs and a 51 % reduction in tidal interface. At Marshall Point, ditch plugging resulted in increased pool habitat and nekton use at the experimental versus control area pools. Both fish and decapods occupied newly created pools within weeks and within months new pools were not different from old pools in terms of nekton use. This study documents only initial nekton response to ditch plugging; longer term monitoring is necessary to ensure achieving management goals. ACKNOWLEDGEMENT This study was supported by the USGSBiological Resources Division, administered through the Patuxent Wildlife Research Center, Coastal Field Station at the University of Rhode Island. My deepest thanks also to the USFWS for promoting this research and to the staff at Rachel Carson National Wildlife Refuge for their support and for providing exquisitely beautiful study sites. My gratitude extends particularly to Jan Taylor, USFWS; Ward Feurt, Rachel Carson National Wildlife Refuge; and Mike Morrison, SW AMP, Inc. Thanks also to all those who braved summer heat and autumn chill to help with field work especially Susan Williams, Brandi Bomt, Kenny Raposa, Mike Adamowicz, Sean Ballou, MJ James-Pirri, Ryan Taisch, and Linda Longo. Charles Roman is the best advisor a graduate student could hope to find . Thank you for your support, reality checks, field assistance, and introduction to Fat Boys in Maine. Your ecological and editorial insights have proven critical. Thank you for your patience and your good sense of humor. Thank you for carefully reading all those rewrites and for keeping high standards. My sincere appreciation goes to Candace Oviatt who encouraged me to stop messing around and just get the Ph.D. Your support came at a time when I doubted even the simplest things. Thank you for your staunch backing. James Heltshe opened the world of statistics and proved great help throughout the course of this study. Any mistakes in that realm, however, are my own. Michael Pilson and Peter August also have my gratitude for their help and insights during my studies and particularly for their assistance as committee members.

water throughout a tidal cycle and do not dry as shallower pannes often do . Despite the extensive literature on salt marshes, research specific to marsh pools is quite limited. , , and  have provided the most extensive descriptions of pools but no quantitative analyses. While several authors  have concluded that pools provide important fish and wildlife habitat, no research has been conducted in New England that quantifies basic geomorphic and geographic pool features.
Salt marsh pools are significant because they provide habitat for numerous bird species. In Massachusetts, for example,  noted that shorebirds, wading species, terns, swallows, and crows were strongly attracted to salt marsh pools.
In southern New Jersey,  documented that various shore and wading birds actually aggregated at pools to feed on dense fish populations. Erwin' s (1996) review paper on Mid-Atlantic coastal habitat supported the conclusion that salt marsh pools are vital year-round habitat for dabbling ducks, shorebirds, and wading birds. Pools are valuable to birds because they provide food items including fish and macrophytes, specifically Ruppia maritima L. .
New England salt marshes have been extensively ditched for mosquito control and salt hay farming, resulting in lowered watertable levels and drainage of the marsh surface often including salt marsh pools. Salt hay farming has been a fairly common practice since colonial times , but the most extensive and widespread 3 ditching occurred in the early-to-mid 1900' s for mosquito control ).  stated that by 1938, approximately 90% of Atlantic coastal marshes from Maine to Virginia had been ditched in an effort to reduce breeding habitat for the salt marsh mosquito (Aedes sollicitans (Walker)). Many natural pools, therefore, were eliminated before the middle of the century.  determined that the effect of ditching on bird habitat was negative --while prey items were still abundant in marshes, they were located in ditches and were harder to capture due to the ditches ' steep sides and fluctuating water levels.
Given their ecological benefits, restoring pool habitat should be an important salt marsh restoration objective. To enhance the success of such work, welldocumented targets for pool creation are necessary. The objective of this paper is to address this need by providing a quantitative baseline analysis of geomorphic (e.g. pool depth, size (m 2 )) and geographic (e.g. density, spatial arrangement, and coverage) attributes of salt marsh pools for a number of ditched and unditched salt marshes throughout New England ( Figure 1).
Not all marshes, however, have the same density of ditches. Therefore, it should be possible to determine whether ditch density correlates with pool density and pool surface area per hectare of marsh. While expected trends might seem intuitively obvious,  suggested that ditch levees caused large pools to form, whereas  concluded that ditching resulted in decreased pool size and number.

METHODS
Salt marsh pools were assayed using both field investigations and aerial photographic surveys coupled with geographic information system (GIS) analysis. The complete effort extended from the Long Island Sound shoreline of Connecticut to midcoast Maine. A total of 27 sites were surveyed by either or both methods.

Field Survey
Field surveys were conducted at 17 salt marshes throughout New England during the 1999 and 2000 growing seasons (Table 1 ). Impounded, filled, tidally flowrestricted, and privately owned marshes were not sampled. Unditched marshes were very uncommon; in Rhode Island, the few unditched sites were substantially smaller than other surveyed marshes and were not included in this study. After much investigation, one unditched site was located in Connecticut but it was included only in the aerial photographic analysis described below.
At each salt marsh, a study area was identified as a discrete region bounded by major creeks, roads, upland, etc. Line transects were used to select sample pools for a number of measures described below. The initial transect was randomly located along the main tidal channel or shoreline with subsequent transects set at 1 OOm intervals.
Transects extended from the edge of the marsh to the upland border and were measured using a metric survey tape. Only pools in direct contact with transect lines were included in the study. The number of pools located on the transects was used to calculate a measure of pool density(# pools/km transect). Adapting the standard line-5 intercept method used in vegetation ecology , pool coverage was defined as the total portion of the transects in contact with pools (m of pool/km transect).
For each individual pool encountered along the transects, pool surface area was estimated by measuring the major and minor axes of regularly shaped pools ) and applying the formula for the surface area of an ellipse. Irregular pools were subdivided into simple geometric shapes, with appropriate data taken to calculate surface area. For each pool, geographic parameters such as distance to nearest neighboring pool and distance to tidal flow (either a ditch or creek) were also measured.
Total pool depth was measured as the distance from the surface of the marsh to the pool bottom. Water depth was recorded for use in comparison with other studies, but it is less consistent than total pool depth since the depth of water alone depends on tides, rainfall, etc. (Figure 2).
In order to determine if pools were restricted to certain locations within a marsh, study areas were divided into three zones based on vegetation. The low marsh zone was dominated by Spartina alterniflora Loisel (tall form), the high marsh was dominated by S. patens, while a third zone was dominated by S. alterniflora comm.). In addition, color infrared (CIR) images were available for a limited number of sites in Connecticut and Maine. A summary of data sources is provided in Table 1; additional metadata are provided in Appendix A. A total of 21 ditched and 6 unditched marshes were surveyed with ArcView 3.2 geographic information system (GIS) software applied to the DOQs and CIR photographs.
Study sites were selected based on availability of high quality digital images and minimal degree of anthropogenic alteration. Site boundaries were delineated in the same manner as in the field surveys. Tidal channels greater than 1 Om wide were excluded from the salt marsh study area while natural channels less than 10 m wide ("creeks") and all ditches were included. Large channels were excluded as representative of open water or aquatic habitat .
Within the study boundary for each site, all pools, creeks, and ditches were delineated with Arc View shape files to calculate pool density and total pool surface area per hectare of salt marsh. Transects were established following methods used in the field survey. Pool data (individual pool surface area, density, distance measures, 7 coverage) were calculated from the transects to compare with the field survey methods and to permit some combining of the GIS and field survey data.

Comparison of Survey Methods
The two methods (field survey transects and DOQ transects) used to survey salt marsh pools were compared. Paired t-tests were run for each variable (pool density, pool coverage, nearest neighbor distance, distance to tidal flow, and average individual pool size) obtained from both methods. Pool size data were log-transformed prior to analysis; no other transformations were necessary.

Statistical Analyses
Pool densities for field and digital orthophotographic quadrangle (DOQ) transect studies were combined to test for differences attributable to marsh type (ditch vs. unditched) using a two-sample t-test. Relying on field data alone, ANOV A was used to determine whether the proportion of pools differed among the three marsh zones (S. alterniflora (tall), S. patens, S. alterniflora (intermediate)) or between marsh types.
Logistic regression was applied to determine which environmental variable(s) best predicted pool location . The effect of marsh type on a series of pool variables measured from DOQs was determined by two-sample t-test.
An index of dispersion was calculated for each marsh surveyed with the DOQ/GIS method. The index (I) is based on nearest neighbor distances (NND) from a minimum of 10 to a maximum of 20 randomly selected pools , Moore 8 1954. The resulting values were then averaged by marsh type (ditched or unditched) and subjected to a two-sample t-test.
While it is instructive to examine ditched versus unditched marshes, it is also clear that some marshes have been ditched more intensively than others. Similarly, ditched and unditched marshes also varied in the amount of natural creek length each possessed. Because ditches and creeks both convey water to and from the marsh, these parameters may be correlated with pool density and the amount of pool surface area per hectare of marsh. Tide range and latitude were also suggested as possibly affecting these parameters. Therefore, simple linear regressions were used to determine the relationship between these factors and pool density and pool surface area per hectare obtained from DOQ sources. Mean tidal range data for each site was obtained from Tides and Currents for Windows (Nautical Software, Inc.). Latitude was obtained from DeLorme TopoUSA Regional Series topographic maps, a commercially available compact disc product.
In order to more closely examine the effect of ditching on salt marsh pools and to control for potential effects of tidal range and latitude, ten areas (five paired sites) were selected from the DOQ data set that were large marshes where a ditched section adjoined or immediately adjacent to an unditched portion of marsh . These pairings ranged from Long Island Sound (Connecticut shore) to southern Maine. The same procedures were followed for delineating study areas, ditches, creeks, and pools as previously noted. Linear regressions and paired t-tests were used to determine the effect of ditching on pool density(# pools/ha salt marsh) and total pool surface area per hectare salt marsh (m 2 pool/ha salt marsh) .

RESULTS
Pool Density, Size, Spatial Pattern, and Other Geographic Measures Transect data from field and DOQ evaluations showed that unditched marshes had a significantly greater density of pools than ditched marshes ( When calculated on an area basis, pool density(# pools/ha of salt marsh) was three times greater in unditched marshes compared to ditched marshes (Table 3). Other parameters with significant differences between marsh types included total pool surface area per hectare (m 2 pool/ha salt marsh), pool coverage (m pool/ km transect), ditch length (m ditch/ ha salt marsh), and creek length (m creek/ha salt marsh). While there was no significant difference in the Index of Dispersion (I), a measure of spatial arrangement, between the two marsh types, fully one-third of the ditched marshes had an index value greater than 1.0 (indicating a uniform pattern), whereas the I-values for unditched marshes were all less than 1.0 (indicating an aggregated pattern; . These results suggest a non-significant tendency for pools to be more spread out in ditched marshes compared to unditched sites. Somewhat unexpectedly, several pool measures were the same between ditched and unditched marshes. Pool size averaged nearly 200 m 2 and the distribution of pool size (log-transformed data, t-tests) was the same in both marsh types. Pools in ditched marshes were no further from nearest neighboring pools or tidal flow than pools in unditched marshes.
Influence of Ditching, Creeks, Tidal Range, and Latitude Along a gradient from unditched to highly ditched marshes, both pool density (# pools/ha salt marsh) and total pool surface area per hectare (m 2 pool/ha salt marsh) declined significantly ( Figure 2). The correlation coefficient reflects variability in the data, especially among marshes with little or no ditching; not all unditched marshes have abundant pool habitat. When pool parameters were regressed against creek length per hectare (Figure 3) there was a positive relationship in each case; as creek density increased, so too did pool density and total pool surface area (m 2 pool/ha salt marsh) also increased. The opposing relationship between surface creeks and ditches is most evident in Figure 4; the regression equation indicates that 314 m of ditching per hectare of salt marsh was sufficient to eliminate the creeks.
Mean tidal range and latitude were not correlated to pool density (p=0.08, p=0.52, respectively). Alternatively, both mean tidal range and latitude were correlated to total pool surface area per hectare (p=0.02, p=0.04, respectively) but their coefficients of determination values (R 2 =0.20 for tidal range, R 2 =0.16 for latitude) were low. When subjected to multiple linear regression, creek length per hectare and tidal range were the most significant variables (p<0.05, both), but the coefficient of determination also was very low (R 2 =0. l 7).

Methodology Comparison
Nine sites (6 ditched, 3 unditched), where both DOQ and field surveys had been conducted, were used to compare transect results obtained from the two methodologies.
Because some surveys (variously among the field and DOQ methods) did not encounter pools, sample size was either 7 or 9 sites for the 5 parameters of concern. Paired t-test results indicated no significant difference between the methods for pool density(#  Redfield ' s (1972) hypothesis that increased ditching resulted in decreased pool density and size.
The opposing relationship of ditch and creek density to pools (increased ditch density corresponded to a decrease in pool density and total area whereas increased creek density corresponded to high pool values) may be due to the way these channels drain a marsh. It is suggested that ditches are generally deeper than the first and second order tidal creeks (as defined by Bayliss   The extensive ditching found throughout New England salt marshes is largely due to public works programs of the Depression Era that enabled mosquito ditching to proceed on a large-scale, rapid basis. By the late 1930s, many East Coast states were reporting near complete drainage of designated salt marshes. Deleterious effects of ditching on wildlife habitat were reported early on (Urner 193 5) with the discussion continuing throughout the 1930s  and the remainder of the century .
The numerous alterations to salt marshes, from ditching, draining, and filling, to diking and restricting tidal flow , have inspired a number ofremediation efforts.
As early as 193 8, resource managers sought a means to balance mosquito control with wildlife conservation. Open marsh water management ("OMWM'') emerged as the premier method to utilize natural processes for mosquito reduction (Cottam 1938, Ferrigno and. Although OMWM techniques have been available for many years, they require more planning than traditional ditching and more labor than pesticide application. The methodology, however, has gained prominence in the last two decades and has been the focus of numerous fish and wildlife studies

Restoration Recommendations
If increasing bird habitat is a goal of salt marsh pool restoration and creation, then the preferred pool dimensions depend upon the species under consideration and the time of year . The size recommendations given by  for 17 construction of large pools (greater than 800 -1000 m 2 for autumn waterfowl and summer shorebird use) lie well within the range of pools on New England salt marshes (3 -3 786 m2). Recommended small pond sizes (less than 200 m 2 for autumn waterfowl) approximate the average size of pools in ditched and unditched marshes.
While  proposed a depth ofless than 15 cm, salt marsh pools in New England had a water depth range of 3 -71 cm. Rather than build "perfect" and identical pools, managers, perhaps, should heed the variability of pool dimensions found in the field. Such variation in pool size and depth over a large number of pools ensures that at least some of them will remain desirable to birds regardless of natural fluctuations in tidal flushing, rainfall, or drought.

Comparison of New England to Other Regions
There have been very few quantitative studies of salt marsh pools. Three recent and one historic reference are provided in Table 4. Pool density ranges from a low of zero in temperate Australia (where  presumably expected to find pools in those grass dominated systems) to a high of 1300 pool/km 2 salt marsh in the unditched marshes of this study. Pool sizes documented across all regions ranged from lm 2 to a high of 55 ha. An exceptionally large pool was located at 100 Acre Cove in Rhode Island where DOQ/GIS analysis estimates pool size to be approximately 3 5 hectares.
While pools ofthis size have not been encountered in other New England salt marshes, evidently such large pools are found elsewhere as demonstrated by the work of Master (1992). Water depths were similar at all locations except for the historic note of a pool 1.2 m deep at an unspecified location in Rhode Island (Price 193 8 assistance, and critical reviews of early manuscript drafts.
20 Table I. List of salt marshes and assessment techniques at each site included field transect surveys ("field surveys"), analysis of digital orthophotographic images ("DOQ ") and color infrared images (" CIR") and other mapping or photographic base images ("Other"      Figure 2. Total pool depth was measured as the distance from the marsh surface to the bottom of a pool while water depth was simply the distance from the water surface to the pool bottom.      Color infra-red photographs were obtained from Ron Rozsa at the Long Island Sound Project, CT DEP, Hartford, CT. These photographs were flown on August 1995 at a scale of 1: 12,000. They were gee-rectified using the Image Analyst extension of ArcView 3.2

Rhode Island
Black and white DOQ images were obtained from the RIGIS Environmental Data Center at the University of Rhode Island. The aerial photography was flown in April 1997. The DOQs are NAD 83 , state plane feet, MrSID compression with 0.6 m/pixel. Additional metadata are available from RIGIS at http ://www.edc.uri.edu.

Massachusetts
Black and white DOQ images were obtained from the Massachusetts Geographic Information System Office at ht1.P_~/.!~~tat~_Jn.~_,_ \!_~Lmgj_~/.m_ass_gj_~,.htm . Aerial photographs were flown at a scale of 1 :5000 in 1999 for the Felix Neck site and in 1994 for the other Massachusetts sites in this study. The images were converted to DOQs at NAD 83 , state plane meters, MrSID compression with 0.5m/pixel. Black and white DOQ images were obtained through the courtesy of the US Fish and Wildlife Service, Gulf of Maine Program. Aerial photographs were flown at a scale of 1 :24,000 in 1997 and were converted to DOQs at NAD83 , UTM Zone 19, meters, MrSID compression with 1. 0 m/pixel. A color infrared aerial photograph was used for the Wells National Estuarine Research Reserve. It was compressed as a MrSID image, UTM Zone 19, NAD 27, meters. The photograph was taken in August 1998 at a scale of 1: 12000 .
A color infrared aerial photograph was used for Moody Marsh. The photograph was flown June 1992 at a scale of 1: 8000. A true color aerial photograph was used for Granite Point Marsh flown on May 1986 at a scale of 1 :2400. These photographs were made available by the Rachel Carson National Wildlife Refuge. Both images were digitally scanned then geo-rectified to USGS 7.5 topographic quadrangle maps using ArcView 3.2 Extension Image Analyst. Due to the low quality of the final product, analysis was limited to calculation of study site area and delineation of creeks and ditches that had been field checked. going management efforts to increase pool habitat in New England are proceeding without a regional benchmark for the composition and distribution of macrophytes in pools and their relationship to pool environmental conditions. The purpose of this study is to provide such a benchmark. Pool macrophytes were surveyed at 12 New England salt marshes using a 0.25m 2 quadrat and the Braun-Blanquet scale for coverage estimation. Eight cover types were identified with the most common being "bare" (pool bottom visible), filamentous green algae, algal flocculation, and Ruppia maritima. Pool cover types differed by region (p<0.001 , ANOSIM) with the "bare" category more prevalent at northern sites. Southern New England pools exhibited slightly more filamentous green algae, algal flocculation, and Ruppia maritima, accounting for 52% of the total dissimilarity between regions. Based on canonical correspondence analysis (CCA), R. maritima, valuable as waterfowl forage, was associated with pools that were deeper, more saline, and had less soft sediment and lower water column nitrogen than the mean values for pools in this study. Canonical correlation analysis indicated that R. maritima and algal flocculation coverage were separated by water depth, soft sediment depth, and water column nitrogen concentrations.

INTRODUCTION
New England salt marshes are not monotonous meadows ofhalophytic grasses, but rather a mosaic of vegetated and aquatic habitats. In addition to creeks and numerous human-made ditches, salt marshes also contain pools, which are softbottomed depressions that hold water throughout a tidal cycle and unlike the shallower pannes, do not tend to dry out . In New England, these pools are steep sided  and have an average total pool depth of approximately 3 0 cm (Adamowicz, Chapter 1). Based on a survey of 26 marshes from Connecticut to southern Maine, they range in size from 1 m 2 to 3 5 ha and cover an average of 2% of ditched marshes and 9% ofunditched marshes (Adamowicz, Chapter 1).
Pools are important habitats, supporting macrophytes , nekton (Able andFahay 1998, Raposa 2000), and foraging birds . Macrophytes provide nekton-nesting sites (FitzGerald 1983), are associated with high fish densities Able 1991 , Smith 1995) and elevated secondary production , and are important wildlife forage . Some factors influencing macrophyte species composition and abundance include salinity, water depth, nutrients, sediment composition, and sediment depth . While some studies  have examined macrophyte coverage in pools, their scope was limited to a small number of pools associated with just a few salt marshes. No work has detailed the range of conditions in pools found throughout New England, especially with respect to macrophyte coverage. 43 Currently, open water marsh management (OMWM) techniques are being used throughout New England and include the construction of new pool habitats.
Macrophytes in pools, however, have not been sampled in a comprehensive fashion to provide a baseline for establishing management goals. The purpose of this study is to provide a New England benchmark for macrophyte species composition, distribution, and relationships with environmental variables in order to better understand the role of pools in overall salt marsh ecosystem functioning and to provide guidelines for management efforts.

Broad-Scale Survey
Twelve salt marshes throughout New England (Table I) were part of the broadscale survey of macrophytes in pools. Study pools were selected by establishing a series of transect lines at each salt marsh. The first transect was randomly located and 44 extended from the estuary/marsh edge to the marsh/upland edge. All pools falling on the transect line (a survey tape) were sampled. Subsequent transects were set parallel to the first one at 100 m intervals. At sample pools, a 0.25 m 2 quadrat was set along a separate transect placed on the pool' s long axis. Initial quadrats were placed randomly within the first 10 m while subsequent quadrats were placed at regular intervals to obtain at least 4 sub-samples whenever pool size allowed.
Percent coverage was taken by visual estimation and recorded into cover classes according to the Braun-Blanquet scale . Since identification of filamentous green algae to species was not possible in the field and the species often occurred intertwined, they were lumped into one cover category. A species list of filamentous greens was based on samples taken from northern and southern salt marsh pools identified in the laboratory (Villard-Bohnsack 1995). Besides filamentous green algae, other cover categories included "bare" (pool bottom visible), "epiphytes" (algae epiphytic on macrophytes ), and "algal floes" (algal flocculation consisting of decayed benthic algae or decayed filamentous algae). Spartina alterniflora, an emergent species, was included as a cover category only when individual stems grew out of the pool bottom and did not represent emergent islands or surrounding marsh surface.
Other variables measured at each quadrat included water depth and soft and total sediment depth. Both sediment depths were measured by inserting a graduated metal rod through the pool bottom. Soft sediment was defined as that surface layer with minimal resistance. At the boundary between soft sediment and lower layers, a change in resistance occurred. Insertion of the rod continued until firm resistance was 45 encountered or total sediment depth exceeded 140 cm. Sediment depths in excess of 140 cm were recorded as > 140 cm.

Intensive Survey
Pools involved in the intensive survey were selected by the same transect method described above and were sampled once during the summer of 1999.
Additional parameters examined at 33 pools on 6 marshes included water pH, salinity, dissolved oxygen, and water and surface sediment redox potential, all of which were measured at 1 m from the pool ' s edge and at the pool ' s center. A YSl-85 handheld meter provided salinity and dissolved oxygen measures. An Orion model 250A with Ir and platinum redox electrodes yielded water pH and redox measures. These parameters were measured as pools were encountered along each transect (once during a field day of 0800 -1800 hr). While fully cognizant that these instantaneous measurements do not capture diel patterns within pools they were intended only to provide an initial estimate of regional pool conditions.
A further subset of 12 pools from 3 Rhode Island salt marshes was sampled for sediment and water column nutrients. Sediment and water samples were taken at a pool' s edge and center; water column nutrient samples were filtered in the field, all samples were stored on ice until they were placed in a freezer later in the day. Water samples were analyzed colorimetrically for inorganic nutrients (N0 2 , N0 3 , NH 3 , P04) on an autoanalyzer .
Three replicate surface sediment samples were taken to a depth of 10 cm from the center of each pool, combined, and iced until frozen later in the day. Rather than 46 total sediment nutrient content, sediments were analyzed for readily available nutrients.
In the lab, sediment samples were thawed and homogenized thoroughly by hand mixing. A 10-gram sub-sample was then combined with 40 ml of artificial seawater.
The mixture was shaken by hand for two 30-second intervals and then centrifuged at 8000 rpm for 20 minutes. The supernatant was removed by syringe, filtered, and subjected to the same nutrient analysis as the water column samples. Sediment dry weight and organic matter (loss on ignition) followed the methods of . Sediment grain-size analysis was obtained with the wet-sieve method (Folk 1974) using a series of standard sieves.

Statistical Analyses
Differences in cover type composition and abundance were analyzed using the PRIMER non-parametric randomization tests of similarity (Clarke and Warwick 1994, Roman et al. in press) applied to a similarity matrix. An analysis of similarities test (ANO SIM) was used to detect differences between groups of pools. Groups defined a priori included the individual marshes and the region (northern or southern) within New England. Contributions of individual cover types to dissimilarities between groups were calculated using percent similarity tests (SIMPER), with percent contribution based on Euclidean distances.
Canonical correspondence analysis (CCA) was used to relate cover types to environmental conditions for pools from all study marshes. All environmental measurements were standardized prior to analysis as suggested by . Cover types were plotted in 2-dimensional space while environmental variables 47 were presented as vectors. The vector's length represents the correlation between that environmental variable and the 2 ordination axes. The longer the vector is, the stronger the relationship between that variable and the community (McCune and Mefford 1999).  indicated that CCA is preferable over detrended correspondence analysis (DCA) where a "good set of environmental data" is available. CCA calculations and graphics were produced by PC-ORD version 4.0 (McCune and Medford 1999).
Fisher's Exact Test  was used to indicate whether the presence of macrophytes in one pool was linked to the presence of macrophytes in the nearest neighboring pool (also known as "contagion"). Thus Fisher's Exact Test can provide resource managers with an indication of whether macrophytes are isolated within pools or if there is a likelihood that propagules can be transferred among nearest neighboring pools. Such information is useful in restoration efforts in determining whether newly created pools can be colonized naturally.

Macrophytes/ Pool Cover Categories
Twelve salt marshes from Maine to Long Island Sound were surveyed resulting in eight cover types ( Table 2). Four of the cover categories were for individual species, while the "filamentous green" category was composed of Enteromorpha compressa, Polysiphonia lanosa, Rhizoclonium tortuosum, Cladophora sericea, Ulothrix flacca, Acrosiphonia arcta and Chaetomorpha linum (scientific names and authorities are given 48 in Appendix I). C. sericea and A. arcta were found at both northern and southern sites.
E. compressa, P. lanosa, and R. tortuosum were sampled only in northern marshes while U flacca and C. linum were observed only in southern marsh pools.
Pool bottoms were mostly bare, that is, they were not covered by macroscopic vegetation of any kind ( Table 2). Filamentous green algae was the next most common cover type. Algal floes and R. maritima ranked third and fourth, respectively, while other individual species were less common. Cover categories and abundances were the same within regions (p>0.17 for southern sites, p>0.44 for northern sites, ANOSIM) and between ditched and unditched marshes in the northern sites (p>0.24, ANOSIM). All marshes in the southern region were ditched . Based on these findings, data were combined within regions. When comparing cover categories and abundances between regions, there were significant differences (p<0.001 , ANOSIM) with SIMPER results indicating that northern pools had more extensive areas of "bare", while southern pools had greater coverage of filamentous green algae, algal floes, and R. maritima (Table 3). 49 Cover Type Canonical Correspondence Analysis Two biplots relating cover type to pool environmental conditions were produced from the survey data. The biplots representing physical pool conditions ( Figure 2) and water column nutrients ( Figure 3) were linked by the placement of the R. maritima and algal floes categories near opposite ends of at least one axis. The "bare" cover category tended to be close to the vector origin in each biplot due to its presence in most pools.
For the biplot with physical pool conditions ( Figure 2), only axis 2 has a significant correlation to cover ordination space and a significant cover typeenvironment correlation (p=0.01 , p=0.02, respectively; Monte Carlo test); therefore, results should be interpreted as a projection of points and vectors onto axis 2. R.
maritima thus was correlated with deeper, more saline pools with less soft sediment than average. Alternatively, algal floes were associated with shallower, less saline pools that had greater depths of soft sediment. Physical pool conditions and cover types plotted in Figure 2 correspond with environmental data summarized by site in Table 4.
As noted above, nutrient analyses of pool conditions were limited to three southern sites ( Figure 3, Table 5). Only axis 1 in Figure 3 was significant (cover type ordination p=0.01 , cover-environmental data r =0.963 , p=0.02; Monte Carlo test) thus results should be taken as projections onto this axis. Water column nutrient conditions associated with R. maritima were low N0 3 +N0 2 , low N02 and high SiQ4 concentrations ( Figure 3, Table 5). Once again, the opposite situation (high N0 3 +N0 2 , high N02 and low Si0 4 concentrations) was associated with algal floes . Ammonia and inorganic phosphate were measured in the water column (Table 5), but their levels were not influential in the CCA analysis. Finally, while detailed data were gathered on other 50 pool variables (Appendix II), resulting biplots did not have significant ordination or cover-environmental correlations and are not presented.

Abundance ofMacrophytes and Bare Areas
Few studies have examined macrophyte cover in salt marsh pools; most have a narrow focus, some mentioning cover types only briefly. A few common points can be elucidated, however. Macrophyte coverage initiates at the start of the growing season in May and peaks at 95 -100% in July and August . R.
maritima  and Cladophora spp.  are the most commonly noted species although Enteromorpha spp. and Percusaria spp. were mentioned for pools in Quebec .
Surveys conducted for this study occurred from June -mid-September when macrophytes should have been abundant. While the number of cover types identified here (8) exceeded those in other pools studies , the significant amount of bare area was an unanticipated result. On average, the bare category accounted for 25% of pool coverage in southern pools and 75% in northern pools (Table 4). Of additional interest are the several factors that appear to influence presence and abundance of macrophyte cover.

51
Factors Influencing Cover Types Cover types were influenced by conditions at two levels: regional and between pool. Regional location determined the relative abundance of cover types (Table 4) with higher levels of vegetation in southern New England and more bare area in northern pools. At the between-pool level, the presence of macrophytes in pools was contagious -nearest neighboring pools were also likely to have macrophytes. This result offers an important suggestion to restoration managers that newly created pools can be naturally colonized by macrophytes if they are located near existing pools that already contain macrophytes. Also between pools, physical and water quality conditions produced a continuum that promoted R. maritima at one end and algal flocculation at the other.
Of the four physical and water quality factors that distinguish between R.
maritima and algal floe cover, nutrient levels and soft sediment depth may indicate cover type segregation along a trophic scale. R. maritima would represent low nutrient conditions and algal floes, the more enriched pools.  indicated that R. maritima survived best in low nutrient conditions and was out-competed at high nutrient levels. This may account for the presence of Ruppia in higher salinity pools when it is normally considered a brackish water species. The interaction of nutrient levels and autotrophic species has been examined extensively for other submerged aquatic vegetation systems (e.g. Dennison et al. 1993) but not for salt marsh pools.
As an alternative hypothesis, the depth of soft sediment might indicate pool age, with pools accumulating soft sediment over time. R. maritima would occur, therefore, 52 in relatively young pools and algal floes in older pools. Whjle several theories exist for pool formation , Petmck 1974, none have been demonstrated in New England salt marshes, and none address the issue of soft sediment accumulation. The process of pool formation and senescence may have a sigruficant influence on cover types, however, and should be investigated further. Lastly, if nutrient levels and sediment accumulation were influenced by anthropogenic factors, there could be sigruficant management implications since R. maritima is important waterfowl forage .
Several factors did not distinguish among pool cover types based on CCA analysis. Water column inorgaruc P concentrations either were not lirruting (total N :P was 2.5 -3.6: 1 (Table 5) compared to Redfield ratio of 16: 1 (Valiela 1995)) or the variability among pools was too great (Table 5) to identify differences among cover types. Sediment nutrient levels, grain size, and orgaruc content (Appendix II) also failed to distinguish among cover types, possibly due to mgh levels in all pools or low sample sizes.

Pool Habitat Quality
Macrophytes enhance estuarine productivity by increasing primary and secondary production (Heck et al. 1989, Sogard and, and providing refuge from predation . As shown in tills study, however, pools, especially in northern New England, have a large amount of bare area. Ruber et al. (1981), in fact, determined that pool primary productivity (514 g/m 2 /yr) was much less than that of the surrounding Spartina salt marsh and more commensurate with that of 53 tidal mudflats. These findings suggest that pools should be relatively unimportant habitats, although as demonstrated by the consistent nekton catches , Adamowicz (Chapter 3)) and bird use  such is not the case.

CONCLUSION
This study documented more cover categories than did previous reports and also differs in reporting a greater predominance of bare areas in New England salt marsh pools. Several factors were associated with different cover types and abundances, including regional location and local pool conditions. Additionally, R. maritima and algal floe coverage were distinguished by conditions that could be indicative of pool     * Only one pool from this site was used in canonical correspondence analysis biplot construction.  Figure 2. Biplot of pool cover types and pool physical variables. Cover types are plotted as open circles (sample pools are filled circles) in two-dimensional cover type ordination space. Environmental variables appear as vectors; a vector' s length represents the correlation between that environmental variable and the two ordination axes. The minimum vector r 2 was 0.30 for this biplot. Relationships between cover types and environmental factors should be interpreted by projections onto Axis 2 since only this axis has significant correlations to cover ordination space (p=0.01) and a significant cover type-environment correlation (p=0.02). in two-dimensional cover type ordination space. Nutrient concentrations appear as vectors; a vector's length represents the correlation between that nutrient and the two ordination axes. The minimum vector r 2 was 0.30 for this biplot. Relationships between cover types and nutrients should be interpreted by projections onto Axis 1 since only this axis has significant correlations to cover ordination space (p=0.01) and a significant cover type-environment correlation (p=0.02).   and give access to the salt marsh surface for foraging ). Fish also use such small places as mussel shells and basal Spartina alterniflora leaves for egg deposition (Daiber 1982, Able and while larvae can grow in shallow puddles on the marsh surface . Along this continuum of nekton habitat are salt marsh pools, soft-bottomed depressions that hold water throughout a tidal cycle and do not dry out like shallower pannes . New England salt marsh pools are steep sided  with an average total depth of approximately 30 cm (Adamowicz, Chapter 1). A survey of 26 marshes from Connecticut to southern Maine showed that pools range in size from 1 m 2 to 35 ha and cover an average of2.3% of ditched marshes and 9.1 % of unditched marshes (Adamowicz, Chapter 1).
Pools are important areas, supporting nekton (Able andFahay 1998, Raposa 2000) and foraging birds . While detailed studies on nekton use of pools have been conducted in neighboring regions (Quebec, FitzGerald and Wootton 1993; New Jersey, Able 1990) New England studies have been limited in scope . Work regarding the range of pool conditions throughout the region, especially with respect to nekton abundance, is lacking. 75 Nekton, i.e. fish and decapod crustaceans, in salt marsh pools is not only a measure of secondary production but also an important prey source for wading and other birds , and provides a trophic link to other ecosystems Ross 1982, Kneib 1986). Variables influencing nekton species distribution and abundance in salt marsh pools include water quality (Audet et al. 1986, Poulin and, pool size (MacArthur and Wilson 1967), water depth , the presence of macrophytic vegetation , and geographic conditions such as elevation ) and distance to tidal flow .
Open water marsh management (OMWM) techniques employed throughout New England include the construction of new pool habitats. Nekton in the pools, however, has not been sampled in a comprehensive fashion to provide a baseline for establishing management goals. The purpose of this study is to provide a New England benchmark for nekton species composition, distribution, and relationships with environmental variables in order to better understand the role of pools in overall salt marsh ecosystem functioning and to provide guidelines for management efforts.

Nekton was sampled at 3 locations in Maine and 3 locations in southern New
England during 1999 ( Figure 1). Nekton sampling continued at Granite Point A and Marshall Point (ME) during 2000. Species composition and abundance of nekton 76 (fishes and decapods) were sampled from May to October in marsh pools using a lm 2 throw trap . The dimensions of the throw trap were 1 m 2 x o. 5 m high and were similar to traps used elsewhere . Trapping efficiencies were estimated at 70 to nearly 100% by  and .
Trap construction and sampling technique followed that of . Trapping was initiated once high tide had receded from a marsh and fish were restricted within pools. If a site had 24 or fewer pools, they were all sampled; otherwise 25 -29 pools were randomly selected. Numbers of pools sampled at each site are given in Table 1.
Samples were obtained by slowly crossing the marsh surface to a randomly selected station on a pool ' s perimeter then tossing the trap 3 -4 m through the air into the water while still at a distance from the pool edge. The bottom of the trap then was pushed into the sediment to ensure that no animals could escape. All animals were removed from the trap using a Im x 0.5 m dip net (3 mm mesh) that fit snugly into the trap . Dip-netting was conducted from at least 3 sides of the trap . Traps were considered empty when 3 consecutive uses of the dip net obtained no nekton. All 77 captured animals were identified to species, measured (total length for fish, carapace width for crabs), and immediately released . Whenever a species was present in large numbers, total length was measured on a subset of 15 -30 individuals. One hundredsixteen pools were sampled in 6 marshes in 1999. Sampling was restricted to 2 sites in intended only to provide an initial estimate of regional pool conditions. Water depth was measured with a meter stick from at least 3 sides of the throw trap . The presence/absence of macrophytes, particularly Ruppia maritima, was also recorded .
In addition to naturally existing pools, a number of pools that were created as part of ditch-plugging/marsh-restoration efforts were sampled. At Granite Point B in Biddeford, Maine, contractors excavated pools of 2 sizes (3 and 9 m diameters) at three different distances (15 , 30, and 50 meters) from tidal flow. This scenario was replicated at 3 locations on the marsh. The purpose of this design was to test hypotheses concerning the influence of pool size and distance from tidal flow on use. Nekton in the created pools was also sampled with a lm 2 throw trap . Pool measurements, water quality data, and the presence/absence of macrophytes also were recorded as noted 78 above. Data from these experimental pools were analyzed separately from the natural pools.

Statistical Analyses
Differences in nekton species richness between the northern and southern sites were determined by t-test Gack-knife estimation following Heltshe and Forrester 1983).
Regional differences in nekton community were determined by non-parametric tests of similarity (PRIMER, . Differences in nekton abundance and length within a marsh over time were ascertained by t-test and ANOVA as appropriate. The association of species to environmental variables was analyzed through canonical correspondence analysis (CCA).   locations but were not included in the analysis due to potential changes in fish behavior as water temperatures declined . All fish densities were fourth-root transformed with rare species ( < 10 individuals) removed as recommended by  Fisher' s Exact Test  was used to indicate whether the presence of nekton in one pool was linked to the presence of nekton in the nearest neighboring pool (also known as "contagion"). The association of nekton with submerged aquatic vegetation (specifically R. maritima) was determined by use of the rank-assignment/ANOVA equivalent of the Kruskal-Wallis test (SAS 1990).
In order to determine if specific pool physical or water quality conditions were associated with either high, medium, or low nekton abundance, fish density data were analyzed first by ANOV AILS Means to determine whether there were pools with densities that remained consistently high (or low) from month to month. For marshes that had such fish density patterns, LS Means testing was used to assign pools to a category of low, medium, or high fish density. Stepwise logistic regression then was applied to environmental variables measured at each pool in order to build a model predicting fish density category. This technique was applied to each site and year combination. 80 Stepwise multiple linear regressions were used to evaluate which environmental conditions were associated with the greatest F heteroclitus density and length at each site and sample date. This analysis was limited to F heteroclitus in order to control for interspecific behavior and the species was further sub-divided into juveniles, adult males, and adult females to control for intraspecific behavioral differences . Best-fit equations presented were those with overall p<0.05 and with significant coefficients Granite Point B). Separate step-wise multiple linear regressions were also performed at 4 sites lacking elevation data and for comparisons sake, at the 3 sites where elevation data were available but were excluded from the regression analysis (Table 7).
The effect ofbiogeographical factors, such as pool size and distance to tidal flow, on nekton abundance was determined by t-test and ANOV A Differences in abundance between created and natural pools also were identified by t-test.

Comparisons among Marshes and over Time
Nekton Community Characterizations. Throw-trap sampling in salt marsh pools resulted in a range of 3 to 6 fish species per site, depending upon location ( Table 2).
Fundulus heteroclitus was the most geographically ubiquitous species, occurring at every site. The least common species were F. majalis and Brevoortia tyrannus, each occurring at only one site. A total of 12 different fish species were sampled throughout the study (May 1999-October 2000; a complete list of scientific and common names is given in Appendix I. Jack-knife estimates of species richness  during September (the only month when all marshes were sampled) indicated greater richness (7.0 ± 0.97 species) for the southern marshes compared to the northern marshes (5 ± 0.94 species, t-test, p<0.005). In addition, fish communities were significantly different between the northern and southern sites (p<0.0001 , ANOSIM).
Percent similarity analysis (SIMPER) indicated that southern sites possessed more F.
heteroclitus and that Cyprinodon variegatus and Lucania parva occurred only in the southern marshes. These three species accounted for 95% of the total dissimilarity between the two regions.
During the study period, a total of 5,820 fish were caught with F. heteroclitus  Table 3).
Fish Length. Average total lengths for species with significant differences between sample dates are given in Decapod Size. In distinct contrast to fish patterns of increasing length through the sampling period, C. maenas at Granite Point A increased in both years to a peak size during September followed by a marked decrease in size during October (Table 4). The absence of large adult C. maenas was very obvious as was the presence of particularly small juvenile crabs.  transformed data, t-test). There were no significant differences in C. maenas carapace widths between natural and created pools (p>0.10, t-test).

Canonical Correspondence
Step-Wise Multiple Linear Regression.
High juvenile F. heteroclitus density was most commonly associated with measures of water depth, salinity, and relative pool elevation (Table 6). Greater juvenile densities during May -September occurred in pools that were shallower and higher in elevation; in October, juvenile densities were greater in deeper pools that were lower in elevation. Salinity correlations with juvenile densities were mixed. F. heteroclitus adult densities were associated with temperature or pool size in approximately 50% of the cases reported in Table 6. Results for adult females suggest a switch to warmer waters in September/October.
When elevation data were absent, F. heteroclitus juvenile densities were most often modeled by salinity, pool size, distance to tidal flow, and water depth ( Table 7).
The best-fit equations predicted more F. heteroclitus juveniles in deeper pools in October; in other months juveniles were more prevalent in shallower pools. F. heteroclitus adult densities most frequently were correlated with temperature and pool size. For F. heteroclitus males, density was greater in smaller and cooler pools (June, September). F. heteroclitus female abundance was also greater in smaller pools overall, with those pools being warmer in October (temperature trends for other months varied).
In the instances when elevation data were available but ignored, log pool surface area (in the case of adults of both sexes), distance to tidal flow Guveniles and males), and temperature Guveniles and males) entered into the regression equations. 89 When stepwise multiple regressions were performed on fish lengths for each class (Table 8, includes elevation data), in over 70% of the cases, larger juveniles were associated with temperature -similar to the frequent correlations between adult densities and temperature. In general, larger juveniles were correlated with lower elevation pools in all months, higher levels of dissolved oxygen, cooler pools during the summer, but warmer and deeper pools in October. Large adults tended to track pool size and dissolved oxygen levels.
When examining F heteroclitus lengths at all sites without elevation data (Table   9), larger juveniles correlated with measures of water temperature and dissolved oxygen in the majority of cases. Large adults tended to track salinity and dissolved oxygen.
When elevation data were available but excluded (Table 9), log pool surface area (for juveniles) and distance to tidal flow (adults) entered into the regression equations.

DISCUSSION
Nekton use of salt marsh pools is best examined on both a large and small scale.
Large-scale views allow comparison among marshes and regions and reveals seasonal trends. Small-scale views address comparisons among pools within a marsh.

Large-scale Patterns
Salt marsh pools were dominated by F heteroclitus and a small assemblage of species common to salt marshes ( Table 2). These findings are consistent with studies of other New England salt marshes ) except that pools represent a subsample of the species caught in other 90 marsh habitats. This subsampling is consistent with the data of  in New Jersey and thus suggests ecosystem rather than local or regional processes.
Pool nekton communities were distinctly different between northern and southern New England. The greater species richness and density in southern New England exemplify the biogeographical break known to occur around Cape Cod, MA for many estuarine and marine species . While examining a number of studies (e.g.  can elucidate this regional division, this is the first investigation of salt marsh pools that has used the same sampling technique in both northern and southern New England. In terms of seasonal trends, several nekton species were shown to increase in density and size from May to October ( Figure 2, Table 4). Others have noted these trends within New England ) and further south . Rather than a steady increase in size, however, green crab (C. maenas) carapace width increased only from May to August (  . It may be that while pools are bare, fish find adequate refuge from predation (a trait associated with macrophyte cover) by burrowing in the soft sediments of pools.
Next, ANOV A/logistic regression results indicated that only one site had pools with consistent fish density levels in each pool (Granite Pool A, smaller, deeper pools had the highest fish densities). When pool size was evaluated in the experimental pools, there was no effect on fish density. It is possible that treatment levels were not extreme enough and subsequently did not affect fish colonization. This in itself is of interest since the "large" created pools (at 254 m 2 ) were nearly 4 times the median size (64.5  , however, did find differences in fish density based on distances to estuarine waters, although they used increments of 75 m -much broader than the area available for experimental manipulation at Granite Point B. Despite these negative results, two overall patterns were observed in nekton distribution among pools. Fisher' s Exact Test revealed contagion of fish among pools (p<0.01). The demonstrated mobility of F. heteroclitus ) and relative proximity of nearest-neighboring pools (Adamowicz, Chapter 1) ensured their spread among pools. Additionally, when comparing newly created pools to existing ones, fish density was more than twice as great and F heteroclitus were 5 -7 mm larger in old pools (Table 5). Similar results were obtained for decapods. These findings may be linked to the poor environmental quality of newly excavated poolsduring the first sampling period dissolved oxygen levels averaged 4.2 mg/I(± 1.8 mg/I) in new pools compared to 7.8 mg/I(± 3.0 mg/I) in old pools.   estimated that these fish could move 32 m across the marsh surface.

Comparison of Environmental Variables among Pools within a Sample
Thus F. heteroclitus has the physical capability of traversing the distance between pools and should be able to encounter a number of pools while moving about when the marsh is flooded at high tide. Stepwise multiple regression analyses demonstrated that larger juveniles were correlated with better habitat conditions (e.g. higher levels of dissolved oxygen and lower water temperatures in summer, Table 8). One could argue that instead of fish choice, better habitat conditions facilitated greater fish growth. While this is possible, the definition of optimal conditions changed from month to month (Table 8), as did the pools that fit the description. So if larger juveniles were always in pools with the best environmental conditions, they would have to move over time.
Perhaps the strongest evidence of active site selection by larger fish is demonstrated later in the season when F. heteroclitus seeks pools (rather than creeks) as overwintering habitat Able 1994, Raposa 2000). This study goes one step further, however, indicating that F. heteroclitus selected among pools (Tables 6 -7).
Nor is pool choice unique to F. heteroclitus. Adults of three species of sticklebacks (Gasterosteidae) were shown to avoid pools that dried out and to settle more often in pools that retained water, suggesting active habitat choice .

Role of Pools In Nekton Life Histories.
While fish may select among pools, why should they occupy pool habitat at all? This study has documented that pools can be devoid of vegetation and exhibit extremes in temperature and dissolved oxygen (Figures 7 -9).
94 Such stressful environments seem unlikely places to support growing nekton populations. Several studies, however, have documented that food, at least, is abundant in pools ).  work indicated that protection from predation is an additional incentive for pool use by juvenile fish. Adult F. heteroclitus foraging on flood tides restricted juveniles to the high salt marsh. When adult fish retreated to creeks with the ebb tide, juveniles remained on the marsh except in those cases where adults had been experimentally removed. In the latter circumstances, juveniles also went to the creeks.
This separation of adults and juveniles was confirmed in the present study by the high relative abundance of juveniles in marsh pools (Table 3) and by larger (more predation resistant) juveniles being found in pools at lower elevations closer to open water (Table   8).  obtained similar results with shoreline fish; small fish occurred in greater numbers in shallow, unvegetated waters, in contrast to other studies where high fish densities were associated with submerged aquatic vegetation (Heck et al. 1989, Rozas and. Thus the preponderance of F. heteroclitus juveniles in pools was likely due to their use of pool habitat as a refuge from predation. As fish grew, they were able to choose among pools for improved habitat and then leave them for tidal creeks or other nearby waters when size and tides allowed, only to return to pools in fall for overwintering. Other Species. While F. heteroclitus was the dominant nekton species, it was not the only one sampled in salt marsh pools; the other eleven fish and 4 decapod species sampled provide a variety of insights to pool habitat quality. For example, three-and four-spine sticklebacks (Gasterosteus aculeatus, Apeltes quadracus, respectively) were sampled only intermittently in pools at several study sites, while P. pungitius was more common (Tables 2 and 4).  found P. pungitius to be a pool resident unlike other sticklebacks, which stayed only for 1 tidal cycle to spawn ). The longer residence time of P. pungitius was thought to be due to its greater resistance to low oxygen concentrations of July and August . This corroborates the correlation of P. pungitius to lower dissolved oxygen levels in the CCA biplots ( Figures 5 and 6).
Several other fish were sampled regularly but in much reduced numbers.
Cyprinodon variegatus were found only in southern New England sites, corresponding to the reported range of Cape Cod, Massachusetts to Florida (Robins and Ray 1986).
CCA results indicated C. variegatus densities were correlated with larger, shallower pools ( Figure 4). The association of C. variegatus with these particular circumstances despite the availability of more moderate pool conditions could be an indication of either competitive displacement or predation avoidance as noted above for F.

heteroclitus.
Menidia menidia are common to marsh creeks and are intolerant of low dissolved oxygen levels (Able and Fahay 1998); therefore, their presence in pools throughout the summer may represent strandings. Menidia beryllina are known to use pools as nurseries , though only larger young of the year (Table   96 4) were sampled in pools during September. Of the two Menidia species, M beryllina is more common to fresh/ brackish waters  Incidental species included Anguilla rostrata, B. tyrannus, F luciae, and F majalis. F luciae is a high marsh species often located in very shallow pools and puddles . Although reported only as a stray in Connecticut (Whitworth I 996), they were regularly sampled at Barn Island.

Restoration Recommendations
Restoration of salt marsh pools should proceed with caution and particular attention to detail. For example, the pool characteristics associated with high F heteroclitus juvenile densities from May -September (shallower pools in higher elevations) were not the same as those in October (deeper pools, lower in elevation) (Table 6). Additionally, C. variegatus and F luciae preferred shallow high-marsh pools whereas sticklebacks required deeper pools that were less likely to dry out                 A. q_uadracus -- England shoreline . They are considered one of the most productive ecosystems in the world  and are thought to be important to estuarine and nearshore production , although see Nixon 1980. Salt marshes, however, are also known as mosquito breeding areas and the serious diseases that mosquitoes sometimes carry have made their control a public policy issue. With the aid of the Civil Works Program during the 193 0' s, nearly 90% of the coastal marshes from Maine to Virginia were ditched in order to drain mosquitobreeding areas . Some ditching was also done during Colonial times and afterwards to facilitate salt hay farming .

"""
Ditching has affected marshes far beyond the original intents of agricultural improvements and mosquito control.  in their study of a Connecticut salt marsh gave this scathing review, " .. . ditching, at best a violent activity which, though it destroys the mosquitoes, also destroys the permanent pools so valuable to wildlife, completely rearranges the mosaic of natural plant communities, and eventually produces other pools of the same kind that the ditches were designed to eliminate. " Adamowicz (Chapter 1) quantified the high correlation between ditching intensity (m of ditch/ha marsh) and reduction of salt marsh pools.  141 and  reviewed numerous articles describing the deleterious effects of ditching on wildlife ranging from invertebrates to birds.
Despite the difficulties attendant with ditching, alternative mosquito control measures such as the application of pesticides, have proven less desirable. It was in this milieu that another method, Open Marsh Water Management (OMWM), came into use during the second half of the last century . It was determined that OMWM could be used not only to control mosquitoes by providing larvivorous fish access to mosquito breeding areas, but also to enhance fish and wildlife habitat . New England salt marsh complexes  with Spartina patens, Juncus gerardii, and Distich/is spicata in the high marsh, while Spartina alterniflora occupies low marsh positions.

BACI Study Design
Each study site was divided into a control and experimental area as part of a BACI (Before-After-Control-Impact) study design ). Control  Table 1.

Pool Sampling
Salt marsh pool sampling included measurement of pool physical parameters, water quality conditions (salinity, temperature, dissolved oxygen), and nekton species composition and abundance. Nekton was sampled at salt marsh pools using a l-m 2 throw trap with 3mrn mesh sides . Trap construction and sampling technique followed that of .
Trapping was initiated once high tide had receded from a marsh and fish were restricted within pools. Samples were obtained by slowly crossing the marsh surface to 144 a randomly selected station on a pool's perimeter then tossing the trap 3-4 m through the air into the water while still at a distance from the pool edge. After landing in the pool, the trap was rapidly pushed further into the sediment so as to prevent animals from escaping from under the bottom of the trap frame. All animals were removed from the trap using a lm x 0.5 m dip net (3mm mesh) that fit snugly into the trap. Dip netting was conducted from at least 3 sides of the trap. Traps were considered empty when 3 consecutive uses of the dip net obtained no nekton. Captured nekton were identified to species, measured (total length for fish, carapace width for crabs), and immediately released. Variables recorded included species, abundance, and individual organism lengths (up to 30 individual for each species). Water quality measures obtained with a YSI-85 hand-held dissolved oxygen meter included dissolved oxygen (mg/I), salinity (ppt), and temperature (C). Water depth was recorded from 3 sides of a deployed throw-trap using a meter stick. Pool size (m 2 ) was estimated by measuring the major and minor axes ofregularly shaped pools ) and applying the formula of an ellipse. Irregularly shaped pools were subdivided into simple geometric shapes, with appropriate data taken to calculate surface area. All pool sizes were remeasured following ditch plugging and whenever water levels expanded or contracted the pools from their well-defined margins. Distances of pools to nearest tidal flow Differences in nekton community were analyzed using the PRIMER nonparametric randomization tests of similarity (Clarke and Warwick 1994, Roman et al. in press) applied to a similarity matrix calculated from the Bray-Curtis similarity index.
An analysis of similarities test (ANOSIM) was used to detect differences between groups of pools in the BACI design ("Before-Control, After-Control, Before-Impact, After-Impact) for each site. Contributions of individual nekton species to dissimilarities between groups were calculated using percent similarity tests (SIMPER). It should be noted that for all the BACI comparisons, regardless of statistical test used, the data sets were limited to containing sampling from the same months in both years (May, June, September, October).
At Granite Point and Moody marsh, changes in nekton densities and sizes were assessed with ANOVA with least squared means (LSMeans, SAS 1985) post-hoc testing of the interaction term (site x time). In a BACI design, there is a time element between "Before" and "After." One might argue that the difference between "Before" and "After" was simply due to time. Thus the interaction term was used to examine the change from "Before" to "After" at the control area and compare it with the change from "Before" to "After" for the impact (experimental) area. If the interaction term was not significant, then natural variability was considered the cause of change rather than the site alterations.
Testing the interaction term also was used to compare physical (e.g. water depth, pool size) and water quality data (e.g. salinity, dissolved oxygen). Nekton densities 146 were square root transformed(.../ (x+0.5)) and pool sizes were log transformed (log (x+ 1 )) prior to analyses in order to achieve normality .
A comparison of species richness among the BACI groups for each site was made using the jack-knife estimate procedure of . In order to test for the interaction term a modified t-test was used as noted below: t where She is the jack-knife estimate of species richness for the "Before-Control" group, (similarly the subscripts bi = "Before-Impact"; ac ="After-Control"; ai= "After- Clearly it would have been preferable to GPS the study areas and their pools in conjunction with the nekton sampling.

Total Fish Abundance Calculations
Changes in total fish abundance in the "Before" and "After" experimental areas were calculated two ways. First, the average fish density for all pools was multiplied by the total surface water area (pools and plugged ditches) based on September 1999 and 148 September 2000 nekton data at Granite Point and Moody marsh. September data were used since it represented the peak summer abundance. Surface water area used was that calculated from aerial photographic ("Before") and GPS ("After") mapping. Resulting "Before" -"After" differences in total fish abundance were expressed as a percent increase or decrease relative to the total fish abundance of both years. Second, the same method (surface water area x average fish density) was applied to all sample months during the "Before" and "After" periods. Monthly averages from 1999 and 2000 were subjected to paired t-testing. Data at Marshall Point was similarly subjected to paired ttesting, although groups were control versus experimental areas paired by month for 2000 data. All data were log (x+ 1) transformed prior to analysis to achieve normality .

Throw-trap sampling was initiated in May 1999 with 4 sample periods through
October, resuming in May of 2000 with 5 sampling periods through October of that year. A total of244 throw-trap samples was taken in 1999 and 395 in 2000 (Table 1).
Seven fish and 3 decapod species were sampled in the two years ( Table 2). Of the 7822 fish caught over the course of the study, 89% were F heteroclitus (mummichog); the next most common fish was P. pungitius (nine-spine stickleback) at 7%. Apeltes quadracus (four spine stickleback), Anguilla rostrata (American eel), and Brevoortia tyrannus (menhaden) were the least common fish species. Among the decapods, Palaemonetes pugio (daggerblade grass shrimp) was most common overall accounting 149 for 54% of the 666 decapods caught. A complete list of scientific and common names is given in Appendix Table I.
"Before" and "After" Comparisons at Granite Point and Moody Marsh At Granite Point, the nekton community remained unchanged in both control and experimental area pools from before to after ditch plugging (p>0.05, both, ANOSIM). The same results were obtained for Moody marsh (Table 3). Similarly total fish and total decapod densities at Granite Point were not altered due to ditch plugging (p>0.05 , both; ANOVA). Again, the same results occurred at Moody marsh (Table 3, Appendix II).
Despite these similarities between "Before" and "After" conditions at the two study sites, there were some differences. Species richness decreased by 3 species at the Granite Point experimental area (Bonferroni adjusted alpha 0.025, p<0.001 , Table 3) despite an increase of 5. 9 species at the control area. At Moody marsh there was no change in species richness due to ditch plugging (p>0.05 , Bonferonni adjusted alpha 0.025,  (Appendix II). At Granite Point, the only individual species that changed from the 150 "Before" to "After" condition was P. pugio which decreased significantly at pools in the control area (p<0.05, ANOV A, Appendix II).

Marshall Point Control and Experimental Area Comparisons
Since the Marshall Point site was ditch plugged prior to this project, it is only possible to compare the experimental and control areas for the 2 years following plug installation. Additionally, since all the pools at the experimental site were created, it is assumed that whatever nekton occurred in created pools was a result of the site alteration.
Nekton communities between the control and experimental areas began the same, but differed over time. In 1999, the nekton community was the same between experimental and control area pools (p>0.05, ANOSIM) but not in 2000 (p<0.05 , ANOSIM). The species most responsible for differences in the experimental and control area in 2000 were F. heteroclitus (39%), C. maenas (22%), and P. pugio (8%) in the experimental area and the sticklebacks P. pungitius (20%) and A. quadracus (8%) in the control area (Table 4).
Experimental area nekton communities also differed between 1999 and 2000 (p<0.01 , ANOSIM). The majority of the difference (57%) was due to greater F.
Marshall Point control and experimental areas also differed in almost all other measures of nekton use (  (Table 8). Nekton densities were also greater at old pools for 7 of 8 species where there were differences (Granite Point, Table 9).
Interestingly, there was only one difference in species density at Moody; C. maenas density was greatest at enlarged pools (p<0.05 , Table 9). In terms of nekton size, only F heteroclitus juveniles and F heteroclitus overall differed among pool types, being larger in old pools at Granite Point (Table 10).
While attempts were made to control for pool size and distance to tidal flow, at Granite Point, enlarged pools were both larger and further from tidal flow than old or new pools (p<0.001 , p<0.01 , respectively; Table 11). The exceptional average size of enlarged pools was due in part to the fact that contractors expanded already large pools. 153 At Moody marsh, there was no difference in pool size among the three pool types; however, old pools were 12.2 m further from tidal flow than new pools (p<0.01 , Table   11). Interestingly, at both sites new pools were 15 -20 cm shallower compared to old or enlarged pools (p<0.001 for both sites).

Habitat Availability and Total Fish Abundance
Maps of pre-existing and post-plugging surface water are provided in Figures 2 -6. The total area of surface water (pools and plugged ditches) for each site and study area are listed in Table 12. Percent of marsh occupied by surface water increased for the experimental sites following ditch plugging (p<0.05, t-test), while the control sites remained the same (p>O. 0 5, t-test). Plugged ditches constituted the majority of new surface water at each site followed by new pools and enlarged pools (Table 13).
Total fish abundance (average fish density per month x total surface water habitat area) did not vary from Before to After for either Moody marsh or Granite Point.
Paired t-tests (sample months in each year as a pair) did not result in differences for "Before" versus "After" conditions at the Moody experimental area (p>0.05 , n=4) and there was no change in control area pools over the same period (p>0.05, n=4 Using just September fish densities, however, total fish abundance at the Moody experimental area declined by 12% from the "Before" to the "After" condition. In contrast, total fish abundance was 13% greater in the "After" (versus "Before') condition at Granite Point experimental area pools. Over the same time, total fish abundances increased from 9 -11 % at the Moody and Granite Point control areas, respectively.

DISCUSSION
The success of ditch plugging efforts should be assessed using a variety of indicators. Nekton use is one category of such indicators and is particularly beneficial on two levels. First, nekton has been shown to respond quickly to salt marsh alterations . Second, nekton use integrates a number of other factors such as water quality and food availability . Thus initial levels of nekton use can be used as an early indicator of habitat function .

Nekton Community and Density Measures
Given the rapid response of nekton in other salt marsh projects where tidal flow had been restored , it was 155 surprising to find very little change in nekton use of pools at Moody marsh and Granite Point following ditch plugging. At these two Maine sites nekton community, total fish density, and total decapod density (Table 3) were not influenced by ditch plugging. A possible explanation for this apparent inconsistency may lie in the different types of marsh alteration -restoration of tidal flow versus ditch plugging.
Upon restoration of tidal flow to a marsh where it had been severely restricted,  found that F. heteroclitus followed the leading edge of water into the most upstream portions of newly created tidal channels. In this case it was both the restoration of tidal flow and fish behavior that resulted in increased nekton use in the recovering marsh.
Experimental area pools in this study, however, differ from the tidally restored marsh just described in 2 important ways. First, except for distance to tidal flow, average environmental conditions (dissolved oxygen, salinity, temperature, water depth; Table 6) across experimental area pools did not change as a result of ditch plugging.
Secondly, not only were nekton present in these pools prior to ditch plugging, but significantly, the dominant species, F. heteroclitus, overwintered in the pools and thus already were present (unaffected by ditch plugging) and able to spawn and produce the juveniles that were sampled in the following weeks and months. While these two factors can account for a lack of an initial change in nekton use of pools, several nekton parameters (species richness, F. heteroclitus juvenile density, and total fish abundance) did vary indicating effects that may become more prominent over a longer span of time.

Species Richness
While at least some F. heteroclitus successfully overwintered in pools and provided a starting population for the post-plug sampling period, other species are known to overwinter in deeper waters outside the marsh . Due to the relatively low numbers of these other species, impacts associated with ditch plugging would not necessarily be seen in density measures (  (Table 3). These opposing trends can be resolved by understanding how fish gain access to salt marsh pools and by comparing alterations at both sites. Just as other studies have documented a rapid increase in nekton following removal of tidal barriers  in press), the rapid decrease in species richness at the Granite Point experimental area may be a result of imposing tidal barriers. The ditch plugs used in Maine salt marshes were up to 0.3 m above the marsh surface and extended across the marsh for several meters by small berm "wings. " Thus ditch plugs created a physical barrier to nekton movement onto the marsh in addition to their designed capacity to retain water.
Neill and Turner (1987) and  documented that canal plugging and low-level weirs in Louisiana decreased fish movement to and from marshes. Both obstructions resulted in decreased numbers of migrant species using the marshes and could account for the declined species richness at Granite Point. But why wasn't there also a decline at Moody? In the Louisiana papers, the canal plugs and weirs were placed at the primary access points where tidal flow (and fish) entered the study areas.
At the Maine study sites, each marsh had a long interface (through ditches, creeks, and shoreline) with tidal flow.
Since fish gain access to the salt marsh surface (and thence to pools) through creeks and channels on flood tides   In fact, that total fish abundance did not increase either at Moody or at Granite Point despite the increase in habitat area (Table 12) suggests that the new habitat areas could be "diluting" the existing population offish on the marsh. Nekton use of plugged ditches was not evaluated in this study, but would be a useful addition to pool sampling to determine more definitively whether ditch plugging increases total fish numbers on the entire marsh. This is a particularly important matter since one of the motivations of the marsh project was to enhance bird foraging habitat. Steady or decreased total fish abundances despite increased surface water habitat could result in increased foraging effort by target bird species.

Marshall Point
At Marshall Point, control area pools lay far from the creeks and ditches. In contrast, the experimental area was fairly narrow and bisected by a large, unobstructed tidal creek (Figure 4). While the experimental area was narrow to start with, this creek restricted placement of new pools to areas adjacent to ditch plugs -only a short distance from tidal flow (Table 7). Given both the small distance to tidal flow and the large bisecting creek (a nekton source), it was not surprising to find greater species richness and densities in experimental area pools compared to the control.
Pool environmental conditions also differed markedly between the experimental and control areas. Water temperatures were cooler and salinities were higher at the experimental area (Table 7) reflecting proximity to tidal flow. These factors together may have been responsible for the differences in nekton community (ANOSIM, Table   7) between the control and experimental areas.

Effect of Pool Type on Nekton Use and Survival
Since post-alteration nekton sampling proceeded almost immediately after construction of ditch plugs at Granite Point, this study was able to capture the earliest stages of nekton response. At Granite Point, old (naturally existing) pools were better nekton habitat as indicated by a number of nekton use parameters (species richness, nekton density, and nekton size; Tables 8, 9, 10). While not directly measured, habitat conditions in new pools may be harsh. Excavation of new pools exposes anoxic peat to oxic conditions, prompting radical changes in redox conditions, pH, and chemical states of minerals such as sulfur and iron . Being relatively shallow (Table 11 ), new pools are subject to extremes in temperature. New pools also have no vegetation in contrast to old pools, which have some amount of filamentous green algae or macrophytes (Adamowicz, Chapter 2).
At the experimental area in Moody marsh, however, several weeks had passed between ditch plugging and the first nekton sampling. This evidently was a long enough period to allow amelioration of initial, potentially harsh conditions in excavated 160 pools so that overall there was no significant difference among pool types. Marshall Point sampling began several months after site alteration and revealed that the experimental area had significantly more nekton use than the control area (Table 5).
Thus within a year's time, nekton use of new pools at Moody marsh and Marshall Point had equaled or exceeded use of old pools.
Fish not only occupy newly restored sites (including pools), but as other studies have documented with gut content analysis, they are able to find food there . Pool fish feed on a variety of planktonic organisms Fitzgerald 1989, James-Pirri et al. 2001), which are readily conveyed by flood tides. Thus the successful establishment of nekton communities in salt marsh pools can occur rapidly and represents the integration of 2 highly mobile trophic levels.

Future Management Recommendations
Evaluating the success of habitat enhancement requires determining to what extent nekton use changes over time. Marshall Point experimental pools have already demonstrated some changes in nekton community (Table 4), individual species densities (Appendix II), individual species size (Appendix III), and pool size (Appendix IV) during the 2 years since their creation.  have documented that different marsh functions return at different rates. Future monitoring would indicate whether the initial responses recorded here were short-lived phenomena or indicators of long-term trends . Additionally, nekton use of plugged ditches must be assessed since they form a significant percentage of the total permanent surface water created on the marsh site. Project success should also be evaluated in 161 terms of trophic exchanges, i.e. whether nekton productivity transfers to target organisms are enhanced or hindered in the newly created habitats. Finally, changes in sediment geochemistry and elevation levels should be closely monitored given the findings of  and in light of the severe consequences ofrising sea level.

CONCLUSIONS
Ditch plugging is an intensive form of habitat enhancement. At Moody and Granite Point there were few initial differences between the "Before" and "After" conditions of nekton use at experimental area pools that could be attributed to the plugging itself New pools, however, were not as good habitat as old pools, at least at the outset. Alterations across plugged areas (i .e. decreased species richness, decreased F. heteroclitus juvenile density, and increased distance to tidal flow) deserve further investigation. All that said, ditch plugging did increase surface water on the marshes and thus provides the potential for an increase in total nekton abundance. Such was the case at Marshall Point where nekton use did increase at the experimental area.
Managers, however, should carefully balance potential direct and indirect effects of ditch plugging on a site-by-site basis prior to adopting it as the best way to amend the effects of drainage ditches.