Lower Bay Complex
and long-term temporal community structure of the benthic fauna in the Lower Bay Complex. I will try to characterize communitymembership ingeneral terms, attempt to relate community structure to the physical environment, and finally, describe how community structure has changed over time. In addition, I will describe the process of analyzing large faunal data sets and assess the suitability of the techniques commonly used.
Unfortunately, no two regional studies used exactly the same methods (Table 18.1). There are differences in sampling locations, season of collection, sampling device, screen sizes used to sieve the samples, and taxonomy (since different taxonomists do not always identify an organism the same way). Corrections for differences inmethods do not exist, and the differences can have a substantial impact on results. Both Diaz and Boesch (1979) and Berg and Levinton (1984), in trying to compare Dean
(1975) and McGrath (1974), noted that differences in sampling locations, season of collection, and sieve sizes limited the conclusions that could be reached. Thus, any approach to comparing the regional benthic surveys must be able to overcome differences in sampling methods.
Because of large differences in methods, I will not consider Stainken et al. (1984) and Adams et al. (1998) in detail. The Stainken et al. (1984) study enumerated only very large benthic animals, and it did not provide results that can be reasonably compared to other studies. Adams et al. (1998) sampled randomly, and there was little overlap in locations between years or with other studies. To circumvent other sampling differences, I will take advantage of three factors. The survey by Cerrato et al. (1989) utilized methods that made it comparable to most of regional surveys that preceded it. It is also a good match to NOAA-USACE (2001) where both season and locations overlap, leaving only sieve size (0.5 mm vs. 1 mm) and identification differences. Secondly, species will be assigned to functional groups to reduce the effect of taxonomic differences. Finally, I will use a method called a Mantel test to compare the data sets in an indirect way that should be less sensitive to differences in methods.
Data from the regional studies consist of counts of individuals for each species present in the samples. Since there are hundreds of species and in some of the studies more than a hundred sampling locations involved, these data sets are unwieldy to examine in raw form and must be summarized in a way that allows community structure analysis. As a first step to examining large data sets, investigators often identify species that were representative of the diverse life histories present and good indicators of community structure and community change. In the present study, I assembled species lists from each regional survey using a variety of criteria and then formed a composite list. Criteria included identifying numerically abundant, cosmopolitan, and commercial species, species with interesting life history attributes, those sensitive to anthropogenic stress, important prey species, and those contributing most to community structure based on the multivariate analyses described below.
In addition to analyzing the data at the level of individual taxa, species in each data set were also assigned to functional groups on the basis of similar lifestyles. Tabulating a large taxonom-ically diverse benthic assemblage into functional groups loses information since the abundances of many species are added together. Functional groups do, however, reduce taxonomic discrepancies between studies and considerably reduce data sets to a smaller number of ecologically meaningful descriptors. Criteria for assigning species vary but it seems reasonable to consider the animal's primary feeding mode combined with whether the organism was infaunal or epifaunal, whether it constructed a tube or was free living, and whether it was mobile or sessile. These criteria merge two prior attempts (Woodin and Jackson, 1979; Fauchild and Jumars, 1979) at classifying the marine benthos in terms of similar lifestyles.
Most ecologists summarize large data sets for further analysis by calculating an index of ecological resemblance or association between the sampling sites or species present (Legendre and Legendre,
1998). In this study, the Bray-Curtis index was used:
D_ E"=1 1*1 j - X2 j I Yl"j=1 (x1 j + x2 j)
If the benthic communities at two sampling locations are being compared, the xij are the abundances (usually root or log transformed to decrease the influence of dominants) of each species present at each location, and the approach is called normal analysis. If associations between two species are being compared, the xij are abundances (again usually transformed) of the species at each sampling location in the study, and the approach is called inverse analysis. The Bray-Curtis indexvaries between 0 and 1, with 0 representing perfect ecological resemblance, and 1 being no similarity. When calculated for all pairs of sampling stations or species, the values can be assembled together to form a matrix of index values, called an ecological resemblance matrix.
Ordination and cluster analysis are two common, multivariate methods used to visualize relationships contained in an ecological resemblance matrix (Field, Clarke, and Warwick, 1982; Legendre and Legendre, 1998). Ordination attempts to plot the sampling stations or species in two- or three-dimensions such that the distances betweenpoints are related to the values in the ecological resemblance matrix. Cluster analysis combines stations or species into groups based on the similarity of their ecological resemblance values. Goodness-of-fit criteria exist for both methods to evaluate how well theyrepresent the originalresemblance matrix (Rohlf, 1993).
Much like the sampling methods, no two investigators studying the Lower Bay Complex used the same data analysis methods. In the present study, I attempted to examine the regional studies using a common set of multivariate methods. Abundance data were loge(x + 1) transformed and relationships were determined using the Bray-Curtis index. Data sets for inverse analysis were reduced to eliminate rare species (<5% of a sample at all stations in entire data set) as suggested by Field et al. (1982). Numerical analysis consisted of both UPGMA (Unweighted Pair-Group Method using Arithmetic averages) clustering and ordination by nonmetric multidimensional scaling (MDS) (Legendre and Legendre, 1998). Environmental data were standardized and matrices of Euclidean distances were calculated.
In the present study, one additional technique called a Mantel test was used to examine relationships among data sets (Mantel, 1967). In the Mantel test, two resemblance matrices (e.g., sets of Bray-Curtis indices for January 1973 and January 1984 data sets) are tested for correspondence. The statistic calculated is a correlation coefficient that ranges in value from — 1 to +1. Values close to +1 or —1 indicate strong positive or negative relationships, respectively. A value near zero indicates no relationship. If the resemblance matrices were calculated for normal analysis, the Mantel test determines how well the spatial community distribution, i.e., the habitat structure, matches between the two data sets. For inverse analysis, the test measures the strength of faunal associations by examining whether species or functional groups in the two data sets tend to co-occur in the same way. The Mantel test compares the relationships among sites or species and not the raw data directly. It is, therefore, somewhat less sensitive to differences in sampling methods and taxonomy than a direct comparison.
Combining all regional studies, a total of 328 benthic species have been identified in the Lower Bay Complex. The dominant taxonomic groups are polychaetes (43 percent), mollusks (17 percent), and crustaceans (31 percent). Of the taxa identified, ninety-five species emerged as being representative of the diverse life histories present and/or good indicators of community structure and community change (Tables 18.2 and 18.3). Several taxa (Polydora cornuta and P. ligni; Ampelisca sp. and A. abdita) were assigned different names but were probably the same organism. Because of the different methods used in the surveys, it is best to restrict comparisons to surveys with similar season and sieve size. For that purpose, both the 1.0 and 1.5 mm sieve results for July 1986 are included from Cerrato et al. (1989). Comparisons matching season and sieve size are always possible between the 1986-87 data and earlier surveys. Comparing the 1994-95
data to earlier studies is more problematic. Many striking differences occur for small organisms (for example, oligochaetes, the polychaetes Mediom-astus spp. and Streblospio benedicti, the bivalve Gemma gemma, and the amphipod Ampelisca abdita), suggesting a large sieve size effect.
The composition of species in Tables 18.2 and 18.3 held few surprises and generally included species common to estuarine and coastal regions of the Northeast and Mid-Atlantic. The most abundant organism in the Lower Bay Complex was the amphipod Ampelisca abdita. This is a small (4-8 mm), tube building, surface deposit feeder that was often found at average densities exceeding 1,000 per m2. It often represented >50% of the fauna collected and was commonly present at 70 to 90 percent of the sampling stations. Its tube building activities substantially modify the physical characteristics of the bottom, increasing the deposition of fine-grained sediments by trapping and incorporating particles into densely-packed tube mats (Rhoads, 1974). A. abdita is sensitive to pollutants and is extensively used in sediment toxic-ity tests (Redmond et al., 1994). It is an extremely important food source for winter flounder (Pseu-dopleuronectes americanus), windowpane flounder (Scophthalmus aquosus), scup (Stenotomus chrysops), weakfish (Cynoscion regalis), and silver hake (Merluccius bilinearis) (Franz and Tancredi, 1992; Steimle etal., 2000).
The polychaete Streblospio benedicti was another dominant. S. benedicti was widespread throughout the region and was found to occur on average at 37 percent of stations sampled during the regional studies. In June 1995, it was present at 97 percent of the sampling stations. Like A. abdita, it is a small (20 x 1 mm), tube building, surface deposit feeder (McCall, 1977). It is found in a variety of sediments and is highly opportunistic, i.e., an early colonizer on disturbed habitats with the ability to grow, mature, and reproduce quickly (McCall, 1977). S. benedicti is also tolerant of high levels of organic enrichment, organic contaminants, and low concentrations of dissolved oxygen (Llanso, 1991; Chandler, Shipp, and Donelan, 1997). Other widespread or cosmopolitan species, along with the average percent of stations at which they were found, included Mulinia lateralis (53 percent), Ilyanassa trivittata (50 percent), Glycera
Was this article helpful?