Printer Friendly



Many sessile organisms that inhabit the intertidal zone, such as the blue mussel Mytilus edulis, exhibit a distinct elevational zonation (e.g., Suchanek 1978), with their upper vertical limit often set by physical constraints and their lower vertical limit set by biological interactions (Raffaelli & Hawkins 1996). Environmental variables that influence species distribution in estuaries vary with elevation, and even small changes in vertical position can have profound effects on the prevailing physical and chemical conditions. Temperature, for example, often varies with elevation, with organisms occupying higher elevations being exposed to an increased risk of desiccation (Raffaelli & Hawkins 1996). Biological variables that influence species distributions also vary with elevation, such as predation intensity, which can be greater at lower elevations (e.g., Lin 1989). Elevational gradients are particularly relevant for organisms that inhabit southeastern salt marshes, where a range in elevation of 26 cm can encompass 80% of the marsh (Morris et al. 2005).

The ribbed mussel Geukensia demissa is a mytilid bivalve that inhabits salt marshes along the eastern coast of North America, from the Gulf of St. Lawrence to Florida, forming dense aggregations in the marsh substrate (Bertness 1984, Bertness & Grosholz 1985). At the landscape scale, G. demissa increases nitrogen removal through filter feeding (Jordan & Valiela 1982, Bilkovic et al. 2017), increases smooth cordgrass (Spartina alterniflora) productivity (Bertness 1984, Angelini et al. 2015), S. alterniflora drought resistance (Angelini et al. 2016), and invertebrate abundance and diversity (Altieri et al. 2007, Angelini et al. 2015). Through these functions, G. demissa is recognized as an ecosystem engineer (sensu Jones et al. 1997) and constitutes an integral part of the salt marsh community.

Previous studies attempting to characterize the distribution of Geukensia demissa have mostly focused on its relationship with biological variables, with several studies concluding that G. demissa distribution differs depending on salt marsh vegetation characteristics and location within the salt marsh. For example, Kuenzler (1961) noted that at a site on Sapelo Island, GA, G. demissa was most abundant near the mouths of creeks that supported tall-form Spartina alterniflora (size morphologies of S. alterniflora are described by Valiela et al. 1978). Kuenzler (1961) also observed lower abundances of G. demissa at higher elevations associated with short-form S. alterniflora. Other studies (e.g., Bertness & Grosholz 1985, Stiven & Gardner 1992, Franz 2001) have focused on the importance of salt marsh vegetation to the distribution of G. demissa, noting that G. demissa tends to occur in association with tall-form S. alterniflora. These studies only focused on small-scale factors that influence the distribution of G. demissa (e.g., the importance of a microhabitat because of its biological characteristics), however, and would be difficult to apply to landscape-scale habitat characterization efforts. Furthermore, although the upper and lower salinity tolerances of G. demissa have been previously described (Wells 1961, Lent 1969), there have been no studies that explicitly examine the relationship between its distribution and salinity, an important environmental variable for sessile marine bivalves (Remane 1934, Deaton & Greenberg 1986).

The purpose of this study was to develop a methodology that characterizes the landscape-scale distribution of Geukensia demissa, specifically using elevation and salinity, with the goal of creating a tool that can be used to predict the distribution of G. demissa habitat in estuaries throughout South Carolina. In South Carolina, where commercial harvest of G. demissa has increased substantially over the past 5 y (South Carolina Department of Natural Resources, unpublished data), such a tool could also have valuable application in the spatial management of this estuarine species. This study also established a methodology for determining the inundation period for G. demissa across a range of elevations, providing temporal insights into the biology of G. demissa such as opportunities for filter feeding and exposure to tidally migrating predators.


Study Area and Experimental Design

To establish the relationships among Geukensia demissa occurrence, elevation, and salinity, transect grid surveys were performed within the salt marsh habitat at eight stations along a salinity gradient in the Ashley River in Charleston, SC (Fig. 1). The Ashley River is a tidally influenced river that flows into the Charleston Harbor, with an average tidal range of approximately 1.6 m. Salt marshes dominated by Spartina alterniflora and black needlerush (Juncus roemerianus) fringe the river. Salinity data at 10 stations along the Ashley River were collected during a monitoring effort from May to September 2015 (McClellan 2017). It is important to note that these values represent average summer salinities; average winter salinities may be lower than those reported in this study. For example, at the nearby Charleston Harbor tidal gauge, the average monthly salinity fell by more than 20% between August 2015 and February 2016 (USGS 2016). For the purposes of the present study, average summer surface water salinities (n = 40 observations per station over the 5-mo period) were calculated from McClellan (2017), and the salinities were interpolated using a logistic model (Fig. 2) so that the sites used for the transect surveys could be assigned a salinity value. The salinity data were recorded in the main stem of the Ashley River, rather than the areas of marsh that were surveyed for G. demissa presence. Salinity was plotted as a function of distance along the Ashley River, which was measured using Google Earth Pro software (Version

Eight areas of marsh were surveyed along the salinity gradient, from 8.35 to 23. For each station at which Geukensia demissa presence-absence surveys were performed, a 30-point grid consisting of five adjacent transects was superimposed over the marsh using ESRI ArcMap GIS software

version 10.3.1 (Fig. 1, inset). The grid was approximately 100 m x 120 m, with approximately 20 m between each point. Each transect began at the edge of the marsh and ran perpendicular to the shoreline into the marsh interior, ensuring that a range of elevations was included in each transect. In the field, each point of the 30-point grid was scored for the presence or absence of G. demissa within a 1-m radius, and the horizontal position and elevation were recorded using a Trimble R8--integrated global navigation satellite system (hereafter the R8).

Measurement of Elevation

All elevation data were collected with the R8 using a virtual reference station survey method and corrected in real time using solutions provided by the South Carolina Real-Time Network (Allahyari et al. 2018). The elevation values were recorded as North American Vertical Datum of 1988 (NAVD88) orthometric heights using the GEOID12B geoid model to transform them from the North American Datum of 1983. To prevent the R8 antenna from sinking into the mud, the antenna base was placed in a plastic bucket; the thickness of the base material of this bucket was later subtracted from the elevation data.

Before field data collection, the expected error of the R8 was empirically estimated at two National Geodetic Survey benchmark sites located on James Island, SC using the same methods that were subsequently used at the sampling sites. For these trials, 150 data points were collected over 5 days. The horizontal and vertical root mean square errors were estimated, where the error of an individual point measurement was the absolute distance from the position measured by the R8 to the position of the benchmark surveyed by the National Geodetic Survey. The R8 had a horizontal and vertical root mean square error of 0.02 m (n = 150). Such a small vertical error highlights the centimeter-level vertical accuracy of the R8. Measurements collected by the R8 were compared with LiDAR-extracted values (n = 240) to quantify the difference between the two techniques. LiDAR-extracted data at each sampling point almost always (93% of the time) yielded higher elevation values than the R8 measurements, averaging 0.19 m ([+ or -]0.17 m SD) above the field-collected data.

The raw elevation data were transformed to local tidal datums. The NAVD88 vertical datum is a geodetic mean sea level datum and the standard vertical datum used for surveying and engineering in the continental United States. Tidal datums relate to local tidal water levels and are more ecologically relevant for use in analyzing the distribution of organisms that respond to tidal cycles. Therefore, elevations collected in the field were transformed to local mean high water (MHW) and mean low water (MLW) datums using the National Oceanic and Atmospheric Administration's (NOAA) VDatum software tool (v. 3.6.1). It should be noted that there was an error associated with the VDatum transformation from NAVD88 to MHW. In the South Carolina/Georgia region, the maximum cumulative uncertainty, representing the SD, was approximately 0.13 m (NOAA 2009). In addition, to account for potential differences in the magnitude of the tidal frame (i.e., distance between MLW and MHW) across sampling locations within the Ashley River, the elevations were converted to a relative tidal position, which was calculated for each observation using the following formula:

[[meters from MLW]/[(meters from MLW - meters from MHW)]]

Statistics and Model Selection

Logistic regression models were used to describe the relationship between Geukensia demissa occurrence (i.e., presence-absence), elevation, and salinity. A forward model selection process was used in which every model was considered between the following:

Pr(occurrence) ~ elevation + [epsilon]

Pr(occurrence) ~ elevation + [elevation.sup.2] + salinity + [salinity.sup.2] + elevation x salinity + [epsilon]

where [epsilon] represents a binomial error term. The models were compared using likelihood ratio tests (LRT) and the Akaike information criteria (AIC). Receiver-operator characteristic (ROC) curves were calculated for each model using the pROC package in R (Robin et al. 2011) to assess their goodness-of-fit, with the area under the ROC curve representing the ability of the model to discriminate true positives and true negatives. McFadden pseudo [R.sup.2] values were also calculated to assess the goodness-of-fit of each model using the pscl package in R (Jackman 2017). The model selection process was performed with elevation data relative to MHW and with the relative tidal position.

Application of the Model to the Ashley River

LiDAR raster data for the surveyed portion of the Ashley River were transformed so that they were relative to MHW using VDatum. A second raster surface was created for the Ashley River using inverse distance weighting to interpolate the salinity dataset provided by McClellan (2017) using ArcMap GIS software version 10.3.1. Pixel values for this surface were compared with the salinity logistic model using the Profile Graph tool within the 3D Analyst extension to ensure consistency between datasets. The pixel values were subsequently converted into a heat map depicting the probability of Geukensia demissa occurrence using the Raster Calculator tool to apply the best model (i.e., Eq. 1, described in the following paragraphs).

Inundation Period

Pendant data loggers (Onset HOBO; n = 12) set to record temperature data at 3-min intervals were placed at three of the surveyed sites from June 13 to July 14, 2017. Locations of the three sites on the Ashley River (upstream, midway along the salinity gradient, and downstream) were selected to capture inundation patterns across the salinity gradient investigated in this study. The data loggers were attached to PVC pipes placed in the substrate so that the loggers were held at the same elevation as the marsh surface. The elevation of each logger was recorded using the R8. Inundation period was inferred from changes in the temperature profile, when water and air temperature could be distinguished. The water temperature during the night was significantly higher than the air temperature, and so nighttime high tides were used. The times during which the data loggers were inundated were compared with high tides recorded by the Charleston Harbor tidal gauge.

A mixed-effects model was used to describe the relationships between inundation period, site elevation, and tidal height in which site elevation and tidal height were considered fixed effects, and the individual logger and site were considered random effects. The mixed-effects model was carried out using the lme 4 package in R (Bates et al. 2015). The Charleston Harbor tidal gauge, located within 20 km of all surveyed sites, provided observed water levels for the entirety of the sampling period. Although both site elevation and tidal height are expressed relative to a tidal datum, it is important to distinguish these two terms: site elevation refers to the elevation of the sampling point relative to MHW, whereas tidal height (relative to mean lower low water) refers to the maximum elevation of the surface of the water for a given high tide. The coefficients for the fixed effects were extracted and used to predict the inundation period for Geukensia demissa using elevation data collected during the surveys. All analyses were conducted using R version 3.3.2.


Model Selection

Relative tidal position did not appear to change with distance along the Ashley River; it was highly correlated with elevation (Pearson's correlation coefficient = 0.96) regardless of the sampling location and was, therefore, not considered in subsequent analyses. All the models described in the following paragraphs are relative to the local MHW datum (in meter).

Models that used the square of MHW ([MHW.sup.2]) were better at explaining variation in Geukensia demissa presence--absence than those that only used MHW (LRT, [[chi square].sub.(1)] = 54.8, P < 0.01), which was also reflected in their lower AIC scores. The same was true for models that used [salinity.sup.2]; these models were better at explaining variation in G. demissa presence--absence than those that only used salinity (LRT, [[chi square].sub.(1)] = 23.6, P < 0.01). The best model, as determined by LRT, included both the [MHW.sup.2] and [salinity.sup.2] terms, but not the interaction between MHW and salinity (LRT, [[chi square].sub.(3)] = 1.6, P = 0.67). The most likely model also had the lowest AIC (216.9; i.e., it was the most parsimonious model), the second highest McFadden pseudo [R.sup.2] value (0.37), and the highest ROC area under the curve (0.87; i.e., it had the lowest prediction error) (Table 1). It is described by the following equation, where Pr represents the probability of G. demissa occurrence:

In ([Pr/[1 - Pr]]) = - 16.70 - 17.57 x MHW + 1.85 x Salinity - 52.93 x [MHW.sup.2] - 0.05 X [Salinity.sup.2] (1)

Presence--Absence Surveys for Geukensia demissa

Both elevation and salinity had a significant effect on the probability of Geukensia demissa occurrence (logistic regression, df = 235, P < 0.01), which exhibited a distinct maximum probability at 0.17 m below MHW and at a salinity of 18 (Fig. 3), according to the model that best explained the variability in G. demissa presence--absence (Eq. 1). The probability of occurrence exhibited a quadratic curve in relation to both elevation and salinity, that is, the occurrence of G. demissa was less likely at elevations and salinities that were both higher and lower than these most probable values.

Based on the surveys conducted in the Ashley River, Geukensia demissa occurred from as high as 0.06 m above to as low as 0.41 m below MHW. No G. demissa were present in the most freshwater site (salinity = 8.35), even though the elevation and dominant vegetation (Spartina alterniflora and Juncus roemerianus) were similar to those at the other sites.

The predictions of Geukensia demissa habitat using the best model (Eq. 1) along the Ashley River are shown in Figure 4.

Inundation Period

The range of elevations at which the loggers were placed (i.e., site elevation) corresponded to the elevational range of Geukensia demissa in the Ashley River, from approximately 0.4 m below to 0.1 m above MHW. Because the loggers were deployed during the summer months, only nighttime high tides could be used; during the daytime, water and air temperatures in the marsh were both high and were difficult to distinguish. Loggers were considered to be inundated when the temperature suddenly increased and were considered to be inundated for as long as the temperature remained stable.

Each temperature logger yielded a different number of high tides that could be analyzed, ranging from 7 to 24 high tides per logger (average = 16). The relationship between inundation period, site elevation, and tidal height is presented in Figure 5. Depending on site elevation and tidal height, inundation periods ranged from 45 min to more than 6 h, with an average of approximately 3.5 h per high tide. Using the coefficients from the mixed-effects model and data from the transect grid surveys, the amount of time that Geukensia demissa spends immersed on an average high tide (1.65 m from mean lower low water, according to the NOAA tidal gauge in Charleston Harbor) can range from 1.27 to 3.71 h. At 0.17 m below MHW, the elevation where G. demissa is most likely to occur according to the logistic regression model, G. demissa would be inundated for 2.46 h on an average high tide. It is important to note that because G. demissa exhibits a highly aggregated distribution, there is a small vertical elevation component to the patch itself, with individuals located on and above the marsh surface. Therefore, the inundation period predictions in the present study can be considered the maximum inundation period for G. demissa at a given elevation because these predictions assume that all individuals are located on the marsh surface.


The surveys conducted in this study enabled the development of a model capable of predicting the probability of the occurrence of Geukensia demissa based on elevation and salinity data, with the potential for the model to be applied to other estuaries as a mapping tool. The model predicted a "sweet spot" where G. demissa was most likely to occur at a high intertidal elevation (0.17 m below MHW) and intermediate salinity (i.e., 18). The elevational range of G. demissa (within 0.41 m of MHW) presented here is similar to that reported by Lent (1968) who observed G. demissa at a site in Delaware located within 1 ft (0.31 m) of MHW.

In the present study, the model captured both the upper and lower elevational distribution limits for Geukensia demissa within the Ashley River, with occurrence at the extremes of this range less likely. Regarding salinity, the lower limit was well established, with G. demissa completely absent from the least saline sampling location, where salinity measured 8.35, suggesting that G. demissa populations are not able to persist in low salinity environments. Although G. demissa is euryhaline, capable of tolerating salinities as low as 5 (Wells 1961) and as high as 75 (Lent 1969), the Ashley River often experiences prolonged decreases in salinity during the wet season, potentially limiting G. demissa survival in less-saline areas. The minimum salinity at which G. demissa was observed in the present study was also consistent with an established decline in the occurrence of marine bivalve species at salinities below 5-8 (Remane 1934, Deaton & Greenberg 1986). Considering the known tolerance of G. demissa for high salinities, the upper salinity limit may not have been captured by the model, as the most saline site in the present study was estimated to have a mean summer salinity of 23 (McClellan 2017); the model should, therefore, not be used to predict G. demissa occurrence at salinities higher than 23.

The potential application of the model on a large scale is novel, as previous studies have primarily focused on local patterns of distribution because of biological variables (e.g., Kuenzler 1961, Bertness & Grosholz 1985, Franz 2001). The model presented in this study could easily be applied to other estuaries in South Carolina and along the eastern coast of North America, using elevation and salinity data to generate high-resolution and landscape-scale information on the distribution of Geukensia demissa. In the context of the commercial harvest of G. demissa in South Carolina, a mapping tool could be used to inform management decisions on setting harvesting boundaries, although first it must be ground-truthed and tested for accuracy.

Despite the mapping value of the model, there are sources of error that must be addressed. The error in transforming elevation data so that it is relative to MHW using VDatum, which NOAA (2009) estimates to be at most 0.13 m, presents a challenge. Furthermore, high-resolution temporally and spatially explicit salinity data are resource intensive to collect, potentially limiting the use of the model in different regions of South Carolina. The salinity dataset used in this study incorporated robust spatial and temporal variability, but such coverage is not common. To circumvent this issue in areas lacking comprehensive salinity data, a different model using only elevation data could be used to characterize Geukensia demissa habitat.

The accuracy of the R8 highlights its value in determining elevation in salt marshes, as opposed to relying solely on Li-DAR elevation data. LiDAR-extracted elevation values for each sampling point were consistently higher than those collected by the R8, a discrepancy that has been documented in other studies performed in South Carolina estuaries (e.g., Montane & Torres 2006, Chassereau et al. 2011). The tendency of LiDAR data to overestimate elevation in salt marshes is attributable to dense vegetation. Chassereau et al. (2011) found that elevation data were particularly inaccurate at lower elevations, where dense stands of tall-form Spartina alterniflora obstruct the path of the LiDAR sensing equipment. Landscape-scale mapping of Geukensia demissa habitat that relies on LiDAR data (e.g., Fig. 4) may require adjustment to account for this vertical bias.

The results from the inundation period experiment largely confirm previous observations (e.g., Kuenzler 1961, Lent 1968) that Geukensia demissa spends a greater amount of time aerially exposed than immersed. The present study goes a step further, providing a method for estimating the inundation period using site elevation and tidal data, and reveals that the high intertidal position of G. demissa allows for a relatively short opportunity for filter feeding. On average, the present study predicts that the inundation period is limited to approximately 2.5 h per high tide at elevations where G. demissa is most likely to occur. It should be noted, however, that the flow of water over the marsh is influenced not only by elevation and tidal height, but also by subtle three-dimensional features of the marsh surface, which can in turn influence inundation period (Torres & Styles 2007); this may have contributed to some of the variability observed in the present study. Determining the amount of time G. demissa spends immersed could be useful in creating models of physiological energetics (e.g., Widdows & Shick 1985).

The distribution of Geukensia demissa poses a regional-scale ecological question, as it appears to vary in relation to latitude. At northern latitudes, such as in the Northeastern United States, G. demissa is most abundant at the edge of the marsh, forming a dense, continuous band (Bertness 1984, Bertness & Grosholz 1985). In the high marsh platform, which is characterized by short-form Spartina alterniflora (Valiela et al. 1978), G. demissa abundance is low (Bertness & Grosholz 1985). At southern latitudes, including the Southeastern

United States region, G. demissa is not found at the edge of the salt marsh in a continuous band; instead, G. demissa forms discrete, dense patches in the high marsh platform and near the mouths of tidal creeks (Kuenzler 1961, Stiven & Gardner 1992). The driving mechanisms responsible for these two distinct distribution patterns are unclear, although the difference has been attributed to variation in predation intensity (Nielsen & Franz 1995). The use of elevation-based tools may improve the understanding of large-scale differences in the habitat distribution of G. demissa.


We wish to thank Ben Stone, Austin Sturkie, Stevie Czwartacki, and Kelsey McClellan for their help with field-work. Kelsey McClellan also provided the salinity dataset used in the construction of the model used to characterize ribbed mussel habitat. Funding for this study was provided by a grant awarded to the South Carolina Department of Natural Resources (SCDNR) through the USFWS State Wildlife Grants Program, as well as from funding allocated to the SCDNR from the South Carolina Saltwater Recreational Fishing License Fund and USFWS Wildlife and Sports Fish Restoration (WSFR) Program. This publication represents the South Carolina Department of Natural Resources' Marine Resources Research Institute contribution number 802.


Allahyari, M., M. J. Olsen, D. T. Gillins & M. L. Dennis. 2018. Tale of two RTNs: rigorous evaluation of real-time network GNSS observations. J. Surv. Eng. 144:05018001.

Altieri, A. H., B. R. Silliman & M. D. Bertness. 2007. Hierarchical organization via a facilitation cascade in intertidal cordgrass bed communities. Am. Nat. 169:195-206.

Angelini, C, J. N. Griffin, J. Van de Koppel, L. P. M. Lamers, A. J. P. Smolders, M. Derksen-Hooijberg, T. van der Heide & B. R. Silliman. 2016. A keystone mutualism underpins resilience of a coastal ecosystem to drought. Nat. Commun. 7:12473.

Angelini, C, T. Heide. J. N. Griffin, J. P. Morton, M. Derksen-Hooijberg, L. P. M. Lamers, A. J. P. Smolders & B. R. Silliman. 2015. Foundation species' overlap enhances biodiversity and multifunctionality from the patch to landscape scale in southeastern United States salt marshes. Proc. R. Soc. 282:20150421.

Bates, D., M. Maechler, B. Bolker & S. Walker. 2015. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67:-48.

Bertness, M. D. 1984. Ribbed mussels and Spartina alterniflora production in a New England salt marsh. Ecology 65:1794-1807.

Bertness, M. D. & E. Grosholz. 1985. Population dynamics of the ribbed mussel, Geukensia demissa: the costs and benefits of an aggregated distribution. Oecologia 67:192-204.

Bilkovic, D. M., M. M. Mitchell, R. E. Isdell, M. Schliep & A. R. Smyth. 2017. Mutualism between ribbed mussels and cordgrass enhances salt marsh nitrogen removal. Eeosphere 8:e0l795.

Chassereau, J. E., J. M. Bell & R. Torres. 2011. A comparison of GPS and lidar salt marsh DEMs. Earth Surf. Proc. Land. 36:1770-1775.

Deaton, L. & M. J. Greenberg. 1986. There is no horohalinicum. Estuaries 9:20-30.

Franz, D. R. 2001. Recruitment, survivorship, and age structure of a New York ribbed mussel population (Geukensia demissa) in relation to shore level: a nine year study. Estuaries 24:319-327.

Jackman, S. 2017. Classes and methods for R developed in the Political Science Computational Laboratory. R package version 1.5.2. Sydney, Australia: United States Studies Centre, University of Sydney. Available at:

Jones, C. G, J. H. Lawton & M. Shachak. 1997. Positive and negative effects of organisms as physical ecosystem engineers. Ecology 78:1946-1957.

Jordan, T. & I. Valiela. 1982. A nitrogen budget of the ribbed mussel, Geukensia demissa, and its significance in nitrogen flow in a New England salt marsh. Limnol. Oceanogr. 27:75-90.

Kuenzler, E. J. 1961. Structure and energy flow of a mussel population in a Georgia salt marsh. Limnol. Oceanogr. 6:191-204.

Lent, C. M. 1968. Air-gaping by the ribbed mussel, Modiolus demissus (Dillwyn): effects and adaptive significance. Biol. Bull. 134:60-73.

Lent, C. M. 1969. Adaptations of the ribbed mussel, Modiolus demissus (Dillwyn), to the intertidal habitat. Am. Zool. 9:283-292.

Lin, J. L. 1989. Influence of location in a salt marsh on survivorship of ribbed mussels. Mar. Ecol. Prog. Ser. 56:105-110.

McClellan, K. L. 2017. Assessing the effect of climate variability on the South Carolina recreational and commercial blue crab fishery. MS thesis. College of Charleston's Master's in Environmental Studies Program. 138 pp.

Montane, J. M. & R. Torres. 2006. Accuracy assessment of Lidar saltmarsh topographic data using RTK GPS. Photogramm. Eng. Remote Sens. 72:961-967.

Morris, J. T., D. Porter, M. Neet, P. A. Noble, L. Schmidt, L. A. Lapine & J. R. Jensen. 2005. Integrating LIDAR elevation data, multi-spectral imagery and neural network modelling for marsh characterization. Int. J. Remote Sens. 26:5221-5234.

National Oceanic and Atmospheric Association (NOAA). 2009. Estimation of vertical uncertainties in VDatum. Vertical Datum Transformation. Accessed May 15, 2018. Available at:

Nielsen, K. J. & D. R. Franz. 1995. The influence of adult conspecifics and shore level on recruitment of the ribbed mussel Geukensia demissa (Dillwyn). J. Exp. Mar. Biol. Ecol. 188:89-98.

Raffaelli, D. & S. Hawkins. 1996. Intertidal ecology. London, United Kingdom: Chapman & Hall. 356 pp.

Remane, A. 1934. Die brackwasserfauna. Verh. Deut. Zool. 36: 34-74.

Robin, X., N. Turck, A. Hainard, N. Tiberti, F. Lisacek, J. Sanchez & M. Muller. 2011. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 12:77.

Stiven, A. E. & S. A. Gardner. 1992. Population processes in the ribbed mussel Geukensia demissa (Dillwyn) in a North Carolina salt marsh tidal gradient: spatial pattern, predation, growth and mortality. J. Exp. Mar. Biol. Ecol. 160:81-102.

Suchanek, T. H. 1978. The ecology of Mytilus edulis L. in exposed rocky intertidal communities. J. Exp. Mar. Biol. Ecol. 31:105-120.

Torres, R. & R. Styles. 2007. Effects of topographic structure on salt marsh currents. J. Geophys. Res. 112:.

U.S. Geological Survey (USGS). 2016. National water information system. USGS Water Data for the Nation. Accessed August 16, 2018. Available at:

Valiela, I., J. M. Teal & W. G. Deuser. 1978. The nature of growth forms in the salt marsh grass, Spartina alterniflora. Am. Nat. 112:461-470.

Wells, H. W. 1961. The fauna of oyster beds, with special reference to the salinity factor. Ecol. Monogr. 31:239-266.

Widdows, J. & J. M. Shick. 1985. Physiological responses of Mytilus edulis and Cardium edule to aerial exposure. Mar. Biol. 85:217-232.


(1) South Carolina Department of Natural Resources, Marine Resources Research Institute, 217 Fort Johnson Road, Charleston, SC 29412; (2) Department of Biology, College of Charleston, 66 George Street, Charleston, SC 29424

(*) Corresponding author. E-mail:

DOI: 10.2983/035.038.0105
Partial list of models considered to characterize Geukensia demissa
habitat, with AIC and ROC scores.

Factors included in model             AIC     ROC area under curve

MHW                                   331.0   0.50
Salinity                              292.7   0.75
MHW and salinity                      285.8   0.75
MHW and [MHW.sup.2]                   278.3   0.74
Salinity and [salinity.sup.2]         271.1   0.77
MHW, [MHW.sup.2], and salinity        246.5   0.83
MHW. salinity, and [salinity.sup.2]   263.3   0.80
MHW, [MHW.sup.2],                     216.9   0.87
salinity, and [salinity.sup.2]
MHW, [MHW.sup.2], salinity,           221.3   0.87
[salinity.sup.2], and
MHW x salinity

These models use elevation relative to MHW. The best model is shown in
COPYRIGHT 2019 National Shellfisheries Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Julien, Asa R.; Tweel, Andrew W.; Mcglinn, Daniee J.; Sundin, Gary W.; Hadley, Nancy H.; Ley-Smith,
Publication:Journal of Shellfish Research
Article Type:Report
Geographic Code:1USA
Date:Apr 1, 2019

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters