INTEGRATION OF BIOTIC AND ABIOTIC DATA TO MAP BENTHIC HABITATS WITHIN BLOCK ISLAND AND RHODE ISLAND SOUNDS

The mapping of benthic habitats presents the distribution and extent of seafloor environments, including biotic and abiotic characteristics, in a geo-spatial context. This thesis aims to improve methodologies used in the field of benthic habitat mapping and works towards establishing a standard mapping protocol to facilitate more effective communication both among scientists and resource managers in effort to further the goal of science-based decision making. This study is in response to interest in wind turbine construction within Rhode Island waters. A thorough understanding of benthic habitats is essential for making scientifically valid management decisions to minimize ecological and economical development impacts. Two major challenges facing benthic habitat mapping are: 1.) Appropriate methodology; and 2.) Producing maps that can easily and effectively convey information important to a broad range of users (e.g. scientists, management agencies, non-profit organizations, individual citizens). The first challenge is examined in Chapter 1, which investigates the effectiveness of two mapping approaches, top-down and bottom-up, for classifying and mapping offshore marine environments. Both methods incorporate acoustic data (side-scan sonar and bathymetry), along with sediment and benthic macrofauna samples. The traditional top-down mapping approach identifies biological community patterns based on geologically-defined habitat map units , whereas the bottom-up approach aims to establish units based on biological similarity and then use statistics to determine relationships with associated environmental parameters. Both methods showed statistically strong and significant abiotic-biotic relationships and produced habitat units with distinct macrofaunal assemblages. Overall, the bottom-up approach was more effective at mapping benthic habitats, producing more clearly defined macrofaunal assemblages . However, the spatial heterogeneity prevented development of full-coverage maps with the currently available number of ground-truth samples . Therefore, for the mapping needs of RI, the top-down method is recommended because it can produce full-coverage maps. Chapter 2 addresses the second challenge . Commonly, maps characterize habitats according dominant species or general community type . While useful, such maps do not always offer practical information to managers and can inadequately represent important habitat characteristics and relationships. In response , benthic habitats were classified according to biological and environmental metrics considered important to the existence of healthy, productive benthic habitats. The weighted metrics were totaled to develop an overall index of benthic habitat value. The index also provides individual metric scores, allowing habitats to be evaluated based on metrics relevant to the user. Furthermore, indices can be used to discern biotic-abiotic relationships between and among habitats and index metrics. The indices identified habitats that scored considerably higher than the others . In general , though, the indices did not indicate specific biological or environmental characteristics that lend to high habitat value, signifying management efforts need to consider all habitat types. However, a correlation was found between tube-building species and species richness, indicating tube mat structures lead to increased biodiversity. The indices also show that habitats within each study area have different relationships with the index metrics, indicating macrofauna have their own associations with the environment within each study area. ACKNOWLEDGMENTS First and foremost, I would like to thank John King for his support and guidance and for providing me with the wonderful opportunity to be a graduate student and allow me to follow my research interests. I would also like to acknowledge my other committee members, Peter August and Susanne Menden-Deuer, for their generous time and valuable advice and encouragement. In addition, I would like to recognize the Graduate School of Oceanography at the University of Rhode Island for fostering an environment that is supportive and facilitates learning. Funding for these studies were provided by the RI Ocean SAMP project. Special thanks to Rhonda Kenny for taking care of the paperwork to ensure my funding. I greatly appreciate all of the support received from my fellow co-workers: Fred Hegg for sharing his extensive sonar expertise; Bryan Oakley for developing the geologic depositional environments data layers; Sheldon Pratt for macrofauna identification and enumeration; Chris Damon for development of the surface roughness data layer; Emily Shumchenia for engaging in thoughtful conversations early on in my student career and for assistance with statistical analyses; Danielle Cares for grain size analysis and, more importantly, for being a great friend and coffee break buddy; Carol Gibson for helping me solve random problems; Anna Malek for fish stomach contents data; and Chip Heil, Dane Sheldon, Nate Vinhateiro, Steve Smith, and Anthony Rossato, for assistance with fieldwork and/or data processing . Also, this research would not have been possible without the help of Gary Savoie, Cathy Cipolla, or Sam DeBow and the crew of the RIV Endeavor.


DEAN OF THE GRADUATE SCHOOL
UNIVERSITY OF RHODE ISLAND individual metric scores, allowing habitats to be evaluated based on metrics relevant to the user. Furthermore, indices can be used to discern biotic-abiotic relationships between and among habitats and index metrics. The indices identified habitats that scored considerably higher than the others . In general , though, the indices did not indicate specific biological or environmental characteristics that lend to high habitat value, signifying management efforts need to consider all habitat types. However, a correlation was found between tube-building species and species richness, indicating tube mat structures lead to increased biodiversity. The indices also show that habitats within each study area have different relationships with the index metrics, indicating macrofauna have their own associations with the environment within each study area.

ACKNOWLEDGMENTS
First and foremost, I would like to thank John King for his support and guidance and for providing me with the wonderful opportunity to be a graduate student and allow me to follow my research interests. I would also like to acknowledge my other committee members, Peter August and Susanne Menden-Deuer, for their generous time and valuable advice and encouragement. In addition, I would like to recognize the Graduate School of Oceanography at the University of Rhode Island for fostering an environment that is supportive and facilitates learning. Funding for these studies were provided by the RI Ocean SAMP project. Special thanks to Rhonda Kenny for taking care of the paperwork to ensure my funding.
I greatly appreciate all of the support received from my fellow co-workers: Fred Hegg for sharing his extensive sonar expertise; Bryan Oakley for developing the geologic depositional environments data layers; Sheldon Pratt for macrofauna identification and enumeration; Chris Damon for development of the surface roughness data layer; Emily Shumchenia for engaging in thoughtful conversations early on in my student career and for assistance with statistical analyses; Danielle Cares for grain size analysis and, more importantly, for being a great friend and coffee break buddy; Carol Gibson for helping me solve random problems; Anna Malek for fish stomach contents data; and Chip Heil, Dane Sheldon, Nate Vinhateiro, Steve Smith, and Anthony Rossato, for assistance with fieldwork and/or data processing .

Introduction
Benthic habitat is described as "a spatially defined area where the physical, chemical, and biological environment is distinctly different from the surrounding environment" . Therefore, distinct biological assemblages are thought to represent distinct environmental conditions . The mapping of benthic habitats presents the distribution and extent of biotic and abiotic characteristics of seafloor environments in a geospatial context ).
Typically, geologic and water depth parameters define the abiotic characteristics.
Benthic habitat maps are valuable tools for numerous ecological and management reasons, including understanding benthic habitat and faunal species and/or community distribution patterns and processes (Valesini et al., 2010;Zajac et al., 2000); defining essential fish habitat (Rooper and Zimmermann, 2007;; establishing environmental baselines ; and implementing appropriate management strategies, such as marine spatial planning, resource regulation, restoration, conservation, monitoring, and impact assessment (Last et al., 201 0;Diaz et al., 2004;Zajac et al., 1999;. There are typically two components to benthic habitat mapping: seafloor imaging and ground-truth studies (Rooper and Zimmermann, 2007). Seafloor imaging is often performed with side-scan sonar and swath bathymetry. These data sets can offer continuous coverage, high-resolution data of large areas (Kenny et al., 2003) and can be acquired relatively rapidly and affordably (Collier and Brown, 2005). Bathymetry maps indicate the depths and topography of the seafloor. Side-scan sonar backscatter intensity reflects the amount of sound returning to the sonar after hitting the seafloor and is indicative of the density, slope, and roughness of the seafloor (Goff et al., 2000). Backscatter intensity has also been linked to seafloor sediment characteristics (Brown and Collier, 2008;Collier and Brown, 2005) . Therefore, side-scan has traditionally been used to map the spatial complexity and heterogeneity of seafloor sedimentary and geological features . Acoustic data are less able to capture biological characteristics of the seafloor (Zajac, 1999). However, side-scan may delineate biological features when the biota modifies the physical structure of the seafloor and produces unique acoustic return patterns, such as with coral reefs ( e.g. Kendall et al., 2005;Collier and Humber, 2007;Roberts et al., 2005;Mumby et al., 2004), shellfish beds (e. g. van Overmeeren et al., 2009;Kostylev et al., 2003), and submerged aquatic vegetation (e.g. Lefebvre et al., 2009;Jones et al., 2007;Sabol et al., 2002).
Ground-truth studies refer to the acquisition of surficial seafloor grab samples, cores, trawl data, and/or underwater imagery (Brown and Collier, 2008;Kenny et al., 2003). These data offer point-or transect-coverage over small areas (Rooper and Zimmermann, 2007) and are usually collected at coarse spatial resolutions . Ground-truth studies are performed to obtain fine-scale information of seafloor characteristics (such as biota, sediment grain size, geological formations, wave/current processes) (Brown and Collier, 2008), often to assist with interpretation and classification of acoustic data. While ground-truth samples can offer detailed point data, the low sampling resolution usually prevents such data from being stand-alone mapping tools, as they may be unable to detect habitat and/or biological structure changes, particularly over small spatial scales and in heterogeneous areas . In addition, interpolating between point samples can produce inaccurate results .
Commonly, benthic habitat mapping employs a top-down approach (Shumchenia and King, 201 0;. This methodology develops habitat map units based on geological similarity, following the assumption that geologic environments or features, such as sediment type, contain distinct biological assemblages. The approach involves acoustically mapping an area of seafloor and then interpreting the data into distinct regions according to backscatter patterns and/or depth ( either visually or using automated classification software). The biological characteristics of each map unit type is identified from ground-truth data and integrated into the map unit description (Shumchenia and King, 201 0;Solan et al., 2003;Ellingsen et al., 2002;.
Using acoustic methods as the primary tools to delineate benthic habitats is attractive because it is less time-and cost-intensive, and requires minimal ground-truth data . However, since side-scan data primarily reflect physical characteristics of the seafloor, the top-down approach tends to produce geology-based habitats and inadequately represent biological communities (Valesini et al., 2010;Shumchenia and King, 201 0;. In addition, the validity and cohesiveness of the biological assemblages defined among these habitats is often not statistically examined (Last et al., 201 0;.
Often, studies employing the top-down approach find that benthic fauna tend to transcend acoustically-derived habitat boundaries -that is, biological communities are present in multiple habitats and a defined habitat exhibits a range of biological communities . However, this finding does not indicate organism-sediment relationships do not exist. Many studies have found links between sediment type and benthic fauna community structure (Verfaillie et al., 2009;Brown and Collier, 2008;Zajac et al., 2000;Snelgrove and Butman, 1994;Rhoads, 1974). The discrepancy may be because sediment grain size is not the sole determinant of species distribution (Snelgrove and Butman, 1994) and some acoustically defined habitats have similar sediment characteristics. In addition, it is likely that a combination of environmental parameters define the range limits of biological assemblages, such as water depth, nutrient and food supply, hypoxia/anoxia, current patterns, disturbance events, competition and predator-prey interactions. For example, in Long Island Sound, community structure changes occur with bathymetric and meso-scale circulation patterns (Zajac et al., 2000).  (Shumchenia and King, 201 0;. The spatial distribution of the habitat map units can be determined objectively through interpolation of the meaningful pointsource parameters  . This extrapolation allows the creation of full-coverage, benthic habitat maps McBreen et al., 2008;.
The bottom-up approach has many advantages. It has the potential to preserve species-environment relationships preserved (Shumchenia and King, 201 0;Rooper and Zimmermann, 2007), biological assemblages are more well-defined (Shumchenia and King, 201 0;, finer-scale habitat attributes can be discerned, and the multivariate analyses employed indicate how well biological assemblage variability is captured by abiotic parameters. This approach is especially useful in benthic environments characterized by gradual transition zones, low relief, and relatively homogenous sediment types, such as gravel and sand  and soft-sediment . In these environments where the ability of acoustic methods to distinguish benthic habitats is limited, the bottom-up method may be better able to detect habitats (Shumchenia and King, 201 0;.
The bottom-up method, however, requires a higher density of point-samples compared to the top-down method Zajac, 1999), causing it to be more resource-intensive (Shumchenia and King, 201 0;. Furthermore, ground-truth surveys must be extensive enough to sample all habitats within the study area; habitats not sampled will not be represented in the final benthic habitat classification map (Rooper and Zimmermann, 2007) .
The primary purpose of this study was to investigate the effectiveness of two mapping approaches, top-down and bottom-up, in offshore marine environments. This comparison is important for advancing methodologies and working towards a standard protocol within the field of benthic habitat mapping that can be applied regardless of study location. Furthermore, to my knowledge, the application of the bottom-up approach in offshore waters , where data density tends to be lower, has not been done before. Secondly, this study aims to classify benthic habitats to assist in determining appropriate locations for wind turbine installation.

Study Area
Rhode Island Sound (RIS) and Block Island Sound (BIS) are transitional waters that separate the estuaries of Narragansett Bay and Long Island Sound from the outer continental shelf . RIS and BIS are environmentally , economically, and culturally valuable for renewable energy development, fishing, boating, ferry and shipping routes, and tourism (RI CRMC, 2010). The benthic habitats of two areas identified as primary potential wind farm locations through a Tier 1 screening process (Spaulding et al., 2010) were examined in detail (Figure 1.1). The BI study area is a 138.6 sq km survey area located within state waters to the south of Block Island, and the FED study area is 178.7 sq km and is located in federal waters in eastern RIS.

a. Acquisition
Side-scan and swath bathymetric data were simultaneously collected within the study areas using an interferometric sonar (C3D, Teledyne Benthos) (Figure 1.2). The Hypack navigation software was use to plan surveys and log in real-time . The acoustic surveys were composed of parallel track lines, with line spacing between 100 m and 150 m. In order to obtain 100% coverage, line spacing was such that each swath overlapped at least 25% with its neighboring swaths and resulted in every portion of the seafloor being imaged at least once.

b. Processing
The raw files were processed using OIC CleanSweep software. For the side-scan backscatter, 2 m resolution mosaics were created. Bottom tracking, angle-varying gains (A VG) and look-up tables (LUT) were applied to the data as necessary to correct for water column returns, arrival angle, and to increase the signal-to-noise ratio of the backscatter returns. These corrections helped create a uniform image to effectively display the features of the seafloor. The backscatter intensity mosaic is displayed on a false color scale as an inverse grey-scale image, ranging from zero (black) to 255 (white). Stronger backscatter is depicted by lighter pixels and represents highly reflective (usually harder or rougher) surfaces, whereas weaker backscatter (darker pixels) represents acoustically absorbent (usually softer or smoother) bottoms (Wille, 2005  For the bathymetry, each swath was corrected for tide, vessel motion, and sonar mount angle. An angle filter was applied to remove potential outlier soundings.
Partial overlap of adjacent swaths allowed the data to be filtered to 6-8X the water depth, ensuring the highest quality soundings were used to build the mosaics. The

c. Analysis
Although both side-scan and bathymetry datasets were collected at very high resolution (2 m and 10 m pixels, respectively), creating habitat maps at this level of detail would be prohibitive ( computation time , file sizes). Therefore, 100 m pixel size was chosen, a scale at which major geophysical changes and boundaries across both study areas were still visible in the mosaics. The mean, minimum, maximum and standard deviation of both the side-scan and bathymetry were calculated at 100 m resolution. These parameters were calculated using ArcMap 9.3 with the Block Statistics feature in the Spatial Analyst Toolbox. Slope was derived using the slope function in Neighborhood Statistics in the Spatial Analyst extension.
In addition, a set of 1.9 million National Ocean Service (NOS) soundings was also compiled . These soundings were used to create a data layer that is a broad-scale measure of surface roughness throughout RIS and BIS. Using the Neighborhood Statistics function, this surface roughness layer was derived by calculating the standard deviation of the slope ( 100 m resolution) within a search radius of 10 pixels (i.e. 1,000 m) using a moving widow algorithm (Damon, Pers. Comm.). Therefore, the resulting data layer has a 100 m pixel resolution and each pixel has a value that is the standard deviation of the slope of the surrounding 1,000 m.

a. Sediment samples
A sub-sample was taken from the surface of each bottom sample and sediment properties characterized using a particle size analyzer (Malvern Mastersizer 2000E).
The Mastersizer generated the weight percent of each Wentworth particle size fraction (very fine sand, fine sand, medium sand, etc.), along with the standard deviation of the particle size distribution for the entire sample.

b. Macrofauna samples
The remaining material from each bottom sample was sieved on 1 mm mesh and macrofauna were retained. All individuals were counted and identified to at least the genus level. In addition, a functional group designation (e.g. surface burrower, tubebuilder, mobile) for each genus was made. The macrofauna abundances(# of individuals) from the BI and FED study areas were pooled and only the genera contributing to 97% of the total abundance between the two areas were included in further analyses. This eliminated genera with very low abundances(< 0.09% of the total abundance, equivalent to < 19 individuals) and resulted in the removal of 663 Examining these three genera at the species-level allows for investigation into if the individual species have distinct relationships with their respective environments.

a. Habitat map units
Geologic depositional environment types define the extent of the habitat map units for the top-down approach (Figure 1.6). The environments were visually interpreted for both the BI and FED study areas from high-resolution side-scan and bathymetry mosaics, sub-bottom seismic reflection profiles, surficial sediment samples, and underwater video  classification framework, (Madden et al., 2010).

b. Multivariate analyses
Analysis of similarity (ANOS IM) was performed on the Bray-Curtis similarity matrix to test the null hypothesis that there were no differences between macrofaunal assemblages among geologic depositional environment types. The test was permuted 999 times to generate a significance level (p < 0.05). The similarity percentages (SIMPER) routine was then used to compare the degree (percentage) to which each individual genus contributes to the within-environment similarity and amongenvironment dissimilarity . SIMPER also reports the percent average within-environment similarity and among-environment dissimilarity.
All analyses were executed in PRIMER 6.

c. Classification
Habitat map units were classified according to the average most abundant genus (# of individuals) within the bottom samples retrieved there, following CMECS protocol. To show biotic-abiotic associations, map units were also labeled by geologic depositional environment type.

a. Multivariate analyses
A suite of abiotic variables was generated from the multiple data layers (i.e. sidescan backscatter, bathymetry , sediment samples , NOS soundings) at each of the 78 bottom sample stations (Table 1.1 ). The variables were normalized to correct for differences in units, and a resemblance matrix created based on the Euclidean distance metric. All analyses were performed in PRIMER 6 (refer to Clarke et al. (2008) or  for further details of statistical analyses).
The biotic Bray-Curtis similarity matrix and the abiotic Euclidean distance resemblance matrix were subject to the BIOENV procedure. BIOENV identifies a subset of abiotic variables that best "explain" the patterns in the macrofaunal composition. BIOENV searches for high rank correlations between the Bray-Curtis and Euclidean matrices and .outputs the highest Spearman rank correlation, p, between combinations of abiotic variables and the macrofaunal assemblages. The maximum number of variables permitted in the output was capped at ten. The BIOENV routine was permuted 999 times to allow for the significance of the results to be assessed.
Statistical significance was assigned when p < 0.05.
The BIO EVN procedure was performed twice, once using all of the abiotic variables and once removing variables that were highly correlated , and therefore , redundant (r > 0.85), as assessed from a draftsman plot was created to assess correlations between the abiotic variables. The more sensible variable was chosen for analysis (for example, mean water depth was chosen over minimum water depth). The test was permuted 999 times to assess significance.
An ANOSIM was performed on the LINKTREE classes to test the null hypothesis that there were no significant differences in the macrofaunal assemblages among classes. SIMPER was used to determine both the overall and individual contributions of each genus to the within-group similarity and between-group dissimilarity of the resulting LINK TREE classes.

b. Habitat map units
To develop full coverage habitat map units, interpolation of the grain size point sample dataset is necessary. However, attempts to interpolate using traditional methods (Oridinary Kriging, Inverse Distance Weighting) in ArcMap 9.3 were unsuccessful due to semi-variograms that failed to show similarity (low semivariance) at short lag distances . This results from point samples being spaced too far apart resulting in a lack of spatial autocorrelation. Using continuous coverage data (water depth , side-scan backscatter, surface roughness) to predict sediment properties was also not successful. For example, the best linear model explaining variation in coarse grain size based on surface roughness and minimum depth had an r 2 of 0.59 .
This was considered was too weak to develop a predictive map of grain size using surrogate data and a least-squares regression model approach . Because full-coverage map units could not be confidently developed , the bottom-up maps were constructed by classifying pixels for which all abiotic data were available and at the original extent (i.e. 78, 100 m pixels). This conservative approach was taken to preserve the accuracy of the maps. This concern for retaining accuracy is echoed by Brown and Collier (2008) who remarked that interpolation methods can often lead to erroneous assumptions in the resulting map, particularly if the degree of seafloor heterogeneity reflected by surficial geology and biota is high (as it is in this study).

c. Classification
The habitat classes follow the LINKTREE output. Each class is described by the average most abundant genus(# of individuals) across all samples within the class (following CMECS protocol) and its relevant abiotic variables to indicate bioticabiotic relationships. and B. serrata (12.6%), both tube-building amphipods, followed by N annulata (8.3%), a deposit feeding bivalve.

a. Sediment samples
Of the 78 bottom samples, 30 genera/species were most abundant within one or more samples. The 48 samples within BI were dominated by 25 genera/species, whereas seven genera/species dominated the 30 samples within FED.

a. Multivariate analyses
There were strong and significant differences in macrofaunal assemblages among the geologic depositional environments (ANOSIM global R = 0.60, p = 0.001).
Judith-Buzzards Bay (PJ-BB) Moraine with sand sheets, sand sheets with gravel, and sand waves" exhibited the most similarity (59.4%), followed by "A. agasizzi -Glacial Lake Floor with sand sheets" (58.3%) and "N. annulata/A. agasizzi -Glacial Lake Floor with fine or coarse sands" (56.1 %). The contribution for the genera/species most responsible for the within-environment similarity ranged between 7. 9% and 100%. The genus/species most responsible for the similarity of each unit varied (18 genera/species identified). Some units were labeled by multiple genera because they contribute equally or nearly equally . The percent dissimilarity between map units ranged from 40.7% to 97.3%, having an average of 77.3%. B. serrata, A. vadorum, A. agasizzi, and N. annulata were the most responsible for the dissimilarity.

b. Classification
The top-down benthic habitat mapping approach generated 18 map units, none of which were present within both study areas. There were 12 map units within BI and six within FED (Figure 1 In total, 10 genera/species defined or co-defined the 18 habitat map units, with eight genera/species representing the units in BI and four representing the units in FED. Five of the 10 genera/species were tube-building amphipods, three were burrowing polychaetes, and there was one tube-building polychaete and one bivalve.

a. Multivariate analyses
The BIOENV procedure identified a subset of six abiotic variables as being the most correlated the macrofaunal composition (p = 0.70, p = 0.001). The variables responsible were percent medium sand, percent coarse sand, standard deviation of the grain size (µm), maximum backscatter intensity, mean depth (m), and surface roughness. Mean depth was the single variable having the highest correlation (p = 0.52) with the macrofaunal assemblage. These results persisted whether highly correlated variables were included or excluded in the analysis.
The LINK TREE identified 22 classes, each of which was defined by a series of abiotic quantitative thresholds of the six input variables (Figure 1. 9, Table 1.6). Each of the class breaks was significant(> 5%) and ANOSIM R values were between 0.36 and 0.81. Six of the thresholds were defined by percent medium sand, five by surface roughness, four by mean water depth, three by percent coarse sand, two by standard deviation of the grain size, and one by maximum backscatter intensity. Some of these thresholds were defined over a narrow range. For example, split "J" divided to the left at surface roughness less than 0.120 and to the right at greater than 0.124, and split "M" was defined by mean water depth less than 19.0 m to the left and greater than 19. 7 m to the right. hudsoni" (58.3%). The contribution for the genera/species most responsible for the within-class similarity ranged between 8.8% and 100%. The genus/species most responsible for the similarity of each class varied, with 19 genera/species identified.
The average between class dissimilarity was 78.9%, ranging from 44.5% to 98.8%.
The species most responsible for the dissimilarity were B. serrata, A. vadorum, N. annulata, and J falcata.

b. Classification
The bottom-up benthic habitat mapping approach resulted in the classification of 78, 100 m pixels ( Figure 1.10). The approach generated a total of 22 habitat classes, 18 of which were present in BI and 9 in FED. The two study areas had 5 classes in common. There were between 2 and 14 bottom samples within each class. In cases where the same genus/species was dominant, classes were distinguished with roman numerals, since the macrofaunal communities among the LINKTREE derived classes were significantly distinct. Classes were identified by two dominant genera/species when their abundances were nearly identical or very high compared to the remaining abundances within that class.
Tube-building amphipods dominated the BI and FED habitats, defining or co- vadorum dominated or co-dominated four classes and three classes, respectively, and A. agasizzi shared dominance for one class. The bivalve, N. annulata, defined or codefined three classes (21 pixels) and the burrowing polychaete, N. nigripes , defined one class (2 pixels).
Overall, 17 genera/species described or co-described the 22 habitat classes.
Specifically, 14 genera/species represented the BI classes and five represented the FED classes. Five of the 17 genera/species were burrowing polychaetes, four were tube-building amphipods, two were tube-building polychaetes, three bivalves, two amphipods, and one species of Oligochaeta.

Discussion
Maps of the distribution of benthic habitats are valuable tools for numerous ecological and management purposes, including understanding ecosystem patterns and processes, determining environmental baselines, impact assessments, and conservation efforts. The goal of this study was to construct and compare the effectiveness of benthic habitat maps for two areas, using the traditional top-down method and the alternative bottom-up method, which has not before been applied to offshore environments.

Comparison of benthic habitat mapping approaches
The top-down classification was advantageous because it produced full-coverage habitat map units containing significantly distinct macrofaunal communities (ANOSIM global R = 0.60, p = 0.001) and described broad-scale biological and geological resources. Furthermore, because the habitats were based on geological similarity, data collection, processing, and analysis were relatively less time-and effort-intensive.
While successful, the top-down approach also had disadvantages. As is frequently found in other top-down studies, some benthic communities and fauna transcend the habitat boundaries as defined by depositional environment type. This is a concern for the top-down approach because it defies the assumption that distinct geological environments will contain distinct biological communities. The A .

Comparison of study areas
The benthic habitats of the two study areas, BI and FED, differ in their biotic and abiotic characteristics, suggesting macrofaunal assemblages primarily have their own associations with the environment. This difference can be seen in the results of the where the thresholds used to define habitat classes occur over narrow ranges of the abiotic variables. The biological communities within the study areas likely vary over a similar spatial scale.
The BI study area exhibits a higher degree of benthic habitat heterogeneity than FED, as evidenced by the top-down and bottom-up approaches both producing twice as many habitats in BI. The side-scan and bathymetry mosaics, depositional environment maps, and grain size data also reflect the increased physical heterogeneity of BI compared to FED. With regard to biological characteristics, at least twice as many genera/species define the habitats in BI than in the larger FED site, and over 3x as many were found to be most abundant genera/species in one or more of the bottom samples.

Biotic-abiotic relationships
The scale at which the environmental parameters and acoustic patterns are examined is important in assessing abiotic-biotic relationships. This importance can be seen in the results of the bottom-up (via the BIOENV procedure) and top-down mapping approaches. For example, the results indicate macrofauna patterns within BI and FED are linked to geologic characteristics at both fine and broad spatial scales.
The point-sample grain size, specifically percent medium and coarse sand, represents the fine scale link. Such sediment-macrofauna associations have been commonly observed in bottom-up mapping approaches (Todd and Kostylev, 2011;, as well as other studies (Verfaillie et al., 2009;Zajac et al., 2000;Snelgrove and Butman, 1994;Chang et al., 1992;Rhoads, 1974). The relationship was also proposed for Block Island Sound by , who suggested the biological communities are gradational, probably related to small-to large-scale differences in sediment texture. The broad-scale geologic-biotic link is with depositional environment type, a relationship which other studies have had mixed results establishing (Todd and Kostylev, 2011;Solan et al., 2003;.
Maximum side-scan backscatter intensity may be another broad-scale geologic connection. Studies have shown positive correlations between backscatter intensity and grain size (Goff et al., 2000, Collier and Brown, 2005.
Therefore, the maximum backscatter intensity may reflect sediment characteristics.
The BIOENV analysis also revealed connections between macrofauna patterns and small and broad scale environmental heterogeneity, as reflected by the standard deviation of the sediment grain size and surface roughness datasets, respectively. That the macrofauna have such a close relationship to these two datasets is interesting because they are very different measures of environmental heterogeneity. The standard deviation of the sediment is a point sample data set that measures variation in the size of grains of sediment within a sample, perhaps representing habitat variety, following the rationale that a greater degree of sediment heterogeneity offers more potential niches . Surface roughness, in contrast, is a 100 m resolution dataset calculated as the standard deviation of the slope within a 1,000 m radius and is particularly intriguing since the biology is sampled over 0.05m 2 area and surface roughness integrates data from as far as 1,000 m away. The details behind this macrofauna-large-scale surface roughness relationship remain unresolved. It is possible this large-scale surface roughness is reflecting another environmental parameter, though it is not correlated to any parameter used in this study (see Appendix).
On a broad scale, macrofaunal community composition was found to change with mean water depth. In fact, this broad-scale parameter exhibited the highest correlation with the biology in the BIOENV procedure (p = 0.52). Depth appears to be valuable parameters in bottom-up habitat mapping studies    (Todd and Kostylev, 2011). Therefore, future studies aimed at resolving this variability should involve additional benthic and water column parameters. Furthermore, BIOENV does not demonstrate causality (Clarke et al., 2008). Possible explanations as to how each abiotic variable influences macrofaunal community patterns are discussed, but further investigation is needed to establish the causalities of the correlative links indicated for the BI and FED study areas.

Conclusion
Two benthic habitat classification approaches,          The average within-environment similarity and the genus most responsible for the within-group similarity, both identified by the SIMPER procedure, are also provided. It is interesting to note that for some environments, the same genus is the most abundant and is the most responsible for the within-group similarity. can be used to discern biotic-abiotic relationships between and among habitats and index metrics.

Figure 1.9. LINK.TREE output for BI and FED. A total of 22 classes (red numbers) were identified within BI and FED. Each class is defined by a series of quantitative thresholds of the six abiotic variables identified in the BIOENV procedure. The threshold for each split (black letters) is listed in
Two offshore locations within Rhode Island waters, selected as potential wind farm locations, serve as the basis for developing this methodology. The indices were able to identify habitats that scored considerably higher than the others. In general, though, the indices did not indicate specific biological or environmental characteristics that lend to high habitat value, which suggests management efforts need to consider all habitat types within the study areas, and cannot focus on certain habitat attributes.
However, a correlation was found between tube-building species and species richness, suggesting tube mat structures lead to increased biodiversity. The indices also show that habitats within the two study areas have different relationships with the index criteria, indicating macrofauna have their own associations to the environment within each study area. The proposed relationships between the index metrics and habitats will be evaluated within the two study areas in the near future.
The methodologies applied in this study can be extended to other locations and tailored to meet project objectives. The development of indices that signify habitat value will help bridge the communication gap between scientists and resource managers, and further the goal of science-based decision-making.

Introduction
Recent interest in development of offshore wind farms within Rhode Island waters has initiated a state-supported, collaborative study of marine resources known as the Rhode Island Ocean Special Area Management Plan (Ocean SAMP). The Ocean SAMP is a spatial planning tool to assist in making scientifically valid management decisions, including identifying appropriate locations for wind turbine installation (RI CRMC (a), 2010). A primary task of the Ocean SAMP was to map the distribution of benthic habitats and identify biological-environmental relationships. A thorough understanding of these habitats is essential to minimize the ecological and economical impacts of wind farm development. The mapping of benthic habitats presents characteristics of seafloor environments in a geospatial context (Auster et al.,

2009).
A major challenge facing the benthic habitat mapping community is presenting data and maps in a way that can effectively convey relevant information to a broad range of users ( e.g. scientists, managers, non-profit organizations, general public).
The information habitat maps should portray depends on the goal of the mapping project , which itself can also be difficult to define. Establishing a clear mapping purpose is important, since the type and resolution of data collected will determine the maps that can be produced . In addition, the lack of a standard benthic habitat classification approach has led to the development of numerous frameworks (e.g. Sneider et al., 2005;.
These schemes vary in their level of organization, detail, and geographic focus  such boundaries -that is, biological communities are present in multiple habitats and a defined habitat exhibits a range of biological communities , more information than dominant species is often needed to evaluate and understand benthic habitat distribution and patterns.
The purpose of this study is to develop an alternative to the "dominant species" approach to benthic habitat mapping. Biological and environmental metrics viewed as important to a wide range of users were identified and calculated for each habitat, including abundance, species richness and other biodiversity metrics, habitat stability, habitat-forming species, and habitat value as a food resource for demersal fish. From these criteria, an index of benthic habitat value was produced to identify habitat "hot spots." Methodology to construct an index, including weighting the metrics and summarizing the scores, was also developed. The final index presents the overall benthic habitat value and offers the scores of each metric for each habitat, allowing habitats to be evaluated according to individual metrics relevant to user needs.
Furthermore, indices can be used to discern biotic-abiotic relationships among habitats, have the potential to identify characteristics of benthic habitats that lend to high index scores, and can be further developed as additional data becomes available.
Two locations within the RI Ocean SAMP were selected to serve as the basis for developing this methodology. Previous studies  suggest that these areas differ in biotic and abiotic characteristics, providing a complex environment to classify and compare. The results of this study will be a valuable contribution for making ecosystem-based management decisions for Rhode Island waters, and serve as a pre-development baseline.
Beyond specific interests in our study areas, this study presents a method for describing benthic habitats that can be applied to any study location and can be tailored to any project objective. In addition, presenting benthic habitats thorough an index will help bridge the communication gap between scientists and resource managers, and further the goal of science-based decision making .

Study area
The Rhode Island Ocean SAMP study area is 3,800 sq km, primarily The benthic habitats of BI and FED differ in their abiotic and biotic characteristics . The BI study area exhibits a higher degree of physical heterogeneity than FED, having a wider range of environments, which tend to change over smaller spatial scales(> 2 sq km) (Figures 2.2 and 2.3). In addition, BIS is a more energetic area, subject to intense mixing due to storms, tidal circulation , and powerful current velocities (RI CRMC, 2010b), as evidenced by transitory geologic features such as large-scale sand waves, sheet sands, sand dunes, small-scale sand ripples, and the overall coarse sediment composition seen within the Bl study area. Alternatively, FED, located in the heart of RIS, appears to be a more stable environment, exhibiting milder current velocities (RI CRMC, 201 Ob), an overall finer sediment composition, and fewer transitory geologic features. In addition, RIS exhibits thermal stratification during warmer months .
These differences in physical environment likely influence benthic community structure and patterns within the two study areas. For example, the more stable environments of FED probably promote long-standing communities, whereas the environments that are transitory within BI are more challenging for organisms to withstand. Similarly, the summer-stratified waters of FED may adversely influence benthic communities in terms of food and nutrient supply, whereas the energetic environment of BI may offer favorable conditions. Benthic communities within BI may also be affected by nutrient input from coastal community activity (Block Island), which may lead to an increase in local production.

Previous benthic habitat classification maps
Benthic habitat classification maps have been developed for the BI and FED study areas using map units of depositional environment type (see section 2.5. 1 The average similarity of the macrofaunal assemblage within each habitat ranges from 6.2% to 59 .4%, with a mean of 34.2% (Table 2. 1;. Samples in habitat "B. serrata/A. agassizi -PJ-BB Moraine, sheet sand, sheet sand with gravel, sand waves" exhibited the most similarity (59.4%) , followed by the "A. agassizi -Glacial Lake Floor, sand sheets" habitat (58.3%). A variety of species were the most responsible or shared responsibility for the within-map unit similarity and contributed between 7.9% and 100% to the similarity.

Methods
Benthic habitats were classified according to eight biological and environmental metrics. These metrics were then weighted and used to develop indices of benthic habitat value for the BI and FED study areas. The metrics incorporated are average abundance, four measures of biodiversity (species richness, Shannon-Weiner index, Pielou's evenness, taxonomic diversity), value as fish food resource, presence of habitat-forming fauna, and habitat stability.
The methods section is structured around constructing the indices. As such, the first sub-section describes the habitat map units and how they were derived. The second sub-section focuses on the abundance and the biodiversity metrics -starting with the datasets needed, how the metrics are defined, and, lastly, how they were calculated. The next three sub-sections follow a similar format to describe the other three metrics. The last sub-section explains how the metrics were weighted and the indices developed.

Habitat map units
Geologic depositional environments were chosen as the map units for the index maps of BI and FED because full-coverage maps can be created and so the indices can be used in association with the benthic habitat maps developed previously.
Depositional environments were visually interpreted from high-resolution sidescan and bathymetry mosaics, sub-bottom seismic reflection profiles, surficial sediment samples, and underwater video

a. Macrofaunalsurvey
The macrofaunal survey (Figure 2 The remaining material from each bottom sample was sieved on 1 mm mesh and macrofauna were retained. All individuals were counted, identified to at least the genus level, and described according to functional group ( e.g. surface burrower, tubebuilder, mobile). The macrofauna abundances from BI and FED were pooled and only the genera contributing to 97% of the total abundance were included in further analyses. This eliminated genera with very low abundances(< 0.09% of the total abundance, equivalent to< 19 individuals) and resulted in the removal of 663 individuals from the study ( of 21,862). PRIMER 6 was used to 4 th root transform all abundances to reduce the influence of highly abundant genera and the Bray-Curtis similarity index was used to assess between station similarity.

b. Abundance metric
Abundance, defined as the number of individuals per bottom sample (0.05 m 2 area), was calculated as an average across all samples belonging to each map unit.

c. Biodiversity metrics
Biodiversity metrics are commonly considered to be indicators of ecosystem health  and stability . Thus, though the relationships between biodiversity and benthic ecosystems have not been evaluated for the BI and FED areas, it is anticipated increased biodiversity is associated with higher quality habitats. Biodiversity as an accepted measure of habitat value is the justification for its inclusion in the index.
Three of the biodiversity metrics describe biological assemblage structure (species richness, Shannon-Wiener diversity index, Pielou's evenness). Species richness refers to the total number of species and is a commonly used first-order measure of biodiversity. However, richness does not express how the diversity is distributed . Therefore, the Shannon-Wiener diversity index is often calculated as well because it takes species richness and relative abundance into consideration   Pielou, 1969).
The fourth biodiversity metric, taxonomic diversity, is used to complement assemblage structure metrics. Instead of focusing on the number of species, taxonomic diversity considers how related species are on a taxonomic level. Thus, samples with species belonging to the same taxa (genus, family, etc.) are considered to be less diverse than samples with species that belong to wider variety of taxa . This metric was calculated from species to genus between every pair of individuals (Clarke and Gorely, 2006; .
All of the biodiversity metrics were calculated using PRIMER 6 (for explanations of equations refer to . High values suggest high biodiversity and evenness, and thus, those habitats are considered to be the most valuable.

a. Fish stomach content analysis
Stomach

b. Value as fish food resource metric
The value of each habitat as a demersal fish food resource was included in the index because benthic organisms, particularly amphipods, can be an important trophic link, as they are a valuable food source for demersal fish , including within RIS and BIS (Malek et al., 2010;RI CRMC, 2010b). Food resource value was evaluated by comparing the prey identified in the stomach analysis to the species found within each habitat. The habitats with the highest percent composition of prey available to demersal fish are viewed as most valuable.

Habitat-forming fauna
Habitat-forming species refer to organisms that create biogenic reefs. Habitatformers are ecologically important, as they can stabilize sediment, provide complex structures for other species to utilize as habitat or refuge, and be an important food source for benthic predators (Callaway et al., 201 O;. Functional descriptions of the recovered macrofauna species were used to determine the presence of habitat-forming species within BI and FED. Of the 87 species identified within the study areas, 18 are considered to be habitat-forming. These species include blue mussels , which create structure from calcareous aggregations , and tubebuilding amphipods and polychaetes that form dense mats of sediment tubes  extending 5-10 ems above the surface . The more of these species present within a habitat, the more valuable the habitat, due to its reef-building potential.

a. Underwater video survey
Video transects were taken at 42 of the macrofaunal sample locations within BI using an Applied Microvideo underwater video camera and two LED lights mounted to a PVC sled . At each station, the sled was towed behind the drifting vessel for five minutes, resulting in transects that averaged 130 m in length and ranged from 30 m to 230m .

b. Habitat stability metric
The habitat stability metric was included in the index to infer temporal variability of physical habitat structures and biological communities. Physical stability was assessed based on characteristics of each habitat, as indicated from geologic depositional environment (see section 4.1) and underwater video data , and was classified according to three categories. The first category , "stable benthos and water column," was assigned to environments dominated by fine sediments (i.e. silt to fine sand). The existence of such substrate indicates there is weak water movement (i.e. currents and/or tides) in the area; otherwise the fine material would be carried away . The second category is "stable benthos, active water column" and was used to denote environments dominated by gravel, cobble, and boulders. The benthos here is considered to be stable because the relatively large, heavy substrate is nonmobile, since water movement will not carry it away. However, there is sufficient water movement to prevent the settlement of finer-grained sediments (i.e. silt to very coarse sand). The third category is "active benthos and water column," which includes transitory environments, such as sand waves and ripples, sand dunes, and sheet sands. Such environments are mobile due to intense, high velocity currents, tidal action, or storm activity .

Index development
To develop an index of benthic habitat value for each study area, the seven objectively-derived metrics were weighted. Physical habitat stability was not weighted because it is a subjectively determined categorical metric , and the classifications do not imply negative or positive implications. A scale of zero to three was chosen for the weights to emphasize top-ranking habitats and for practical purposes; ranking the 18 habitats from one to 18 for seven metrics would be unmanageable.
A weight of three was assigned to the map unit considered most valuable for each metric (i.e. the map unit with the highest abundance, the unit with the highest species richness, the unit with the most prey available to demersal fish, etc.). Similarly, a weight of two was given to map units that rank second ( e.g . second highest abundance) and a weight of one given to units that rank third. All other units were assigned a value of zero. The scores for each habitat were then totaled. The highest possible score any habitat can achieve is 21 and the lowest is zero.
The resulting index maps were color coded to emphasize the range of values.
And physical habitat stability was indicated with hatch and stippling patterns.
Additionally , a table describing the scores of each metric for each habitat was created to allow detailed interpretation of the indices.

Abundance
More than 21,000 individuals belonging to seven phyla and 87 genera were sampled across the 78 stations within BI and FED. For both areas, the majority

Index of benthic habitat value
The habitats are numbered 1-18 in each figure and table for referencing purposes.
Habitats 1-12 are found within BI and 13-18 within FED. As described in the methods, the habitat map units are defined by depositional environments and contain significantly distinct macrofaunal assemblages.

a. BI study area
The resulting index for BI contained 12 habitats with scores ranging from zero to 12 (Figure 2.7, Table 2.2). The index reveals that there are no specific dominant species, depositional environment type, or habitat stability category that yields high index scores. Instead, the habitats with the highest index scores exhibit a wide range of abiotic and biotic characteristics, ranging from coarse sand with small dunes to pebble gravel coarse sand to silty sand and from tube-building amphipods to tubebuilding and surface-burrowing polychaetes. The habitats scoring the lowest values also possess a range of characteristics. In fact, the high and low scoring habitats are defined by some of the same features. Furthermore, there are no clear patterns among the index variables. Scores of one, two, and three were distributed across most of the habitats, with the exception of species richness and number of habitat forming species, which scored high in the same three environments.
The highest index value of 12 belongs to habitat 10 and is mainly due to the biodiversity metrics. The habitat is dominated by Polycirrus medusa and Lumbrineries hebes, tube-building and surface burrowing polychaetes, respectively.
In addition to co-dominating the abundance, L. hebes also contributes most to the similarity (14.48%) among all of the macrofaunal samples within that habitat (refer to Geologically, the habitat is part of the inner shelf moraine and exhibits transitory features -coarse sand with small dunes, sheet sands, and sand waves. As such, the area is considered to have an active benthos and water column. The remaining habitats had an index values four or less. Three habitats (2, 6, and 12) scored a value of zero. There were no commonalities between these three habitats.
Dominant species ranged from the tube-builder A. vadorum, B. serrata and P. medusa to the mobile polychaete genus, Syllis. Depositional environments included glacial delta plain -pebble gravel coarse sand, glacial alluvial fan -sheet sand, and glacial alluvial fan -sand waves. Accordingly, the habitats were categorized as having active water columns and stabile or active benthos.

b. FED study area
The index scores for the six habitats within FED ranged from 15 to two ( Figure   2.7, The third highest index value, seven, belongs to habitat 13, exhibiting high average abundance, species richness, number of habitat-forming species, and acting as a valuable food source for demersal fish. The habitat is dominated by A. agassizi, but N delphinodonta and N annulata are most responsible for the within-habitat similarity (7.93% and 6.97%, respectively). Glacial lake floor with sheet sands define the habitat, and it is characterized as having an active benthos and water column.
The remaining habitats scored between two and six. Like with BI habitats, these lowest scoring habitats possess different abiotic and biotic characteristics. They are defined or co-defined by B. serrata, A. agassizi, or N annulata and the depositional environments are hummocky moraine -fine sand, moraine -sheet sand/sheet sand with gravel/sand waves, and glacial lake floor -coarse silt. Habitat stability is defined as either stable benthos and water column or active benthos and water column.
Though, the two lowest scoring habitats share two commonalities, being defined by fine sediments (fine sand or coarse silt) and considered to have a stable benthos and water column.

7. Discussion
The goal of this study was to develop indices of benthic habitat value for two offshore study areas (Bl and FED) targeted as primary sites for potential wind farm

Identifying benthic habitat "hot spots"
Benthic habitat "hot spots" were clearly identified by the BI and FED indices (habitat 1 and 10, respectively). These "hot spots" were relative to each index ; the habitat with the highest index value scored five points more than the second highest scoring habitat.  ( 17), having the highest index value ( 15), scored in every criterion, except one, but did not overshadow the other habitats. In general, the results indicate management efforts need to consider all habitat types, and cannot focus on certain habitat attributes.

Comparison of BI and FED indices
As the top-scoring habitats suggest, the BI and FED indices were quite different.
The relationships between abiotic and biotic characteristics that appear to exist within BI do not within FED, and vice-versa. For example, examination of the BI habitats suggest the highest evenness occurs in environments defined by coarse sand with small dunes and are categorized as having an active benthos and water column.
However, within FED, this relationship does not hold true. In fact, none of the top three ranking evenness habitats within FED are defined by coarse sand with small dunes and habitat stability varies. Rather than disproving potential relationships, however, the differences between BI and FED speak towards the macrofauna having their own associations to the environment within each study area, supporting previous findings of the BI and FED study areas .

Biodiversity and tube-building fauna
Habitat-formers, such as tube-building fauna, are ecologically important, as they can provide complex structures for other species to utilize as habitat or refuge, stabilize sediment, and be an important food source for benthic predators (Callaway et al., 201 0;. Consequently, habitat-formers tend to create areas of increased biodiversity relative to the surrounding environment . Previous studies (e.g.  have reported positive relationships between habitat variety and species diversity, following the rationale that a greater degree of sediment heterogeneity offers more potential niches, and therefore, allows for higher diversity . For example,  reported that suspension feeders (such as tube-building amphipods) physically dominate hard surfaces, but, despite this, a diverse range of fauna ( deposit feeders, predators, browsers) reach high densities in mature epifaunal assemblages.
That habitat forming-species and species richness were correlated and that high biodiversity tended to occur in habitats defined at least in apart by tube-building species, indicates tube-builders and/or their dense mats positively influence benthic ecosystems. Studies have suggested polychaete tube-mat structures increase sediment heterogeneity (i.e. habitat complexity), leading to increased biodiversity . In addition, tube-builders specialize in resource uptake by building tubes that extend 5-10 cm above the seafloor. This strategy allows tube-builders to avoid competition for resources on the seafloor and allows them to obtain the more nutritious food that tends to concentrate a few centimeters above the seafloor in the water column . It is also possible that tube-builders positively interact with other species (predator-prey, competition, mutualism).

Biodiversity and habitat stability
While tube-building fauna are positively related to biodiversity, the indices suggest that the highest biodiversity is achieved when tube-builders co-dominate a habitat. Possibly tube-building fauna are able to out-compete other species for resources, as they do tend to occur in very high densities. In these study sites, three   .
Disturbances can be physical (e.g. storms, currents, tides) or biological (e.g. predation). If a habitat is very stable, diversity is reduced due to the competitive exclusion -species that are optimally adapted for that environment will out-compete others. Conversely, if the intensity and frequency of environmental disturbance is too high, it may present conditions too stressful for many species, also resulting in reduced diversity .  for example, noted that within RIS and BIS organisms living in active environments must be adapted for movement in sand and be able to recover from periodic burial. At intermediate disturbance levels diversity is highest because there is less competitive exclusion, which frees up resources for other species to utilize, and conditions are tolerable to a wider range of species .

7 .5. Biodiversity metrics
Biodiversity metrics played an important role in developing the indices (four metrics are included) because they are considered to be indicators of ecosystem health  and stability , and because they indicate how biological communities respond to their environment . For example, evidence suggests that as species richness increases there is increased primary production, as well as increased resistance to natural disturbances and invasion within a community . Furthermore, biodiversity has been the focus of some conservation efforts ( e.g. . While using four measures of biodiversity may seem repetitive, especially in FED (where three of the metrics were correlated), each measure may be responding to different environmental parameters, and therefore, be valuable independent metrics .  reported even significantly related biodiversity metrics revealed significantly distinct relationships with different environmental variables, and therefore could not be considered redundant. In this study, the relationships between biodiversity and habitat-forming fauna suggest the biodiversity metrics may represent habitat heterogeneity.

Temporal variability
Temporal variability can present a challenge to benthic habitat mapping, both in data collection and in creating final products. Because maps are created using abiotic and biotic datasets representing single sampling/survey events in time, they often do not reflect the temporal dynamics of transitory features. However, qualitative descriptors of temporal variability may be inferred, as was the purpose of including the habitat stability parameter in the indices. For example, within unstable physical environments (mobile sheet sands, sand waves, sand ripples), characteristics (abiotic and biotic) of the benthic habitats are more likely to change. With regard to biotic data, temporal variability may be indicated by the presence of opportunistic species that reflect recent habitat disturbance, or the presence of large, long-lived individuals that indicate a more stable environment and potentially lower temporal variability in macrofauna composition .
It is possible seasonal differences in macrofaunal community composition are reflected in these results and that the indices may become outdated. However, Steimle (1982) reported there were no clearly defined seasonal changes between biological communities examined in February and in September within BIS. Steimle (1982) also presented evidence to suggest these habitats are relatively stable on a time-scale of decades.

Applicability
The methodology presented here can be applied to a broad range of environments, as evidenced by the success of the indices in identifying benthic habitat "hot spots" at two study areas differing in their abiotic and biotic characteristics. Moreover, the criteria incorporated into the indices can be tailored to meet individual project needs, and indices can be further developed as additional data becomes available . A table of relevant habitat attributes identified as important by a range of user groups is nicely presented in . Following this table, other criteria that may be relevant in developing benthic habitat indices include finer-scale sediment data or water column processes, organic carbon content, chlorophyll-a concentration, importance of habitat for larval recruitment, and degree of anthropogenic impact/human-induced attributes (such as from construction, dredging, fishing).
Biologically, the presence of species of interest ( e.g. key species, indicator species, endangered, commerciall y important) and biodiversity metrics of species rarity or taxonomic distinctness may be informative criteria . Along with dominant species and species contributing most to the habitat similarity , it may also be useful to label habitats according to dominant species groups.

Future work
This study represents an initial attempt to construct indices of benthic habitat value. The index results and proposed relationships will be verified through collection and analysis of additional data in the near future. For instance, the relationship between the diet of demersal fish and biodiversity will be evaluated throughout RIS and BIS . The analysis will involve examining demersal fish stomach contents to determine if their diet diversifies in areas where more types of prey are available. In other words, "Do fish take advantage of diverse habitats or just focus on eating amphipods within any given habitat?" If the correlation is positive, it supports that increased biodiversity is beneficial to benthic ecosystems within BI and FED. If there is no relationship, it would indicate certain food types (i.e. amphipods) are preferred, and, thus, the degree of biodiversity is unimportant to demersal fish.
With regard to biodiversity , future studies will assess the appropriateness of including four biodiversity metrics into the index by examining what the metrics represent and if they are repetitive . In addition , though high biodiversity is anticipated to be positively associated with benthic habitat value, future studies will evaluate such relationships within BI and FED.

Conclusion
Resource managers are increasingly faced with dwindling budgets and a lack of easily applicable, science-based methods with which to make far-reaching management decisions, such as locations for wind turbine installation. This paper addresses this issue with the development of an index to identify benthic habitat "hot spots" that can be applied to any study location and be adapted to meet any project objectives. Indices were constructed for two study areas within Rhode Island waters by classifying habitats according a suite of biological and environmental metrics considered relevant to a broad range of user groups. Previous research has shown the habitats contain significantly distinct macrofaunal assemblages. The indices present overall benthic habitat value and offers scores of each metric, allowing habitats to be evaluated based on user need. Each index identified a habitat that scored considerably higher than the other habitats. In general, though, the indices did not indicate specific abiotic or biotic characteristics that lend to high habitat value, which indicates management efforts need to consider all habitat types within the study areas, and cannot focus on certain habitat attributes . However, a correlation was found between tube-building species and species richness, suggesting tube mat structures lead to increased biodiversity. The indices also show that habitats within the two study areas have different relationships with the index criteria, indicating macrofauna have their own associations to the environment within each study area. Biodiversity metrics play a large role in development of the indices, as they are considered to be indicators of ecosystem health and stability. This expectation and the proposed relationships among the index metrics and habitats will be evaluated within the two study areas in the near future.

Map Key·
Bl