2013 NORTHEASTERN NATURALIST 20(1):155–170 Habitat Selection by the Rock Gunnel, Pholis gunnellus L. (Pholidae) Jacob T. Shorty1 and Damon P. Gannon1,2,* Abstract - Pholis gunnellus (Rock Gunnel) is an amphibious fish found along North Atlantic coastlines. Remaining above the waterline at low tide and breathing air, it is ecologically unusual and is a food source for a variety of seabirds, mammals, and fish. We investigated intertidal and subtidal habitat selection by the Rock Gunnel along the US East Coast. To quantify characteristics of intertidal microhabitats and their usage by the Rock Gunnel, we conducted intertidal quadrat surveys in midcoast Maine and New Brunswick. Using datasets from trawl surveys conducted by state (Massachusetts and Connecticut) and federal fishery management agencies, we investigated how subtidal habitat characteristics influenced the occurrence of Rock Gunnel in trawl yields. Logistic regression and classification and regression tree (CART) analysis showed that Rock Gunnels preferred microhabitats in the lower intertidal zone, with sand/pebble/gravel substrata and overlying cobbles, tidepools, and dense algal cover. The occurrence of Rock Gunnel in trawl yields did not depend upon temperature, salinity, or latitude, but decreased with depth and increased moving eastward. In the intertidal zone, the Rock Gunnel appears to select habitat that minimizes the risks of predation and desiccation at low tide, while allowing access to abundant intertidal prey resources at high tide. The Rock Gunnel’s broad physiological tolerances suggest that its selection of habitat in the subtidal zone is driven primarily by the availability of sheltering structure and biotic factors, rather than by temperature or salinity. Introduction The use of a habitat disproportionately to its availability is defined as habitat selection, and can provide insight on the importance of various resources and conditions to the study species (Johnson 1980). In selecting habitat, animals face a tradeoff between a habitat’s benefits (e.g., food availability) and its risks (e.g., predation) (Werner et al. 1983a, b). One simple method proposed for quantifying the fitness consequences of tradeoffs associated with habitat selection is calculating the ratio of mortality (μ) to growth (g) (Werner and Gilliam 1984). Selecting the most profitable habitats from among all those that are available should maximize fitness and minimize μ/g. The intertidal zone is a patchy environment of fine-scale heterogeneity. For the animals that inhabit the intertidal zone, the value of adjacent microhabitats can vary drastically. For example, habitats that are in close spatial proximity can vary with regard to food availability, risk of predation, protection from desiccation, degree of competition, thermal conditions, dissolved oxygen concentration, and turbulence of the water. 1Department of Biology, Bowdoin College, 6500 College Station, Brunswick, ME 04011. 2Bowdoin Scientific Station, Bowdoin College, 6500 College Station, Brunswick, ME 04011. *Corresponding author - dgannon@bowdoin.edu. 156 Northeastern Naturalist Vol. 20, No. 1 Pholis gunnellus L. (Rock Gunnel) is a small, perciform fish, capable of breathing air and remaining in the intertidal zone during low tide (Fig. 1). It is found in two populations of relative isolation along the North Atlantic coastlines of Europe and North America, with the western North Atlantic population ranging from Greenland and Labrador to Delaware Bay (Hickerson and Cunningham 2006, Mecklenberg 2003). Most frequently encountered on rocky shorelines, Rock Gunnel is also recorded in benthic trawl surveys further from the shore, to depths of 183 m (Schroeder 1933). The Rock Gunnel is an important food source for a variety of seabirds, piscivorous fishes, and mammals (Birks and Dunstone 1985, Blackwell et al. 1995, Collette 2002, Cote et al. 2008, Hawksley 1957, Rail and Chapdelaine 1998, Winn 1950). Although the majority of these animals likely feed upon the Rock Gunnel while it is underwater (Cairns 1992), the extent of predation pressure it faces when exposed by the tide is unclear. The Rock Gunnel mainly eats small crustaceans (primarily amphipods and isopods), polychaetes, molluscs, and gastropods (Cheetham and Fives 1990, Qasim 1957, Sawyer 1967, Stroud 1939). While almost nothing is known about the foraging behavior of the Rock Gunnel, its benthic lifestyle and broad diet suggest that it is a sit-and-wait predator. The intertidal fishes of temperate Atlantic waters demonstrate a seasonal variation in presence and abundance within the intertidal zone (Moring 1990). Rock Gunnels are rarely encountered in Maine’s intertidal zone during November and are completely absent between December and early March (Moring 1990, 1993; Sawyer 1967). Noting this absence, Sawyer (1967) hypothesized that Rock Gunnels migrate to deeper waters of the Gulf of Maine Figure 1. Rock Gunnel (Pholis gunnellus) (photograph © Evan Graff). 2013 J.T. Shorty and D.P. Gannon 157 to avoid cold air temperatures. Spawning occurs in late winter, with one or both parents guarding a demersal egg mass (Gudger 1927, Qasim 1957). Upper estuarine habitats serve as key spawning/nursery areas for Rock Gunnels along Long Island Sound and the Gulf of Maine (Chenoweth 1973, Pearcy and Richards 1962). Along the central Maine coast, larval abundance peaks from February to April (Chenoweth 1973). As an intertidal fish, the Rock Gunnel must deal with changes in environmental conditions not experienced by fishes restricted to subtidal waters. Classified as a “remainer” species, the Rock Gunnel is passively exposed by the ebb tide and remains relatively inactive while emersed, in contrast to other intertidal fish, which may actively emerge for shorter durations (Martin 1995). It has several physiological traits allowing it to survive out of water for hours, including low permeability of the skin and gills (Evans et al. 1999) and the ability to breathe air using “gaping behavior” facilitated by a heavily vascularized esophagus (Laming 1983). Additionally, the Rock Gunnel retains nitrogenous waste until reimmersion to prevent unnecessary water loss (Kormanik and Evans 1988). Typically associated with the rocky intertidal zone, the Rock Gunnel can be found beneath rocky cobbles on the shoreline. While there are few data available characterizing the Rock Gunnel’s intertidal habitat use, there is some evidence of habitat preference. Underwater visual surveys of intertidal/ shallow subtidal areas at high tide in Nova Scotia (northeast of Halifax) found Rock Gunnels within beds of Ascophyllum nodosum (L.) Le Jolis (Rockweed) but not Zostera marina L. (Eelgrass) (Schmidt et al. 2011). In Scotland, density of Rock Gunnel decreased with height in the intertidal, with the most fish found closest to the low-tide line (Koop and Gibson 1991). At a study site in Maine, Rock Gunnels were found using specific rock crevices across several seasons, demonstrating fine-scale selection of specific microhabitat from among other available choices, though little homing capability was evident (Moring 1993). However, a study along the Massachusetts coast found that the use of two tidepools by the Rock Gunnel was regular but highly variable across multiple seasons (Collette 1985). There is also evidence for fine-scale habitat selection in other Pholidae, using algal cover matching body coloration (Burgess 1978, Winkler 1988). We investigated the influence of various habitat features on fine-scale habitat selection by the Rock Gunnel in the intertidal and subtidal zones. Using logistic regression, we developed a predictive model to analyze the influence of physical and biological habitat characteristics on Rock Gunnel habitat selection from intertidal quadrat surveys. We identified critical threshold values for intertidal environmental variables by constructing a classification and regression tree. Additionally, using long-term datasets available from trawl surveys conducted by federal and state fisheries management agencies, we examined subtidal habitat selection by the Rock Gunnel using a second logistic regression model. 158 Northeastern Naturalist Vol. 20, No. 1 Field-Site Description Fieldwork was conducted at various intertidal locations in midcoast Maine, from Harpswell north to Rockland along the coast, and on Kent Island, NB, Canada, in the Bay of Fundy (Fig. 2). These locations varied in their intertidal productivity and habitat composition, featuring rocky shores, sand beaches, and mud flats. Additionally, they ranged in their exposure to wave energy from sheltered coves to headlands exposed directly to open ocean. Plant/algal communities were dominated by Rockweed and three Fucus species (F. distichus L., F. spiralis L., and F. vesiculosus L.) along the rocky shores used in this study, while Eelgrass was the most common in the mudflat areas. Methods Intertidal habitat selection To determine which intertidal habitat characteristics correlate with use by the Rock Gunnel, we conducted quadrat surveys along transects within the intertidal zone, parallel to the water’s edge. All quadrats were placed within 1 m of the water. Thus, time from low tide could be used as a proxy for vertical position within the intertidal zone. Each transect consisted of twenty 1-m2 quadrats, with one quadrat surveyed roughly every 5 m over a total distance of Figure 2. Study sites in Midcoast Maine and New Brunswick (inset) used in intertidal surveys conducted between June and August 2011, to examinine habitat selection by Pholis gunnellus (Rock Gunnel). 2013 J.T. Shorty and D.P. Gannon 159 approximately 100 m. Standing at the waterline, quadrats were tossed behind the observer along the shoreline in the direction of the transect. Quadrats that landed partially or entirely in the water were brought to the closest point on the shore. We surveyed at all stages of the tidal cycle to sample the entire vertical range of the intertidal zone. The tidal cycle was divided into thirds (high, middle, and low water periods) and each period was sampled equally. We noted the number of Rock Gunnels present within the quadrat and measured a variety of other biological and physical parameters (Table 1). First, the time at which the quadrat was sampled was recorded to determine the time from low tide (used as a proxy for vertical position in the intertidal zone, as described above). Wave energy was classified using a method similar to that of Ekebom Table 1. Variables measured in intertidal quadrat surveys and examined using logistic regression and classification and regression tree analysis to characterize intertidal habitat selection by Pholis gunnellus (Rock Gunnel) in Maine and New Brunswick. Type of variable Variable name Description Physical Time from low tide Minutes from time of low tide; used to extrapolate relative vertical height of quadrat in intertidal (with the quadrat located within 1 m of the water’s edge). Physical Fetch Maximum of four estimated measurements of fetch in the cardinal directions, as a proxy for wave energy; ranges are categorized as 0, <200 m, 200–1500 m, 1501–5000 m, and >5000 m. Physical Slope Slope of quadrat substrata perpendicular to the waterline; estimated to nearest 10°. Physical Substratum Dominant substratum in quadrat; categorized as either mud, fine sand, sand/pebbles/gravel (mixed), rocky cobbles, or solid rock. Physical Rocky cobbles Presence or absence of any rocky cobbles (rocks larger than roughly 5 x 5 x 5 cm overlying substratum). Physical Standing water Presence or absence of any standing water in quadrat. Physical Tidepools Presence or absence of water bodies larger than 5 cm x 3 cm x 3 cm (large enough to immerse an individual Rock Gunnel). Physical/biological Algal/plant cover Percent cover of plants and algae in quadrat, estimated to nearest 10%. Physical/biological Dominant algae/plant Algae/plant with the greatest percent cover in quadrat; identified to highest specificity possible in the field. Biological Amphipod/isopod density Total amphipods/isopods within quadrat; ranges are categorized as 0, <10, 10–30, 31–50, or >50. Biological Large crustaceans Total Green Crab, Atlantic Rock Crab, Asian Shore Crab, and American Lobster in quadrat. 160 Northeastern Naturalist Vol. 20, No. 1 et al. (2003); the amount of fetch in each of the 4 cardinal directions from the quadrat was classified using 5 broad ordinal categories of distance derived from NOAA nautical charts (charts 13290, 13293, and 13302) (Table 1). We used the maximum of the 4 fetch values recorded in our analyses. In each quadrat, we categorized the dominant substratum and dominant plant or algae, recorded the slope of the substratum, percent cover of algae and plants, and the combined density of amphipods and isopods (Table 1). Additionally, we noted the presence or absence of rocky cobbles, standing water, and tidepools, and counted the number of individuals of several potential competitors of the Rock Gunnel: Carcinas maenas L. (Green Crab), Cancer irroratus Say (Atlantic Rock Crab), Hemigrapsus sanguineus De Haan (Asian Shore Crab), and Homarus americanus Edwards (American Lobster), which were added together in the large crustaceans variable. Using the entire quadrat survey dataset, we performed binary logistic regression analysis using SPSS 19 (IBM, Armond, NY) to create a statistical model for predicting the occurrence of Rock Gunnel. Since the majority of Rock Gunnel occurrences involved only one individual, we converted the number found in each quadrat into a binary presence/absence response variable. Due to correlation with a number of other predictor variables, we excluded the large crustaceans variable from this and all subsequent analyses. In addition to single terms, two-way interactions between the variables with statistically non-significant Spearman rank correlation coefficients were also included in the initial model. First, we examined each variable’s explanatory power using backwards stepwise selection. Predictors were retained in the model when they were at a P-value less than 0.10 based upon the likelihood ratio statistic. We then compared models based on goodness of fit using the corrected Akaike information criteria (AICc), which penalizes model complexity to prevent overfitting (Hurvich and Tsai 1989). The set of predictor variables that resulted in the lowest AICc value was selected as the final model. We performed classification and regression tree (CART) analysis to identify critical thresholds in the intertidal habitat characteristics that predict the occurrence of Rock Gunnel. The suite of variables used was the same as those examined by logistic regression. CART modeling has been used extensively to understand resource selection (Bourg et al. 2005, De’ath and Fabricius 2000, Eby and Crowder 2002, Friedlaender et al. 2006, Rejwan et al. 1999). CART analysis is a recursive partitioning technique in which the values of a single response variable are explained by one or more explanatory variables. In CART analysis, a tree is created by splitting the data into 2 mutually exclusive subsets based upon the values of a single explanatory variable, with the goal of maximizing homogeneity within subsets while maximizing heterogeneity between them (De’ath and Fabricius 2000). This process is repeated on resulting subsets with the aim of separating the data into homogenous groups (De’ath and Fabricius 2000). A conditional inference tree was constructed using the PARTY algorithm in R (Hothorn et al. 2005, 2006). 2013 J.T. Shorty and D.P. Gannon 161 Subtidal habitat selection We obtained Rock Gunnel catch data from trawl surveys to analyze subtidal habitat characteristics associated with its presence. Surveys were conducted by NOAA’s Northeast Fisheries Science Center (NEFSC), the Massachusetts Division of Marine Fisheries, and the Connecticut Marine Fisheries Division. Data from trawls included Rock Gunnel catch yields, bottom temperature, depth, salinity, latitude/longitude, and date for each trawl station. The largest dataset was provided by the NEFSC, consisting primarily of spring and fall trawl surveys of equal temporal/spatial sampling effort on the US East Coast north of Cape Hatteras, NC from 1965 to 2009. The Massachusetts Inshore Trawl Survey (MISTS) covered state waters in Massachusetts from 1978 to 2009, and the Long Island Sound Trawl Survey (LISTS) encompassed the waters of Long Island Sound (state waters of both Connecticut and New York) from 1985 to 2010. Details on the trawl surveys are provided in NEFSC 2012, CTDEP 2012, and MDMF 2012. Similar to the analysis used in the microhabitat surveys, we performed logistic regression to examine the relative importance of environmental and survey characteristics in predicting Rock Gunnel catches using data from the NEFSC, MISTS, and LISTS surveys. Predictor variables included: survey, season, depth, bottom temperature, bottom salinity, latitude, and longitude. Again, backwards stepwise selection was used based upon likelihood ratio and the final model was selected based on minimized AICc. Results Intertidal habitat selection A total of 1548 quadrats were surveyed and used for analysis, from 90 transects. The logistic regression model incorporated 4 single terms and 2 two-way interactions, including time from low tide, slope, tidepools, substratum, the interaction between rocky cobbles and algal/plant cover, and the interaction between fetch and amphipod/isopod density (Table 2). The final model correctly predicted 97.4% of cases and was highly significant (model chi-square = 221.34, df = 24, P < 0.001). A Hosmer and Lemeshow goodness-of-fit test indicated the model was well calibrated (chi-square = 0.26, df = 8, P > 0.999). All predictors retained in the model were statistically significant (P < 0.05) aside from the interaction of fetch and amphipod/isopod density’s overall effect (P = 0.106) and each pairwise combination of these categories (P > 0.191). While the overall effect for substratum was significant (P < 0.001), none of the individual categories comprising this variable was significant (P > 0.995). Occurrence of Rock Gunnel decreased significantly with increasing height in the intertidal zone (P < 0.001). The odds of a Rock Gunnel being present in a quadrat decreased by 4% per min before or after low tide, as demonstrated by the odds ratio (eB = 0.96). Increasing slope resulted in a slight, significant increase (6% per 10°) in the probability of encounter (eB = 1.03, P = 0.048). The presence of tidepools in a quadrat also significantly increased the odds of Rock 162 Northeastern Naturalist Vol. 20, No. 1 Gunnel occurrence (P = 0.008); quadrats with tidepools were 3.2 times more likely to contain Rock Gunnels (eB = 3.16). While the B values of individual substratum categories have high standard error (817.39), the sand/pebbles/ gravel (mixed) category had the greatest increase in odds ratio (eB = 135.86) compared to the overall effect. Finally, when rocky cobbles were present, the odds of Rock Gunnel presence increased by 3% with every 10% increase in total algal cover (eB = 1.03). The final conditional inference tree produced by the CART analysis contained 6 recursive splits resulting in 7 final nodes (Fig. 3). The habitat characteristics contributing the most to the model included time from low tide and substratum (P < 0.001 for each), and produced the final node (Node 4) with the highest Rock Gunnel occurrence rate, in which 47.1% of quadrats examined within 36 minutes of low tide on mixed substratum contained Rock Gunnels. Other variables that improved the model’s predictability when used as splitting criteria included amphipod/isopod density (P < 0.001), rocky cobbles (P = 0.011), and slope (P = 0.020). Splitting criteria used in the tree separated 1060 cases (68.4%) into a single final node (Node 13) with no occurrences of Rock Gunnel. The node prior to this split contained 1102 cases (71.2%) and contained only one quadrat containing Rock Gunnels; only one quadrat surveyed more than 84 minutes from low tide contained Rock Gunnels. Subtidal habitat selection A total of 13,438 trawl stations were included in the analysis, 62 (0.5%) of which reported Rock Gunnel. The final model correctly predicted 98.9% Table 2. Binary logistic regression analysis for variables predicting the presence of Pholis gunnellus (Rock Gunnel) in quadrats from intertidal surveys conducted June through August 2011 in Maine and New Brunswick (model chi-square = 221.342, df = 24, P < 0.001, n = 1548). B values and odds ratios (eB) for substratum categories reference substratum’s overall effect. Interactions between fetch and amphipod/isopod density’s categories are not shown and were not statistically significant factors (P > 0.05). SE = standard error, df = degrees of freedom. Predictor variable B SE of B df P eB Time from low tide -0.04 0.01 1 less than 0.001 0.96 Slope 0.03 0.02 1 0.048 1.03 Tidepools (present) 1.15 0.44 1 0.008 3.16 Substratum (overall) 4 less than 0.001 Mud -13.62 3269.57 1 0.997 0.00 Fine sand 3.94 817.39 1 0.996 51.52 Sand/pebbles/gravel (mixed) 4.91 817.39 1 0.995 135.86 Rocky cobbles 1.08 817.39 1 0.999 2.94 Solid rock 3.69 817.39 1 0.999 40.09 Rocky cobbles (present) * algal/plant cover 0.03 0.01 1 less than 0.001 1.03 Fetch * amphipod/isopod density 16 0.106 Constant -6.58 817.39 1 0.994 0.00 2013 J.T. Shorty and D.P. Gannon 163 Figure 3. Classification and regression tree (CART) output of Pholis gunnellus (Rock Gunnel) occurrence in quadrats from intertidal surveys, in relation to habitat characteristics. Final nodes show the proportions of quadrats containing Rock Gunnels. Min_from_low = minutes from low tide, substrate = dominant substratum, cobble = presence/absence of rocky cobbles, amph_iso = amphipod/isopod density, and slope = quadrat slope. Categories for substratum (a = mud, b = fine sand, c = sand/pebbles/gravel, d = rocky cobbles, and e = solid rock) and amphipod/ isopod density (a = 0 individuals, b = 1–10 individuals, c = 11–30 individuals, d = 31–50 individuals, and e = >50 individuals) are lettered. See Methods for complete list of habitat characterist ics included in analysis. n = 1548. 164 Northeastern Naturalist Vol. 20, No. 1 of cases and was highly significant (model chi-square = 835.03, df = 11, P < 0.001). Of the initial set of terms, only salinity and latitude at the trawl station were removed during the model selection process (Table 3). All retained terms were highly significant (P < 0.001) except for bottom temperature (P = 0.052). The odds of encountering Rock Gunnel at a given trawl station decreased with depth (B = -0.03) and increased moving eastward (decreasingly negative longitude, B = 0.42). Discussion This is the first investigation using quantitative modeling techniques to study habitat selection by the Rock Gunnel. Rock Gunnels used intertidal habitat nonrandomly. Vertical height within the intertidal played a primary role in predicting the habitat choices of the Rock Gunnel, as demonstrated by the significant influence of the time from low tide variable in both the logistic regression and CART analyses. These models also indicated that the Rock Gunnel selected for sand/ pebble/gravel and fine sand substrata with overlying cobbles, dense algal cover, tidepools, and non-flat quadrat slopes. Densities of potential prey (amphipods and isopods) were not influential (the split in the CART model based upon this variable is dictated by the occurrence of only one individual). Rock Gunnels in this study demonstrated a clear preference for habitat in the lower intertidal zone, which is congruent with the findings of Koop and Gibson (1991). Individual fish were found above the waterline no more than two hours from low tide. Another Pholid, Apodichthys flavidus Girard (Penpoint Gunnel) has survived far longer periods of emersion, up to 20 hours (Horn and Riegle 1981). During emersion, the Rock Gunnel faces a lack of mobility, likely inhibiting its ability to forage and to evade terrestrial predators. Additionally, emersion probably causes physiological stress from rapid changes in temperature and desiccation. Mud habitat, which was not used by Rock Gunnels, may be avoided due to low dissolved oxygen content. Despite the Rock Gunnel’s physiological capability to endure conditions out of water (Evans et al. 1999, Kormanik and Evans 1988, Laming 1983), our Table 3. Summary of binary logistic regression analysis for variables predicting the presence of Pholis gunnellus (Rock Gunnel) in trawl catches from the datasets of 3 trawl surveys: the Northeast Fisheries Science Center’s Spring/Fall Bottom Trawl Survey (NEFSC), the Massachusetts Inshore Trawl Survey (MISTS), and the Long Island Sound Trawl Survey (LISTS) (model chi-square = 835.032, df = 11, P < 0.001, n = 35,356). SE = standard error, df = degrees of freedom. Predictor variable B SE of B df P eB Bottom temperature 0.03 0.02 1 0.052 1.03 Longitude 0.42 0.04 1 <0.001 1.53 Depth -0.03 0.01 1 <0.001 0.97 Season 4 <0.001 Survey 3 <0.001 Constant 22.32 704.56 1 0.975 4.93*109 2013 J.T. Shorty and D.P. Gannon 165 results suggest that they minimize exposure to these conditions. Beneath the protection of cobbles and algae, Rock Gunnels can exploit trapped moisture and may be sheltered from foraging terrestrial predators, such as Neovison vison (Schreber) (American Mink). The interaction of the rocky cobbles and algal/plant cover variables retained in the regression model provides a better metric for total protective cover than either variable alone. Based on the CART model, cobbles are valued over organic cover and consequently may provide greater predator protection. The Rock Gunnel also appears to make a trade-off for the fluctuating water temperature, salinity, and dissolved oxygen content of isolated pools. As tidepools and protective cover are not mutually exclusive, it is likely that the ideal habitat for tidally stranded Rock Gunnels includes both. However, protective cover appears to dictate intertidal habitat selection to a large extent; rocky cobbles appear as a significant split in the CART analysis for gunnels encountered between 36 and 84 minutes from low tide, and all gunnel-containing quadrats with tidepools also contained rocky cobbles, ≥50% plant/algal cover, or both. The individual impacts of environmental conditions and predation pressure on the value of protective cover for the Rock Gunnel are unclear. Previous studies have found that the degree of isolation from the sea (i.e., vertical position within the intertidal zone), tidepool morphology (surface area and volume), physiochemical properties of the water (temperature and salinity), and ecological interactions (competition and predation) affect the distributions of intertidal fishes (Gibson 1972, Halpin 2000, Mahon and Mahon 1994, Nakamura 1976, Thomson and Lehner 1976, Zander et al. 1999). These factors appear to act synergistically (Bennett and Griffiths 1984, Macieira and Joyeux 2011). The interpretation that Rock Gunnels remain in the intertidal zone during low tide to facilitate access to abundant prey in this habitat during submersion is also consistent with previous studies on other intertidal species (e.g., Faria and Almada 2006, Rangley and Kramer 1995). From the perspective of the μ/g relationship first proposed by Werner and Gilliam (1984), the Rock Gunnel appears to make a tradeoff between food resources and vulnerability, the costs of which are faced during the low-tide emersion period. During this time, the Rock Gunnel remains inactive and makes habitat choices that minimize the risk of mortality while also curtailing physiological stressors that could impede growth and reproduction. For remaining in the intertidal to be selectively favorable despite the prioritizing of risk abatement at low tide, there must be unobserved benefits for the Rock Gunnel. Emersion may result in a net decrease of predation risk if the Rock Gunnel faces greater predation pressure from fish and diving seabird predators while in the subtidal (Cairns 1992, Collette 2002), though more work is needed to understand the impacts of predators both above and below the waterline. Also, given the lack of evidence for habitat selection based upon food availability during low tide, it is probable that the intertidal zone provides highly profitable food resources during the high-tide period. Primary food sources 166 Northeastern Naturalist Vol. 20, No. 1 of the Rock Gunnel are characteristic of intertidal areas. Across the northern Gulf of Maine, amphipods of the genus Gammarus are dominant species in the rocky intertidal zone (Larsen 2012). In a study of the shallow subtidal in Midcoast Maine, polychaetes, amphipods, and Idotea isopods, together the majority of the Rock Gunnel’s diet in the region (Sawyer 1967), were among the most abundant macroinvertebrates (Ojeda and Dearborn 1989). This study found Idotea only at the shallowest depths surveyed, suggesting that Rock Gunnels may feed in the upper intertidal zone. By remaining within the intertidal zone at low tide, the Rock Gunnel could quickly access foraging areas in the upper intertidal zone when the water level permits. This hypothesis is supported by evidence from the CART analysis, where Rock Gunnels showed some preference for steep-sloped quadrats with rocky cobbles; occupying such areas would facilitate faster movement to shallower waters. In the subtidal zone, the Rock Gunnel showed preference for shallower waters, as demonstrated by the logistic regression. Notably, this model indicated that neither temperature nor salinity dictated the distribution of Rock Gunnels in trawled areas. These factors are probably insufficiently variable to influence the Rock Gunnel’s subtidal habitat choices, especially considering the variation in temperature and salinity it endures in the intertidal. Longitude and depth were significant predictors in the model, as were season and trawl survey program (NOAA/NEFSC fall and winter surveys, Connecticut LISTS survey, and Massachusetts DMF MISTS survey). In addition to nearshore areas, elevated areas of the sea floor such as Georges Bank and Nantucket Shoals also had higher densities of Rock Gunnel than did the deeper waters within the Gulf of Maine. Like the intertidal zone, these regions have high productivity associated with nutrient availability. Gammarids are present in these areas, most abundant in waters less than 50 m deep (Avery et al. 1996). The production of amphipods on Georges Bank is as high as in the intertidal zone (Collie 1985). Also in high abundance are polychaetes, mollusks, and small crustaceans (Collie et al. 1997). Together, these taxa provide a rich prey community for the Rock Gunnel similar to that available in the intertidal zone. In the intertidal zone at low tide, the Rock Gunnel appears to place a high value on refuge. However, it may remain in the intertidal during low tide to facilitate access of abundant prey in the upper intertidal zone at high tide. The Rock Gunnel’s habitat selection in the subtidal zone is not correlated to temperature or salinity, but likely driven by biotic factors and the availability of shelter. Further behavioral observation is needed, both to better characterize the Rock Gunnel’s fine-scale, subtidal habitat use to determine if it moves into the upper intertidal to forage at high tide and to understand the differences in predation pressure faced by Rock Gunnels in both intertidal and subtidal habitats. Acknowledgments Amy Johnson was vital in developing methods for field data collection and provided helpful feedback for the manuscript along with John Lichter. Katie Guttenplan, Alex Fahey, Sara Davenport, and Adam Mortimer assisted with the fieldwork. Trawl survey data 2013 J.T. Shorty and D.P. Gannon 167 were provided by NOAA’s Northeast Fisheries Science Center, the Massachusetts Division of Marine Fisheries, and the Connecticut Marine Fisheries Division. This project was funded by a Bowdoin Scientific Station Summer Fellowship and the Henry L. and Grace Doherty Coastal Studies Research Fellowship. Literature Cited Avery, D.E., J. Green, and E.G. Durbin. 1996. The distribution and abundance of pelagic gammarid amphipods on Georges Bank and Nantucket Shoals. Deep Sea Research II 43(7–8):1521–1532. Bennett, B.A., and C.L. Griffiths. 1984. Factors affecting distribution, abundance, diversity of rock-pool fishes on the Cape Peninsula, South Africa. South African Journal of Zoology19:97–104. Birks, J.D.S., and N. Dunstone. 1985. Sex-related differences in the diet of the mink Mustela vison. Holarctic Ecology 8:245–254. Blackwell, B.F., W.B. Krohn, and R.B. Allen. 1995. Foods of nestling Double-crested Cormorants in Penobscot Bay, Maine: Temporal and spatial comparisons. Colonial Waterbirds 18:199–208. Bourg, N.A., W.J. McShea, and D.E. Gill. 2005. Putting a CART before the search: Successful habitat prediction for a rare forest herb. Ecology 86(1 0):2793–2804. Burgess, T.J. 1978. The comparative ecology of two sympatric polychromatic populations of Xererpes fucorum Jordan and Gilbert (Pisces: Pholididae) from the rocky intertidal zone of central California. Journal of Experimental Marine Biology and Ecology 35:43–58. Cairns, D.K. 1992. Diving behavior of Black Guillemots in northeastern Hudson Bay. Colonial Waterbirds 15(2):245–248. Cheetham, C., and J.M. Fives. 1990. The biology and parasites of the Butterfish, Pholis gunnellus (Linnaeus, 1758), in the Galway Bay area. Proceedings of the Royal Irish Academy 90B:127–149. Chenoweth, S.B. 1973. Fish larvae of the estuaries and coast of central Maine. Fisheries Bulletin 84:200–204. Collette, B.B. 1985. Resilience of the fish assemblage in New England tidepools. Fisheries Bulletin 84:200–204. Collette, B.B. 2002. Gunnels: Family Pholidae. Pp. 481–483, In B.B. Collette and G. Klein-MacPhee (Eds.). Bigelow and Schroeder’s Fishes of the Gulf of Maine. Third edition. Smithsonian Institution Press, Washington, DC. 748 pp. Collie, J.S. 1985. Life history and production of three amphipod species on Georges Bank. Marine Ecology Progress Series 22:229–238. Collie, J.S., G.A. Escanero, and P.C. Valentine. 1997. Effects of bottom fishing on the benthic megafauna of Georges Bank. Marine Ecology Progress Series 155:159–172. Cote, D., H.M.J. Stewart, R.S. Gregory, J. Gosse, J.J. Reynolds, G.B. Stenson, and E.H. Miller. 2008. Prey selection by marine-coastal River Otters (Lontra Canadensis) in Newfoundland, Canada. Journal of Mammology 89(4):1001–1011. Connecticut Department of Energy and Environmental Protection (CTDEP). 2012. Long Island Trawl Survey. Available online at http://www.ct.gov/dep/cwp/view. asp?A=2696&Q=322660. Accessed 24 April 2012. De’ath, G., and K.E. Fabricius. 2000. Classification and regression trees: A powerful yet simple technique for ecological data analysis. Ecology 81(1 1):3178–3192. 168 Northeastern Naturalist Vol. 20, No. 1 Eby, L.A., and L.B. Crowder. 2002. Hypoxia-based habitat compression in the Neuse River Estuary: Context-dependent shifts in behavioral avoidance thresholds. Canadian Journal of Fisheries and Aquatic Sciences 59(6):952–965. Ekebom, J., P. Laihonen, and T. Suominen. 2003. A GIS-based step-wise procedure for assessing physical exposure in fragmented archipelagos. Estuarine, Coastal, and Shelf Science 57:887–898. Evans, D.H., J.B. Claiborne, and G.A. Kormanik. 1999. Osmoregulation, acid-base regulation, and nitrogen excretion. Pp. 79–96, In M.H. Horn, K.L.M. Martin, and M.A. Chotkowski (Eds.). Intertidal Fishes: Life in Two Worlds. Academic Press, San Diego, CA. 399 pp. Faria, C., and V.C. Almada. 2006. Patterns of spatial distribution and behaviour of fish on a rocky intertidal platform at high tide. Marine Ecology Progress Series 316:155–164. Friedlaender, A.S., P.N. Halpin, S.S. Qian, G.L. Lawson, P.H. Wiebe, D. Thiele, and A.J. Read. 2006. Whale distribution in relation to prey abundance and oceanographic processes in shelf waters of the Western Antarctic Peninsula. Marine Ecology Progress Series 317:297–310. Gibson, R.N. 1972. The vertical distribution and feeding relationships of intertidal fish on the Atlantic coast of France. Journal of Animal Ecology 41:189–207. Gudger, E.W. 1927. The nest and nesting habits of the Butterfish or Gunnel, Pholis gunnellus. Natural History 27:65–71. Halpin, P.M. 2000. Habitat use by an intertidal salt-marsh fish: Trade-offs between predation and growth. Marine Ecology Progress Series 198:203–214. Hawksley, O. 1957. Ecology of a breeding population of arctic terns. Bird-Banding 28(2):57–92. Hickerson, M.J., and C.W. Cunningham. 2006. Nearshore fish (Pholis gunnellus) persists across the North Atlantic through multiple glacial episodes. Molecular Ecology 15:4095–4107. Horn, M.H., and K.C. Riegle. 1981. Evaporative water loss and intertidal vertical distribution in relation to body size and morphology of stichaeoid fishes from California. Journal of Experimental Marine Biology and Ecology 50:273–288. Hothorn, T., K. Hornik, and A. Zeileis. 2005. Unbiased recursive partitioning: A conditional inference framework. Journal of Computational and Graphical Statistics 2006:651–674. Hothorn, T., K. Hornik, and A. Zeileis. 2006. Party: A laboratory for recursive partytitioning. Available online at cran.stat.auckland.ac.nz/web/packages/party/vignettes/ party.pdf. Accessed 24 April 2012. Hurvich, C.M., and C. Tsai. 1989. Regression and time-series model selection in small samples. Biometrika 76(2):297–307. Johnson, D.H. 1980. The comparison of usage and availability measurements for evaluating resource preferences. Ecology 61(1):65–71. Koop, J.H., and R.N. Gibson. 1991. Distribution and movements of intertidal Butterfish, Pholis gunnellus. Journal of the Marine Biological Association of the United Kingdom 71:127–136. Kormanik, G.A., and D.H. Evans. 1988. Nitrogenous waste excretion in the intertidal Rock Gunnel (Pholis gunnellus L.): The effects of emersion. Bulletin of the Mount Desert Island Biological Laboratory 27:33–35. Laming, P.R. 1983. Ventilatory rate in the Butterfish (Pholis gunnellus) as a consequence of temperature and previous emersion. Comparative Biochemistry and Physiology 76A:71–73. 2013 J.T. Shorty and D.P. Gannon 169 Larsen, P.F. 2012. The macroinvertebrate fauna of Rockweed (Ascophyllum nodosum)- dominated low-energy rocky shores of the Northern Gulf of Maine. Journal of Coastal Research 28(1):36–42. Macieira, R.M., and J.C. Joyeux. 2011. Distribution patterns of tidepool fishes on a tropical flat reef. Fishery Bulletin 109:305–315. Mahon, R., and S.D. Mahon. 1994. Structure and resilience of a tidepool fish assemblage at Barbados. Environmental Biology of Fishes 41:171–190. Martin, K.L.M. 1995. Time and tide wait for no fish: Intertidal fishes out of water. Environmental Biology of Fishes 44:165–181. Massachusetts Division of Marine Fisheries (MDMF). 2012. Resource assessment surveys project. Available online at http://www.mass.gov/dfwele/dmf/programsandprojects/ resource.htm#resource. Accessed 24 April 2012. Mecklenberg, C.W. 2003. Family Pholidae Gill 1893: Gunnels. California Academy of Sciences Annotated Checklists of Fishes 9. 11 pp. Moring, J.R. 1990. Seasonal absence of fishes in tidepools of a boreal environment (Maine, USA). Hydrobiologia 194:163–168. Moring, J.R. 1993. Rock Gunnels in the intertidal waters of Maine. Maine Naturalist 1:17–26. Nakamura, R. 1976. Experimental assessment of factors influencing microhabitat selection by the two tidepool fishes, Oligocottus maculosus and O. snyderi. Marine Biology 37:97–104. Northeast Fisheries Science Center (NEFSC). 2012. Research vessel information. Available online at http://www.nefsc.noaa.gov/femad/ecosurvey/mainpage/ships.html. Accessed 24 April 2012. Ojeda, F.P., and J.H. Dearborn. 1989. Community structure of macroinvertebrates inhabiting the rocky subtidal zone in the Gulf of Maine: Seasonal and bathymetric distribution. Marine Ecology Progress Series 57:147–161. Pearcy, W.G., and S.W. Richards. 1962. Distribution and ecology of fishes of the Mystic River Estuary, Connecticut. Ecology 43(2):248–259. Qasim, S.J. 1957. The biology of Centronotus gunnelus (L.) (Teleostei). Journal of Animal Ecology 26:389–401. Rail, J.F., and G. Chapdelaine. 1998. Food of Double-crested Cormorants, Phalacrocorax auritus, in the gulf and estuary of the St. Lawrence River, Quebec, Canada. Canadian Journal of Zoology 76:635–643. Rangeley, R.W., and D.L. Kramer. 1995. Use of rocky intertidal habitats by juvenile Pollock, Pollachius virens. Marine Ecology Progress Series 126:9–17. Rejwan, C., N.C. Collins, L.J. Brunner, B.J. Shuter, and M.S. Ridgway. 1999. Tree regression analysis on the nesting habitat of Smallmouth Bass. Ec ology 80:341–348. Sawyer, P.J. 1967. Intertidal life history of the Rock Gunnel, Pholis gunnellus, in the western Atlantic. Copeia 1967:55–61. Schmidt, A.L., M. Coll, T.N. Romanuk, and H.K. Lotze. 2011. Ecosystem structure and services in Eelgrass, Zostera marina, and Rockweed, Ascophyllum nodosum, habitats. Marine Ecology Progress Series 437:51–68. Schroeder, W.C. 1933. Unique records of the Brier Skate and Rock Eel from New England. Bulletin of the Boston Society of Natural History 66:5–6. Stroud, R.H. 1939. Survey of the food of four fishes of the Bay of Fundy. Fourth Annual Report of the Bowdoin College Scientific Station, Bowdoin College, Brunswick, Maine 6:19–24. 170 Northeastern Naturalist Vol. 20, No. 1 Thomson, D.A., and C.E. Lehner. 1976. Resilience of a rocky intertidal fish community in a physically unstable environment. Journal of Experimental Marine Biology and Ecology 22:1–29. Werner, E.E., G.G. Mittelbach, D.J. Hall, and J.F. Gilliam. 1983a. Experimental tests of optimal habitat use in fish: The role of relative habitat profitability. Ecology 64(6):1525–1539. Werner, E.E., J.F. Gilliam, D.J. Hall, and G.G. Mittelbach. 1983b. An experimental test of the effects of predation risk on habitat use in fish. Ecology 64(6):15 40–1548. Werner, E.E., and J.F. Gilliam. 1984. The ontogenetic niche and species interactions in size-structured populations. Annual Review of Ecology and Systematics 15:393–425. Winkler, P. 1988. Environmentally induced color polymorphism in the Penpoint Gunnel, Apodichthys flavidus (Pisces: Pholididae). Copeia 1988(1):240–242. Winn, H.E. 1950. The Black Guillemots of Kent Island, Bay of Fundy. The Auk 67(4):477–485. Zander, C.D., J. Nieder, and K. Martin. 1999. Vertical distribution patterns. Pp. 26–53, In M.H. Horn, K.L.M. Martin, and M.A. Chotkowski (Eds.). Intertidal Fishes: Life in Two Worlds. Academic Press, San Diego, CA. 399 pp.