Links between rain, salinity, and predation in a rocky subtidal community.
Biotic interactions are spatially and temporally heterogeneous. This simple fact has posed a considerable challenge for the development of predictive models of population growth (May 1981, Caswell 1989), community structure (Paine and Levin 1981, Wiens 1989), and ecosystem function (Holling 1992). The recognition that biotic interactions change along environmental gradients was a major advance toward making community ecology a more predictive science, because it provided a framework to define the conditions where biotic interactions might have a large influence on the distribution and abundance of organisms (Connell 1961, Paine 1966, Dayton 1971). Indeed, gradients of environmental stress are a key component of models of community regulation predicting the relative importance of biotic and abiotic factors (Connell 1975, Menge and Sutherland 1976, 1987, Bertness and Callaway 1994). Recent concern over an increasing pace of anthropogenic environmental change has increased the need for a more quantitative understanding of how fluctuations in environmental factors affect populations and communities in general (Karieva et al. 1993, Barry et al. 1995), and has renewed interest in the ecological effects of environmental stress (Dunson and Travis 1991, Ives 1995).
Predation is a key biotic interaction shaping communities that is reduced by harsh environmental conditions in lakes (Kitchell and Kitchell 1980) and in rocky intertidal habitats (Connell 1975) by restricting predator foraging time (Menge 1978, Garrity and Levings 1981), foraging efficiency (Menge 1983), or by causing predator mortality (Denny et al. 1985). There is a surprising lack of quantitative information on the levels of abiotic factors associated with reduced predation in spite of the long-standing prediction that harsh physical factors mediate predation (Connell 1975, Menge and Sutherland 1976).
Subtidal communities in fjords provide an excellent opportunity to examine the effects of environmental stress on ecological interactions because freshwater runoff creates a stratified water column with a sharply defined low-salinity layer overlying a denser saline layer (Officer 1983). The underwater walls of fjords can support dense assemblages of attached marine invertebrates (Tunnicliffe 1981) and mobile predators. In some fjords in New Zealand, a shallow zone of mussels (primarily Mytilus edulis galloprovincialis) spans the depth range of the LSL above an unusually diverse epifaunal community of sponges, bryozoans, corals, brachiopods, and ascidians (Grange et al. 1981, Smith and Witman 1999). Since low salinity is physiologically stressful to most marine species, we hypothesized that the LSL acts as a dynamic barrier to predation by regulating predator abundance. Specifically, we postulated that marine predators were kept below the LSL by their intolerance to low salinity and thus were prevented from feeding on species in the shallow zone. The depth of the predation refuge may be determined by rainfall because the amount of freshwater runoff is a primary determinant of the LSL depth in fjords (Stanton and Pickard 1981, Farmer and Freeland 1983). Predation refuges are significant for stabilizing predator-prey interactions (Huffaker 1958, Murdoch and Oaten 1975, Crowley 1981), increasing species diversity (Woodin 1978, Witman 1985), and for serving as nursery habitats (Kneib 1987).
MATERIALS AND METHODS
We measured two components of predation, the abundance of predatory invertebrates and predation intensity, in the LSL and at depths below it on vertical rock walls in Doubtful Sound, New Zealand. Mussel recruitment and the distribution and abundance of sessile invertebrates were measured within the same time period as the predation studies.
Study sites. - The two study sites were located in the mid-fjord region on the northern side of Doubtful Sound, east and west of the Tricky Cove experimental site established by Grange and his co-workers (Grange et al. 1981: [ILLUSTRATION FOR FIGURE 1 OMITTED]). Although referred to as East and West Tricky Cove, neither of the sites are embayments, but are large sections of near-vertical rock faces that drop off steeply to [greater than]200 m depth. The two areas were chosen as site replicates because they have similar topography and face the same southwesterly direction. East Tricky Cove (abbreviated ET; 45 [degrees] 20.92[minutes] S, 167 [degrees] 02.91[minutes] E) is [approximately] 250 m horizontal distance from West Tricky Cove (abbreviated WT; 45 [degrees] 20.86[minutes] S, 167 [degrees] 02.79[minutes] E). Fetch for wind-generated waves is limited in the narrow fjord and so both sites are much more sheltered from wave action than the coast at the mouth of the fjord. Thermistors (Hobo Temps Onset Computer, Pocasset, Massachusetts, with 0.1 [degrees] C resolution) attached to the wall at WT recorded temperature every 3.6 h for 1 yr. The annual temperature range was 9.0 [degrees] -17.5 [degrees] C in the LSL (3 m depth), 10.5 [degrees] -17.5 [degrees] C at 10 m and 11.2 [degrees] 17.4 [degrees] C at 18 m depth from 1 April 1993 to 31 March 1994 (J. Witman, unpublished data). These temperature ranges are consistent with long-term averages reported by Grange et al. (1991). It seems unlikely that the shallow distribution of predators was affected by temperature since the annual temperature range of the LSL was only 1.5 [degrees] C cooler than at 10 m.
Mytilus edulis galloprovincialis forms aggregations up to several hundred square meters in area at the study sites and at other subtidal locations in Doubtful Sound (Grange et al 1981; F. Smith and K. Grange, unpublished data). We have observed M. e. galloprovincialis overgrowing serpulid polychaetes, encrusting bryozoans, hydroids, and coralline algae. These observations and the commonly observed ability of mytilid mussels to overgrow other space occupiers (Seed and Suchanek 1992) suggests it is a likely competitive dominant in this shallow subtidal community.
Mussel zonation and community structure. - Random quadrat photography was used to quantify the distribution and abundance of mussels, M. e. galloprovincialis, and major taxonomic groups of sessile epifauna and macroalgae along depth gradients at ET and WT sites during January-February 1993. Eighteen 0.25-[m.sup.2] random quadrats were photographed along transects at each of five depths (1, 3, 6, 10, 20 m) using a Nikonos V camera equipped with a 15-mm lens and two strobes (Witman 1985). Percent cover of mussels, other epifaunal invertebrates, and macroalgae were determined by projecting the 35-mm color transparencies onto a backlit slide projector overlaid with 200 random dots on a clear plastic sheet. The number of dots falling on each taxon of interest was then counted, and percentage cover was calculated as the number of dots occupied by each taxon divided by 200 (as in Witman 1987).
Hydrography and rainfall. - The water column was profiled with a Conductivity Temperature Depth recorder (CTD, Datasonde model 3, Hydrolab Corporation, Austin, Texas) during cruises in February, April, and November 1993 to determine the depth of the low-salinity layer. Additional observations of the depth of the LSL and predation were made by SCUBA diving at both sites over 3-5 d periods in January, March, and May 1993 and in April 1994. The boundary between the LSL and the fully saline water mass below was visible to divers as a density and coloration discontinuity (Davies-Colley 1992). Data on daily rainfall accumulation in Doubtful Sound from October 1992 to April 1994 were obtained from a rain gauge (12.7 cm diameter opening) located 3 m aboveground and 10 m above sea level at the Deep Cove Outdoor Education Center. Deep Cove is at the head of Doubtful Sound, 16 km landward of the Tricky Cove study sites [ILLUSTRATION FOR FIGURE 1 OMITTED].
We assumed a simplified two-layered model of fjord circulation in studying the link between rainfall and the thickness of the LSL (Farmer and Freeland 1983). In an unrestricted region of a fjord, such as the midfjord study area in Doubtful Sound, the model assumes that the height of the surface is close to sea level and freshwater discharge at the head of the fjord drives the LSL seaward. Increased rain could increase the thickness of the LSL and depress the depth of the bottom boundary. The terms "thickness of the LSL" and the "depth of the lower margin" or "bottom boundary of the LSL" are used interchangeably in the text. The location of the lower margin of the LSL in relation to the subtidal communities on the walls of the fjord also fluctuates with the tides [ILLUSTRATION FOR FIGURE 2 OMITTED].
Predator abundance. - To compare the abundance of predators in the rock wall communities at depths in and below the LSL, replicate band transects were conducted by stretching a transect tape parallel to the wall so that it followed the 2.5, 5, 10, 20, and 30 m isobaths. A 1.0 m wide swath along the transect was carefully searched for predatory invertebrates using a measured rod as a reference. Transects were 30 m long in February and 25 m long in April and November (1993). The number of sea stars, urchins, gastropods, nudibranchs, octopi, crabs, and spiny lobsters were counted in four replicate band transects per depth, except for February when three transects were surveyed per depth. Therefore, our estimates of predator densities are based on a total area of 450-500 [m.sup.2] of habitat per sampling period. All predator censuses were conducted during the day.
Predation intensity. - Predation intensity refers to rates of consumption of experimental prey (Menge 1983). To document rates of predation on mussels, predation intensity experiments were performed at four of the five depths surveyed for predator abundance. Since prey vulnerability is inversely related to body size in many prey species, including mussels (Paine 1976), the size range of Mytilus edulis galloprovincialis (formerly Mytilus edulis aoteanus) used in the experiments was restricted to 3.5-5.0 cm shell length. This size range included the mean size of M. e. galloprovincialis consumed naturally by the sea stars Coscinasterias calamaria and Patiriella regularis at these sites (4.28 [+ or -] 0.87 cm shell length, mean [+ or -] 1 SD, n = 45, haphazard collections in January-April 1993).
Mussels were collected from each site and transplanted to different depths on the rock walls using velcro, as in Witman (1985). The procedure consisted of gluing a small pad of velcro "hooks" to the middorsal or midventral valve of the mussel. Reciprocal pads were attached with Koppers Splash Zone Compound to patches on the rock walls ([approximately]15 cm diameter) that were cleared to facilitate adhesion of the epoxy. The epoxy blobs with embedded velcro were [approximately] 10 cm long. To eliminate potential experimental artifacts of attracting predators to the disturbance of clearing a patch, the experiment was begun at least 24 h after the velcro pads were set up. Mussels attached to the walls in this fashion were oriented with the posterior end of the shell facing away from the substratum at a slight angle to simulate the orientation of natural mussels. The experimental unit was one mussel on a velcro pad, with individual mussels randomly spaced at least 50 cm horizontal distance apart along a 6-8 m horizontal span of a given depth contour on the walls. The complete experimental design in April 1993 consisted of 10 replicate mussels at each of 2.5, 5.0, 10, and 18 m depths at East and West Tricky Cove (n = 40 mussels per site).
No information on rates of predation in the fjord subtidal communities was available at the beginning of the investigation to aid in the design of the predation intensity experiments. We chose to spread out individual mussels over a large area at each depth, rather than cluster a larger number of replicate mussels at one location to avoid spatial pseudoreplication (Hurlbert 1984), and to maximize the probability that predators would encounter the prey. A considerable amount of diving was required to monitor predation on 40 mussels along the 2.5-18.0 m depth gradient at each site over these large areas of the walls for several days per experiment. These logistical considerations prevented us from monitoring predation on larger numbers of replicate mussels.
The experiments at WT began at 0930 on 20 April 1993 and ended at the same time on 24 April. The ET experiments overlapped with the last 2 d of the WT experiments. We repeated the experiments at both sites in November 1993. The experiment at West Tricky began at 1000 on 22 November and ended 4 d later at 1030. The East Tricky experiment began at 1630 on 22 November and ended at the same time as WT on 26 November. During the November trials, 1 and 3 of the 10 velcro attachments failed at 2.5 m depth at WT and ET respectively, so 9 and 7 mussels per depth were available to predators at 2.5 m depth. It was not practical to use photographic or video techniques to monitor predation events since the experimental mussels were spread out over large ([approximately equal to]8 x 16 m) areas of rock walls. Consequently, predation on mussels was monitored by divers over the 2-4 d experiments. Many predation events were directly observed, since many of the mussels were consumed by sea stars that remained on the prey for 5-30 h. The urchin Evechinus chloroticus left characteristic scraping marks from its Aristotle's lantern and a jagged edge on the posterior end of the mussel shell, which was eventually consumed by fragmenting the shell. This trademark of urchin predation was verified by examining and photographing mussels that were being consumed by urchins (J. Witman, unpublished photographs). Spiny lobsters, Jasus edwardsii, also fragmented the shells, but in most cases carried them to their home crevices. At West Tricky Cove, horizontal ledges protruding out from the wall at 6 and 28 m depth where experimental mussels would have accumulated if they fell off the wall were routinely searched for dislodged mussels. The velcro pad on the experimental mussels made them easy to distinguish from other mussel shells. At East Tricky Cove, a ledge at 6-10 m depth and a gradual, sloping shelf at 25 m were searched for dislodged experimental mussels. The disappearance of mussels was attributed to an unobserved predation event if dislodged experimental mussels were not found on these ledges.
Sea star movement. - To examine the potential range of movement of two mussel predators in relation to the LSL, the dorsal surfaces of the sea stars Patiriella regularis (n = 20 per location) and Coscinasterias calamaria (n = 5 per location) were marked with a soft china marker and released at two locations at 1.3 m depth at WT on 2 April 1994. The distance that the marked sea stars moved was determined when they were recovered 6 h later.
Mussel recruitment. - The hypothesis that the lower limit of the mussel zone was determined by differential mussel recruitment along the depth gradient was tested by a recruitment experiment conducted on the same temporal and spatial scales as the rest of the investigation. Roughened slate tiles ([approximately equal to]25 x 25 cm) attached to the rock walls at 3, 6, 10, and 20 m at both study sites were used as settlement collectors. The effects of predatory urchins, sea stars, gastropods, and fish on mussel recruitment was tested by enclosing tiles in plastic mesh predator exclusion cages. Two treatments and a control were included in the experiments; tiles in exclusion cages (-invertebrate predators, -fish), tiles in topless cages (-invertebrate predators, +fish) and noncaged control tiles (+invertebrate predators, +fish). Mesh openings were 15 x 15 mm. The sides of the topless cages were 10 cm high. Smith and Witman (1999) found no significant differences in fluid fluxes over tiles in caged, noncaged, and topless cages identical to those in this study at shallow subtidal sites in Doubtful Sound, suggesting that there were no obvious hydrodynamic artifacts of cages and topless cages that might affect mussel recruitment to the tiles. The topless cages were effective in excluding urchins, sea stars, and gastropods because they approached the tiles from the sides. Wrasses and blennies were the predominant types of fish observed over noncaged tiles and tiles in topless cages (J. Witman and K. Grange, personal observations).
One of each open (noncaged), predator exclusion, and topless-caged tiles were randomly assigned to locations along a 1.5 m length of PVC pipe. Six of these recruitment tile arrays were then randomly assigned to locations on the rock walls along each of the 3, 6, 10, and 20 m depth contours. The recruitment tile arrays were then bolted to the rock walls so that the tiles did not protrude [greater than]2.5 cm out from the natural invertebrate assemblages on the wall. The complete experimental design consisted of 72 slate tiles at each of the two sites (WT and ET), with 18 tiles at each of the four depths that were allocated to the three control/treatment categories, yielding six replicates for each treatment and control located at a given depth.
The recruitment experiment was begun during the 1st wk of May 1993 and the tiles were photographed 3 wk later, and thereafter in August and November 1993 and in April 1994. Mussel recruitment was measured over 3-4 mo intervals by photographing the central 238 [cm.sup.2] area of each tile in situ. The byssus threads of the mussels distinguished newly recruited mussels from other small, ovoid objects in the projected 35-mm transparencies.
Data analysis. - Mean abundances of invertebrate predators at the five depths were compared by the Kruskal-Wallis test in lieu of a single factorial ANOVA because in many cases the variances of the depth means were significantly different by Cochran's test (Winer et al. 1991) after the data were transformed, thus violating the homogeneity of variances assumption of ANOVA (Sokal and Rohlf 1981, Underwood 1981).
Failure-time analysis was used to analyze the results of the predation intensity experiments because the analysis of survivorship curves by failure-time analysis compares rates of survival, utilizing more information from experiments than survival at the endpoints (Pyke and Thompson 1986, Fox 1993). In this application, failure-time represents time to a prey mortality event. The survivorship data were "censored" in the sense that the actual time that the mussel prey were consumed (removed) was not precisely known in all cases, and because the experiments ended before all the prey died (Pyke and Thompson 1986). In these situations, we only knew that the actual time that the mussel prey were removed by predators was greater than the last time that they were observed alive or being consumed. Statistical analyses of censored survivorship data may be biased if censoring is not taken into account (Muenchow 1986, Pyke and Thompson 1986, Fox 1993). Comparisons of the survivorship of mussel prey were conducted using a product-limit estimate of the survivorship function, the Kaplan-Meirer method, in JMP statistical software (SAS 1995). This nonparametric routine tests the homogeneity between groups (depths) by reporting a chi-square statistic from a Wilcoxon and a log rank test (SAS 1995). Statistical tests of survivorship assume that the experimental units are independent (Fox 1993), which in our application was satisfied by the independent location of individual mussels in the predation intensity experiments. The log rank test was used because it is unbiased by early deaths (Pyke and Thompson 1986). Comparisons were made separately for each month and site (e.g., April, at ET) in two stages with the bulk comparison of all four depths (2.5, 5.0, 10, 18 m) followed by separate, planned pairwise comparisons of 2.5 m vs. each of the three other depths. The Bonferroni procedure was used in the multiple comparisons to increase the level of significance to an [Alpha] of 0.016 (up from [Alpha] of 0.05) to avoid inflating the Type I error rate (Winer et al. 1991), although adjusted alphas may be overly conservative in some cases (Rice 1990).
To estimate predation intensity as a function of the depth of the LSL, we developed a regression between rainfall (x) and the depth of the salinity level influencing predation (y). An upper limit of 30 mg/g salinity was used in the regression model of rainfall on salinity because 30 mg/g salinity defined the upper boundary of dense bands of sea stars and urchins, and salinities [less than or equal to]30 mg/g caused sea stars to move away (J. Witman and K. Grange, personal observations), or were associated with no predation during experiments at 2.5 m depth. Data for the regression analysis were obtained from 19 d during the 1.5-yr study period when both CTD casts from the midfjord region of Doubtful Sound and rainfall data were available. Y variates were square-root transformed to satisfy the equality of variances assumption of least squares linear regression (Sokal and Rohlf 1981). The regression equation was then used to predict the depth of the 30 mg/g boundary from October 1992 to April 1994. A three-point moving average of back-transformed y values was computed in the SERIES routine of Systat 5.21 software (Systat 1992) to yield the average depth of the 30 mg/g boundary. The moving average computation interpolated up to 7 d of missing rain data.
Daily tidal heights were calculated for the 1 October 1992-30 April 1994 study period using equations provided in the New Zealand Hydrographic Survey Tide Tables to correct for Deep Cove from the nearest reference station at Westport. Tidal heights calculated for Deep Cove ranged from 0.9 m during neap tides to 1.7 m during spring tides, with an average height of 1.3 [+ or -] 0.16 m (mean [+ or -] I SD). Daily tidal changes in the average depth of the 30 mg/g boundary relative to the study locations on the rock wall were calculated by adding the tidal amplitude (half the tidal height) to the depth of the 30 mg/g to represent the deepest excursion (lower boundary) of the layer and similarly, by subtracting the tidal amplitude from the depth of the 30 mg/g boundary to represent the shallowest position of the layer (upper boundary, [ILLUSTRATION FOR FIGURE 2 OMITTED]).
Community structure and mussel zonation. - Mytilus edulis galloprovincialis was the most abundant epifaunal invertebrate at depths influenced by the LSL [ILLUSTRATION FOR FIGURES 3 and 4 OMITTED]. Mussel abundance increased with depth from 1.0 to 3.0 m, where they attained maxima of 52.0 and 80.0% cover at ET and WT, respectively. Mytilus abundance declined sharply with depth below the typical lower boundary of the LSL [ILLUSTRATION FOR FIGURE 5 OMITTED]. Barnacles (predominately Chamaesipho sp.) dominated the shallowest subtidal community (1 m depth) at both sites with lesser abundances of encrusting coralline algae and ephemeral green algae (Ulva sp., [ILLUSTRATION FOR FIGURE 4 OMITTED]). Most of the substrate at 1.0 m depth was bare rock (53.9% ET, 67.2% WT, not graphed) which declined sharply in the mussel zone (6.0% cover of bare rock at ET 3 m, 0% at WT 3 m). At both sites, the predominance of coralline algae at 6 m depth gave the ecotone a relatively more "barren" appearance than all other habitats at or below the mussel zone [ILLUSTRATION FOR FIGURE 4 OMITTED]. The percent cover of all other epifaunal invertebrates besides mussels, barnacles, and hydroids increased with depth below the LSL [ILLUSTRATION FOR FIGURES 4 and 5 OMITTED]. The anemone Diadumne sp. was prevalent at 6 m depth. There was no unoccupied space in the diverse deeper communities (Smith and Witman 1999), where a layer of erect bryozoans covered [greater than] 18% of the substrate at 10 m and below at both sites. Solitary and colonial ascidians collectively accounted for [greater than]25% cover at 10-18 m depth at both sites [ILLUSTRATION FOR FIGURE 4 OMITTED]. Calcarid sponges and demosponges were most abundant at 18 m where they covered [greater than]20% of rock walls [ILLUSTRATION FOR FIGURE 4 OMITTED].
Hydrography. - The water column was sharply stratified during the sampling periods [ILLUSTRATION FOR FIGURE 5A OMITTED]. Salinities at the surface were 5.0 to 7.0 mg/g in February and 5.0 mg/g in April. Surface salinity increased to 12.5 mg/g in November. The LSL was deepest in February (4.0 m at 30 mg/g salinity) and shallowest in November when 30 mg/g salinity occurred at 1.5 m depth. Average daily salinities at the shallowest predation site (2.5 m depth) were correspondingly lower in February than in November [ILLUSTRATION FOR FIGURE 5B OMITTED]. There was an inverse relation between average daily rainfall and salinity at 2.5 m depth during the sampling periods, with significantly lower rainfall in November than February ([ILLUSTRATION FOR FIGURE 5B OMITTED]; Mann-Whitney U = 31.0, P = 0.037, df = 11). The upper 1 m of the LSL was not as well mixed in April as it was during the other sampling periods, based on the abrupt shape of the salinity profile (Pickard 1961).
Predator distributions. - The greatest concentration of predators typically occurred just below the LSL at 5.0 m depth ([ILLUSTRATION FOR FIGURE 6 OMITTED], Table 1, Appendix). For example, mean predator densities above and below 5.0 m differed by an order of magnitude at WT in February and April. Kruskal-Wallis and multiple comparisons tests revealed that the average number of predators per depth was significantly greater at 5 m than at 20 and 30 m in February (Table 1). In April, the mean number of predators was significantly greater at 5 m than at 10, 20, and 30 m (Table 1). The predator guild centered around [TABULAR DATA FOR TABLE 1 OMITTED] the 5-m isobath in February and April was primarily composed of sea stars (Patiriella regularis, Coscinasterias calamaria) and the sea urchin Evechinus chloroticus (Table 2, Appendix). The guild was numerically dominated by the bat star, P. regularis. This sea star [TABULAR DATA FOR TABLE 2 OMITTED] was significantly more abundant at 5 m than at 10, 20, and 30 m depths in February and more abundant at 5 m than all other depths in April. (Table 2). Densities of the large sea urchin, E. chloroticus, also peaked at 5 m depth and displayed a significant decrease with depth below 5 m during these two sampling periods (Table 2). Although some Patiriella occurred in the LSL at 2.5 m depth in February and April, it appeared that they were caught by a short-term downwelling of the LSL, since we observed that they were moving deeper and were not feeding (J. Witman and K. Grange, personal observations). Lower densities of gastropods (Maurea punctulata and Turbo granosus), another sea urchin species (Pseudechinus huttoni), and the nudi-branch Atagemma carinata also occurred at 5.0 m (Appendix).
In sharp contrast to the previous sampling periods, predator densities peaked at 2.5 m at both ET and WT sites in November [ILLUSTRATION FOR FIGURE 6B OMITTED]. The upward shift in the dense predator band from 5 m to 2.5 m in November coincided with the shallowing of the LSL from 3.04.0 m to 1.5 m depth. Table 2 shows that the predators concentrated at 2.5 m depth included P. regularis, Coscinasterias calamaria, and Evechinus chloroticus. Patiriella densities were significantly higher at 2.5 m than at three of the four depths (10, 20, 30 m) censused below it at both sites (Table 2). Although this vertical shift in the predator band was striking, Coscinasterias and Evechinus were still abundant at 5.0 m depth in November (Table 2). Spiny lobsters, Jasus edwardsii, were observed in the predator transects only at 10 m depth at East Tricky Cove in November, where they occurred in mean densities of 0.5 individuals/25 [m.sup.2].
Predation intensity. - Predation intensity experiments using individual mussels attached to independent locations on the fjord walls revealed that no mussels were consumed within the LSL (2.5 m depth) in April at either the WT or ET sites [ILLUSTRATION FOR FIGURE 7A, C OMITTED]. Analysis of the survivorship of mussel prey revealed highly significant differences in prey survival among the four depths tested at each site and time that the experiments were conducted (Table 3).
In April, 20-60% of the total number of mussels deployed below the LSL (5-18 m) were consumed within 100 h at the WT site. Lower survival below the LSL was manifested in significantly higher prey survival at 2.5 m than at 5.0 m depth at WT (Table 3). The sea stars Coscinasterias calamaria, Patiriella regularis, and the sea urchin Evechinus chloroticus were responsible for the high predation documented at 5 m at WT [ILLUSTRATION FOR FIGURE 7 OMITTED]. The trend of higher prey survival in the LSL (2.5 m) at East Tricky Cove in April was not significantly different from prey survival at 5.0 m at this site and time by the comparison of survivorship curves (Table 3), despite the fact that no mussels were consumed at 2.5 m and four were eaten at 5.0 m. The P value from the log rank test comparing the 2.5 vs. 5.0 m curves was 0.029, but it was not significant after Bonferroni adjustment. Prey survival at ET was significantly higher at 2.5 m than at 10 m depth, where spiny lobsters, Jasus edwardsii, accounted for 87.5% of the predation (Table 3, [ILLUSTRATION FOR FIGURE 7C OMITTED]). Two instances of Jasus predation were directly observed, and lobster predation was implicated for five other mussels at ET because a total of seven fragmented mussels with velcro still attached to their shells were found and photographed 1 m in front of a crevice at 10 m containing three adult lobsters (J. Witman and K. Grange, unpublished photos). Sea stars preyed on experimental mussels at 5 m depth at ET in April as they did at the WT site.
Forty to sixty percent of the mussels originally set out in the LSL (2.5 m depth) during the April predation experiments were found alive in the same locations at ET and WT 5 mo later, suggesting that the refuge effect held for longer time periods than the 2-4 d duration of the experiments. None of the experimental mussels remaining at depths below the LSL after the April experiments (n = 35, both sites pooled) survived the same 5-too period.
When the predation intensity experiments were repeated in November, some of the mussels transplanted to 2.5 m depth were consumed by the sea stars P. regularis or C. calamaria [ILLUSTRATION FOR FIGURE 7B, D OMITTED]. This contrasts with the April experiments when no mussels were consumed at 2.5 m depth. Comparison of survival curves between depths in November showed prey survival was significantly lower at 5.0 m than at 2.5 m at WT, largely due to high predation by Coscinasterias at 5.0 m. (Table 3, [ILLUSTRATION FOR FIGURE 7B OMITTED]). Consumption of all the mussels at 10 m depth by lobsters at ET resulted in prey survival being significantly lower at 10 m than at 2.5 m in November (Table 3, [ILLUSTRATION FOR FIGURE 7D OMITTED]).
More than one-third, or 61, of the total number of mussels used in all four predation experiments (156) were consumed by predators. The lobster, J. edwardsii, had the greatest per capita effect, consuming a total of 17 mussels, but this predation was restricted to one depth and site (10 m, ET). C. calamaria was ranked second by the total of 14 mussels consumed from the 2.5-5.0 m depth range. P. regularis consumed a total of 9 mussels, while E. chloroticus was observed feeding on 4 mussels. Octopus marorum had the lowest per capita effect in the experiments, where it was observed feeding on only 2 mussels. Losses of 15 of the experimental mussels were attributed to unobserved predation.
Rain and tidal influences on the LSL. - Least squares linear regression indicated that the relationship between the depth of 30 mg/g salinity (y) and rain (x, in millimeters) 1 d prior to water column profiling could be expressed by [-square root of y] = 1.365 + 0.009x (P [less than or equal to] 0.0001, n = 19), with [r.sup.2] = 0.615. Thus, 61.5% of the variation in the depth of the 30 mg/g layer was accounted for by the amount of rain accumulating 1 d before. A regression of the depth of the 30 mg/g salinity (y) vs. rain accumulation the same day that the water column (x) was profiled was not significant, suggesting that a 1-d lag is required for freshwater runoff to depress the LSL.
Fig. 8 shows the frequency and duration at which 30 mg/g salinity occurred at a given depth between 1 October 1992 and 30 April 1994. The average depth of the 30 mg/g boundary predicted from the regression was 2.35 m. The lower margin of the layer represents the average width of the spatial refuge from invertebrate predation. Deep excursions of the LSL to [greater than or equal to]4 m depth occurred three times in 1993 (April, September, December) and once in March 1994. No seasonality in the occurrence of these major rain storms was apparent.
Tidal influences on the location of the predation refuge are shown in Fig. 9. Subtidal habitats shallower than or equal to the depth of the upper tidal excursion of the 30 mg/g boundary (upper boundary in [ILLUSTRATION FOR FIGURE 9 OMITTED]) are always in the predation refuge. The LSL moves downward with the failing tide to bring the depth of the 30 mg/g boundary below the depth predicted from rainfall alone for a few hours per day (lower boundary, [ILLUSTRATION FOR FIGURE 9 OMITTED]). This temporary exposure to low-salinity water would reduce predation by interrupting the feeding process of the stenohaline marine predators and causing them to move away (J. Witman and K. Grange, personal observations). Consequently, we refer to these cases where the 30 mg/g boundary was driven down to a given depth for part of a tidal cycle as partial predation refuges. Due to these tidal influences on the 30 mg/g boundary, the 2.5 m habitat was completely free of invertebrate predators only 10% of the 18-mo period (53 d), yet it was a partial predation refuge 90% of the time (477 d). For a small increase in depth from 2.5 to 3.0 m, the percentage of the time (days) that the zone was a complete refuge from predation decreased to 5%, while the zone was a partial refuge 70% of the time [ILLUSTRATION FOR FIGURE 9 OMITTED]. The 5.0 m depth habitat was completely buffered from predation only half a day (0.01% of the time), and was a partial refuge 2.0% of the time (10.6 d).
Mussel recruitment. - Mytilus recruitment was low, with a total of only 17 mussel spat occurring on the roughened slate tiles throughout the 12-mo experiment. The low numbers of recruits precluded statistical analysis of the data. Nearly all recruitment occurred over the summer sampling period (November 1993, April 1994) and was restricted to the 3-6 m depth range (Table 4). When all treatments were pooled within depths, there was a trend of higher recruitment to 6 m than to 3 m depth. The data also suggest an effect of predator exclusion on mussel recruitment since no recruits were found on the noncaged tiles exposed to predators. Many of these tiles at 6 m had clusters of pentaradial scraping marks on them, which are characteristic of sea urchin grazing (Ayling 1981).
Our results indicate that the abundance of invertebrate predators and predation intensity vary along vertical salinity gradients in subtidal rock wall habitats at two sites in Doubtful Sound, New Zealand. Freshwater runoff maintains a surface low-salinity layer, which we suggest drives predators from shallow depths and reduces predation in the shallow zone. Thus, the areal extent of the spatial refuge from predation seems largely determined by variation in rainfall and other factors influencing freshwater runoff. Striking differences in the composition of sessile invertebrate communities above and below the salinity gradient (halocline have [TABULAR DATA FOR TABLE 3 OMITTED] been previously described from fjords in Sweden (Rosenberg and Moller 1979) and New Zealand (Grange et al. 1981), but only one other study has connected the LSL to an ecological process. Smith and Witman (1999) suggest that the LSL influences the recruitment and diversity of epifaunal communities on landscape spatial scales.
Both rainfall and tides influence the location of the layer relative to the communities and study sites on the subtidal walls of the fjords. Rain is evidently the major determinant of the thickness of the LSL, as a 4.1-m range (1.5-5.6 m) in the average thickness of the LSL was accounted for by rain accumulation a day before the salinity profile. based on calculated tides for the head of the fjord, we estimate that the LSL moved up and down 0.9 to 1.7 m due to daily tidal variation. This indicates that comparatively small vertical movements of predators may occur on a daily basis due to tides, but that the major changes in vertical distribution of predators may be driven by rain on temporal scales of days to months. For example, the 2.5-m difference in the depth of the LSL between April and November exceeded the tidal range.
Successful predation in shallow habitats of this eco-system depends on the interplay between the time required for a predator to consume or remove its prey and the amount of time before predation is interrupted by low-salinity water deflected downward by changes in freshwater runoff, tides, or by other mixing events. Estimates of the time required for P. regularis, C. calamaria, and E. chloroticus to consume single mussels were available from our direct observations of the predation intensity experiments. The large sun star C. calamaria ([greater than]300 mm diameter) usually remained on the prey for 15-20 h, but in two instances it opened M. edulis galloprovincialis in 5 h. Patiriella ([less than]60 mm diameter) fed on individual mussels in aggregations of 3-8 sea stars requiring at least 15 h to consume the viscera. The urchin Evechinus ([approximately]120 mm diameter) removed approximately one-third of the mussel shell in 15-20 h. Mussel shells were fragmented within 5 h in the two instances of lobster predation observed. We assume that rates of consumption by lobsters are generally more rapid than those of the comparatively sluggish echinoderms. As more motile predators, lobsters may be able to avoid the LSL by moving rapidly up the walls to remove mussels, then returning to depths below the influence of the LSL to consume them. In Southern California, Robles et al. (1990) found that lobsters have a large vertical foraging range, extending up into the mid-intertidal zone to feed on mussels. Ten marked sea stars C. calamaria released at 1.3 m depth and 26.3 mg/g salinity moved an average ([+ or -] 1 SD) of 3.91 [+ or -] 1.04 m, or nearly 4 m in a 6-h period. This rate of movement will enable it to track the lower boundary of the LSL and keep pace with the tide, but our estimates indicate that it would not be enough time for a C. calamaria at the lower margin of the LSL to consume a mussel and escape the downward excursion of low-salinity water. Forty marked Patiriella regularis released at 1.3 m depth and 29.7 mg/g salinity moved an average of 1.42 [+ or -] 0.64 m in 6 h, or close to the average tidal amplitude. This suggests that a nonfeeding Patiriella could keep up with an average tide, but a P. regularis feeding close to the lower margin of the LSL clearly would not have enough time (15 h) to consume a mussel and avoid downward excursions of the LSL. A P. regularis at this depth could be stranded in the LSL during a spring tide with a 1.7 m tidal height.
TABLE 4. Densities of Mytilus sp. recruitment to slate tiles at 3 and 6 m depth at West and East Tricky Cove. Recruitment occurred between November 1993 and April 1994. Data represent mean densities (1 SD in parentheses) per 238.0 [cm.sup.2] of six replicates per treatment. There was no mussel recruitment to the tiles at 10 and 18 m depth. Treatment Depth Topless Site (m) Open Caged cage West Tricky Cove 3 0 0 0.16 (0.40) 6 0 0.33(0.51) 0.5 (0.54) East Tricky Cove 3 0 0.33 (0.51) 0 6 0 0.16 (0.40) 1.16 (0.98)
When the LSL moved upward in November there was a dramatic upward shift in the location of the dense band of predatory invertebrates, and 1-4 mussels were eaten in the 2.5 m depth of the former predation refuge in April. Even so, there were still relatively large numbers of Coscinasterias and Evechinus just below at 5.0 m depth at this time. Consequently, predation was still high on the mussels transplanted to 5.0 m (particularly by Coscinasterias, [ILLUSTRATION FOR FIGURE 7 OMITTED]), causing the significant result of lower prey survival at 5.0 vs. 2.5 m (Table 3).
Potential artifacts. - Mussels have been experimentally transplanted to examine patterns of predation in a large number of temperate intertidal habitats (Kitching and Ebling 1967, Seed 1969, Paine 1976, Bertness and Grosholz 1985, Hardwick-Witman 1985, Petraitis 1990, Menge et al. 1994, Robles et al. 1995). Although the mussels transplanted in this study were not strictly tethered, as experimental manipulations of prey organisms they are subject to some of the criticisms of preytethering experiments raised by Peterson and Black (1994). For example, it is possible that rates of predation on restricted prey may not be similar to natural rates of predation on unrestricted prey. However, for the two species of sea stars that we observed feeding on natural beds, the duration of feeding on solitary, transplanted mussels was very similar to the duration of feeding on a mussel in natural aggregations at 5-6 m depth (15-30 h for P. regularis experimental vs. 12-30 h in natural beds, n = 35, 5-30 h for C. calamaria experimental vs. 6-24 h in natural beds, n = 10; Witman and Grange, unpublished data).
Differences in the "ambient" abundance of natural mussel prey along the depth gradient might have biased the predation intensity experiments by reducing predation intensity on the transplanted mussels at depths where the cover of mussels is high. In this scenario, predators might be swamped by prey (Murdoch and Oaten 1975) in the shallow zone (2.5 m), possibly resulting in lowered attack rates on individual mussels and higher prey survival at this depth, thus offering an alternative explanation to a refuge effect created by the LSL. To evaluate this alternate hypothesis, we estimated prey: predator ratios using an estimate of Mytilus percentage cover at the depth where the predation intensity experiments were conducted (2.5 and 5.0 m) and total predator densities at the same depths for the months and sites when both types of data were available (WT in April, ET and WT in November). This yielded a ratio of 65% mussel cover to 77 predators/25 [m.sup.2] at 2.5 m and 30% mussel cover to 207 predators/25 [m.sup.2] at 5.0 m for West Tricky Cove in April. Per capita estimates were 0.84% mussel cover available per predator at 2.5 m vs. 0.14% mussel cover available per predator at 5.0 m (mussel cover from [ILLUSTRATION FOR FIGURE 3 OMITTED]). By this estimate, [approximately]6 times more prey would have been available to each predator (primarily Patiriella regularis) at 2.5 than at 5.0 m at WT in April. A predator-swamping effect thus has potential to explain the low predation observed at 2.5 m. However, in the considerable number of hours spent underwater at these two sites, the P. regularis in the LSL at 2.5 m depth were never observed feeding on mussels (J. D. Witman and K. R. Grange, unpublished observations). Because they were not observed feeding, it is unlikely that the sea stars at 2.5 m depth were satiated and that the low predation at this depth was due to predator swamping effect.
There was apparently less potential for differences in prey:predator ratios to explain the differences in predation intensity between the shallow study areas at either WT or ET sites in November. For example, estimated per capita prey availability at WT was nearly the same between the 2.5 m and 5.0 m depths (0.28% mussel cover per predator at 2.5 m vs. 0.20% cover per predator at 5.0 m), yet there was a significant difference in the survival of mussel prey consumed between the two depths [ILLUSTRATION FOR FIGURE 7 OMITTED]. Similarly, there was a small difference in prey: predator ratios between the two shallowest depths at ET (0.063% natural mussel cover available per predator at 2.5 m vs. 0.075% mussel cover available per predator at 5.0 m), yet two times more prey were eaten at 5 m than at 2.5 m. The necessity of interpolating mussel percent cover values at 2.5 and 5.0 m depth from Fig. 3 renders our prey: predator ratios less precise than if mussel cover was actually measured at these two depths. Due to the large changes in mussel cover over small depth increments [ILLUSTRATION FOR FIGURE 3 OMITTED], it was less desirable to use mussel percent cover values from 3.0 and 6.0 m depth in the prey: predator ratios.
Physical conditions limiting the refuge. - The existence of the predation refuge depends on the salinity stratification of the fjord water column. At least two additional factors could modify our analysis of the depth and stability of the LSL. First, oceanographic or meteorological conditions that reduce stratification could dissipate the refuge effect. Stratification may break down temporarily during periods of upwelling caused by the exchange of deep fjord water with in-flowing coastal shelf water. Turbulence overcoming the stability of the water column may also arise during periods of high winds (Officer 1983) or internal waves (Farmer and Freeland 1983). In other high-latitude regions where rainfall and freshwater discharge is lower in winter than summer (Burrell 1986), the depth of the LSL may show considerable seasonal variation. However, seasonality of rainfall is not apparent in the New Zealand fjords where rain occurs all year round (Pickrill 1987, Grange et al. 1991). In all fjords, the influence of freshwater runoff decreases with horizontal distance from the head of the fjord to the open sea (Stanton and Pickard 1981, Farmer and Freeland 1983) and a distinct LSL is probably absent from many outer regions of fjords. This raises the possibility of a horizontal predation refuge where populations of euryhaline species such as mussels and the barnacles Chamaesipho columna and Eliminus modestus (Batham 1965, Grange et al. 1981) may be exploiting low-predation regimes in the inner and middle regions of the fjord.
Secondly, our regression model predicting the depth of the 30 mg/g interface in the mid-fjord region of Doubtful Sound is simplistic in that it considers rainfall as the only determinant of the depth of the interface. Snowmelt is a form of "stored precipitation" influencing the magnitude of freshwater runoff and halocline depth that is not considered in our model. However, runoff from snowmelt is evidently more important in fjords along the margin of British Columbia and Alaska than in New Zealand (Pickard and Stanton 1980, Burrell 1986, Pickrill 1993). Doubtful Sound does not have glaciated catchments. While freshwater runoff is the major factor influencing both fjord stratification and circulation, any mechanism increasing mixing will increase the depth of the LSL (Yih 1980, Farmer and Freeland 1983). It is likely that a more accurate prediction of the depth of the interface would be obtained by taking additional physical parameters into account, such as density and velocity differences of the two layers (Yih 1980). Even so, our model describes a significant amount of variability in the depth of the LSL as a function of rainfall.
Potential consequences for community structure. - A likely consequence of reduced predation in the LSL is reflected in the vertical distribution of Mytilus edulis galloprovincialis [ILLUSTRATION FOR FIGURE 3 OMITTED]. The lower half of the mussels' vertical range (3-6 m) is exposed to higher predation than the upper half, which is frequently within the LSL. We suggest that the maximum abundance of mussels occurs in low salinities at depths of 3.0 m partly because M. edulis galloprovincialis has a spatial escape from predatory sea stars and urchins, which abut the lower limit of the LSL and mussel zone. Where they were most abundant (3 m), mussels obtained a partial release of predation 70% of the time, or [approximately]20 d/mo and 1.5 predator-free days per month between October 1992 and May 1994 [ILLUSTRATION FOR FIGURES 8 and 9 OMITTED]. High predation from sea stars, urchins, and lobsters a few meters below the average depth of the LSL may regulate the lower limit of the mussel zone. A partial and complete predator release of 10.6 and 0.5 d/1.5 yr, respectively, appears to be insufficient in preventing predators from controlling the distribution of mussels occurring below 5.0 m. The interpretation that the mussel distributions are regulated by predators must be substantiated by controlled reductions of predators at their lower limit (Paine 1966), but it is a logical interpretation for a pattern of high densities of mussel consumers coinciding with a zone of declining mussel percentage cover.
The mussel recruitment experiment provided a preliminary indication of substantial recruitment variation across the 3-18 m depth range. The absence of any recruitment to 10 and 18 m depth suggests that recruitment was confined to the shallower depths of 3 and 6 m, and that recruitment limitation prevents the Mytilus zone from penetrating deeper than 6 m into the subtidal zone at these two sites. There was a trend of higher recruitment to the lower edge of the mussel zone at 6 m than to the center of the zone at 3 m [ILLUSTRATION FOR FIGURE 3 OMITTED]. These recruitment results are consistent with the hypothesis that predation regulated by salinity stress is the major determinant of the lower limit of the mussel zone at 6 m, but because recruitment was so low, the experiment should be repeated.
Predator impact depends on interaction strength (Paine 1980), on the extent to which different species of predators are substitutable or ecologically "redundant" (Lawton and Brown 1994), and on the propensity for interactions among different predator species (Wootton 1994, Navarette and Menge 1996). We doubt that all species of invertebrate predators described here (Appendix) have equal effects, but the lack of basic information on the diets of subtidal predators in New Zealand makes this determination difficult. In northern New Zealand, the endemic sea urchin Evechinus chloroticus consumed algae, sponges, and ascidians, resulting in reduced species diversity (Ayling 1981). E. chloroticus is typically associated with subtidal substrata encrusted by coralline algae (Ayling 1981, Choat and Schiel 1982). Like other urchins in temperate reefs (Witman 1985), E. chloroticus plays an important role in maintaining coralline algal flats in northern New Zealand (Ayling 1981). We suggest that E. chloroticus may be the most important biological agent of space creation at the Tricky Cove sites, where it may have a similar influence on diversity. The regulation of E. chloroticus abundance by fluctuations in the LSL would likely influence the diversity and abundance of other epifaunal species besides mussels, including bryozoans, ascidians, and sponges. Indeed, the high cover of coralline algae at the ecotone where E. chloroticus is abundant (6 m depth, Table 2, [ILLUSTRATION FOR FIGURE 4 OMITTED]) may result from the removal of bryozoans, ascidians, and sponges as well as mussels by urchins. Another omnivorous urchin species, Pseudechinus huttoni, may play a similar role in the deeper fjord communities. Coscinasterias calamaria has a greater per capita effect on mussels than Patiriella regularis, as this sun star consumes mussels more rapidly than P. regularis, which also commonly feed in groups. Spiny lobsters had the greatest per capita effect on mussels, but it is difficult to infer how important lobster predation was in regulating the lower limit of the mussel zone, because we did not observe them foraging at 6 m depth (including nocturnal observations). The large difference in predation intensity at 10 m depth between the two sites was due to the greater abundance of lobsters at the ET site (Appendix). We suspect this is related to more suitable habitats for adult lobsters (overhanging ledges, crevices) at the ET site.
Environmental variability and predation refuges. - We view the role of environmental variation in this system as interrupting predation on shallow subtidal species and determining refuge size and availability. Variation in the key environmental factor, rainfall, was stochastic over the temporal scale of the 1.5-yr study period, as heavy rains could occur any month (Stanton and Pickard 1981, Grange et al. 1991). Temporal variation in rainfall and refuge size may become increasingly predictable over a period of tens of years if seasonality is evident (Chesson 1986).
Although the importance of diel variation in predation refuges has long been in recognized some taxa (Zaret and Suffern 1976, Heinrich 1979), there has been surprisingly little research on the relationship between environmental variability and temporal fluctuations in the size or effectiveness of predation refuges. This topic deserves more attention as a potential consequence of anthropogenic climate change where environmental variability is predicted to increase (Schneider 1992). Clearly, large temporal variability in refuge effectiveness should be more characteristic of refuges driven by physical environmental conditions than those created by persistent biogenic habitat structures or substrate heterogeneity.
Predators often restrict prey to shallow habitats at the margin of lakes (Werner et al. 1983, Mittelbach 1984) and streams (Sih 1982, Power 1987), or to tidal elevations above the foraging range of low intertidal or subtidal predators in the rocky intertidal zone (Paine 1966, 1971, Dayton 1971, Menge and Lubchenco 1985) and salt marshes (Kneib 1987). Many of these "marginal habitats" are environmentally stressful, particularly in high intertidal zones (Garrity and Levings 1981, reviewed in Underwood 1986) and in the shallow subtidal zone of fjords (this study). The ability of prey species to avoid predation by living in environmentally stressful refuges where predators are rare comes with a cost to the prey species of tolerating environmentally severe conditions (Sih 1987). In this example of the Mytilus edulis galloprovincialis zone occurring in the LSL, low salinity is apparently the environmental stressor. While the exact salinity tolerance of M. edulis galloprovincialis is not known, we infer from extensive research on the conspecific Mytilus edulis that mussels in the upper half of the zone (i.e., at or above 3 m, [ILLUSTRATION FOR FIGURE 3 OMITTED]) are living close to their limits of physiological tolerance to low salinity. Due to its broad thermal tolerance, Mytilus edulis galloprovincialis (M. edulis aoteanus) is capable of dominating upper intertidal habitats in New Zealand (Kennedy 1976). Mytilus edulis shows physiological adaptation down to 20 mg/g, but salinities [less than]20 mg/g can cause death (Bayne et al. 1985). Here, the upper limit of the M. edulis galloprovincialis zone may be regulated by low salinity, since salinities from 1.0 m depth to the surface were always [less than]20 mg/g and were usually [less than]10 mg/g ([ILLUSTRATION FOR FIGURE 5 OMITTED], Grange et al. 1991). For mussels living below this extreme, but still within the average range of the LSL, potential costs of avoiding predation by tolerating low-salinity stress may involve reduced growth, reduced reproductive output, or both.
The low-salinity layer is a gradient of environmental (salinity) stress that diminishes with depth. As such, our suggestion that predation in the shallow subtidal is regulated by environmental stress is in accord with the prediction of Menge and Sutherland's (1976, 1987) model of community regulation that the importance of predation decreases with increasing environmental harshness. Compared to intertidal systems (reviewed in Menge and Farrell 1989) this key prediction has received little attention from subtidal researchers. Watanabe (1984) found that the distribution of predation refuges (algal habitats) was a better predictor of predation intensity on gastropods than environmental stress in subtidal communities off central California. Wave action imposes severe physical stress in shallow subtidal communities on the open coast (Witman 1987, Dayton et al. 1992), where it may limit predation. The gradient of diminishing wave action with depth (Riedl 1971) represents another subtidal gradient of environmental harshness where the predictions of these models of community regulation (Menge and Sutherland 1976, 1987) could be tested.
By establishing links between rainfall, the depth of the LSL, and predation, this study has implications for understanding how climatological phenomena altering the depth of the low-salinity layer may influence subtidal community dynamics. Although it is unknown if our results will apply to larger spatiotemporal scales, we speculate that the size of predation refuges and the distribution of salinity-tolerant invertebrates in fjords may increase with a greater influx of freshwater into the coastal zone, because their key marine predators will be displaced by greater downwelling of the low-salinity layer.
We thank F. Smith for invaluable assistance underwater and in the laboratory, and for his discussions of physical effects on fjord communities. R. Singleton, M. Page, and E. Harvey also helped with predator surveys. We are indebted to P. Meredith, L. Shaw, and B. Walker for the logistical support that made our research in remote field sites both possible and pleasurable. Thanks to R. Etter, M. Bertness, S. Genovese, J. Leichter, B. Menge, J. Barry, and an anonoymous reviewer for their critical reviews of the manuscript. We are grateful to B. Cole of the Center for Statistical Science (Brown University) for improving our understanding of survival analysis, and to S. Genovese for computing tidal curves. Thanks to P. Mladenov for facilitating many aspects of this research as Chair of the Department of Marine Sciences at the University of Otago (NZ). Our research was supported by grants INT-9221707 and OCE 9302238 to J. D. Witman from NSF International and Biological Oceanography Programs and by the NZ Foundation for Research, Science and Technology grant CO1121 to K. R. Grange.
Ayling, A. M. 1981. The role of biological disturbance in temperate subtidal encrusting communities. Ecology 62: 830-847.
Barry, J. P., C. H. Baxter, R. D. Sagarin, and S. E. Gilman. 1995. Climate related, long-term faunal changes in a California rocky intertidal community. Science 267:672-675.
Batham, E. J. 1965. Rocky shore ecology of a southern New Zealand fiord. Transactions of the Royal Society of New Zealand 6:215-227.
Bayne, L., D. A. Brown, K. Burns, D. R. Dixon, A. Ivanovici, D. R. Livingstone, D. M. Lowe, M. N. Moore, A. R. D. Stebbing and J. Widdows. 1985. Effects of stress and pollution on marine animals. Praeger, New York, New York USA.
Bertness, M. D., and R. Callaway. 1994. Positive interactions in communities. Trends in Ecology and Evolution 9:191193.
Bertness, M. D., and E. Grosholz. 1985. Population dynamics of the ribbed mussel Geukensia demissa: the costs and benefits of aggregated distribution. Oecologia 67:192-204.
Burrell, D. C. 1986. Interactions between silled fjords and coastal regions. Pages 187-216 in D. W. Hood and S. T. Zimmerman, editors. The Gulf of Alaska. Outer Continental Shelf (OCS) Study MMS86-0095, U.S. Department of the Interior, Washington, D. C., USA.
Caswell, H. 1989. Matrix population models. 1989. Sinauer Associates, Sunderland, Massachusetts, USA.
Chesson, P. L. 1986. Environmental variation and the coexistence of species. Pages 240-256 in J. Diamond and T. J. Case, editors. Community ecology. Harper and Row, New York, New York, USA.
Choat, J. H., and D. R. Schiel 1982. Patterns of distribution and abundance of large brown algae and invertebrate herbivores in subtidal regions of northern New Zealand. Journal of Experimental Marine Biology and Ecology 60: 129162.
Connell, J. H. 1961. The influence of interspecific competition and other factors on the distribution of the barnacle Chthamalus stellatus. Ecology 42:710-723.
-----. 1975. Some mechanisms producing structure in natural communities: a model and evidence from field experiments. Pages 460-490 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Belknap, Cambridge, Massachusetts, USA.
Crowley, P. H. 1981. Dispersal and the stability of predatorprey interactions. American Naturalist 118:673-701.
Davies-Colley, R. J. 1992. Yellow substance in coastal and marine waters around the South Island, New Zealand. New Zealand Journal of Marine and Freshwater Research 26: 311-322.
Dayton, P. K. 1971. Competition, disturbance and community organization: the provision and subsequent utilization of space in a rocky intertidal community. Ecological Monographs 41:351-389.
Dayton, P. K, M. J. Tegner, P. E. Parnell, and P. B. Edwards. 1992. Temporal and spatial patterns of disturbance and recovery in a kelp forest community. Ecological Monographs 62:421-445.
Denny, M. W., T. L. Daniel, and M. A. R. Koehl. 1985. Mechanical limits to size in wave-swept organisms. Ecological Monographs 55:69-102.
Dunson, W. A., and J. Travis. 1991. The role of abiotic factors in community organization. American Naturalist 138: 1067-1097.
Farmer, D. M., and H. J. Freeland. 1983. The physical oceanography of fjords. Progress in Oceanography 12:147-219.
Fox, G. A. 1993. Failure time analysis: emergence, flowering, survivorship and other waiting times. Pages 253289 in S. M. Scheiner, and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA.
Garrity, S. D, and S. C. Levings. 1981. A predator-prey interaction between two physically and biologically constrained tropical rocky shore gastropods: direct, indirect and community effects. Ecological Monographs 51:267286.
Grange, K. R., R. J. Singleton, W. M. Goldberg, and P. J. Hill. 1991. The underwater environment of Doubtful Sound: results from an instrument array moored from November 1987 to June 1989. New Zealand Oceanographic Institute Miscellaneous Publication 105.
Grange, K. R., R. J. Singleton, J. R. Richardson, P. J. Hill, and W. DeL. Main. 1981. Shallow rock-wall Biological associations of some southern fiords of New Zealand. New Zealand Journal of Zoology 8:209-227.
Hardwick-Witman, M. N. 1985. Biological consequences of ice rafting in a New England salt marsh community. Journal of Experimental Marine Biology and Ecology 87:283-298.
Heinrich, B. 1979. Foraging strategies of caterpillars: leaf damage and possible predator avoidance strategies. Oecologia 42:325-337.
Holling, C. S. 1992. Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447502.
Huffaker, C. B. 1958. Experimental studies on predation: dispersion factors and predator-prey oscillations. Hilgardia 27:343-383.
Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54: 187-211.
Ives, A. R. 1995. Predicting the response of populations to environmental change, Ecology 76:926-941.
Karieva, P. M., J. G. Kingsolver, and R. B. Huey. 1993. Biotic interactions and global change. Sinauer Associates, Sunderland, Massachusetts, USA.
Kennedy, V. S. 1976. Dessication, higher temperatures and upper intertidal limits of three species of sea mussels (Mollusca: Bivalvia) in New Zealand. Marine Biology 35: 127137.
Kitchell, J. A., and J. F. Kitchell. 1980. Size selective predation, light transmission, and oxygen stratification: evidence from recent sediments of manipulated lakes. Limnology and Oceanography 25:389-402.
Kitching, A., and F. J. Ebling. 1967. Ecological studies at Lough Ine. Advances in Ecological Research 4:197-291.
Kneib, R. T 1987. Predation risk and use of intertidal habitats by young fishes and shrimp. Ecology 68:379-386.
Lawton, J. H., and V. K. Brown. 1994. Redundancy in ecosystems. Pages 255-270 in E. D. Schulze, and H. A. Mooney, editors. Biodiversity and ecosystem function. Springer-Verlag, Berlin, Germany.
May, R. 1981. Theoretical ecology: principles and applications. Second edition. Sinauer Associates, Sunderland, Massachusetts, USA.
Menge, B. A. 1978. Predation intensity in a rocky intertidal community: effect of an algal canopy, wave action and desiccation on predator feeding rates. Oecologia 34:17-35.
-----. 1983. Components of predation intensity in the low zone of the New England rocky intertidal region. Oecologia 58:141-155.
Menge, B. A., E. L. Berlow, C. A. Blanchette, S. A. Navarrete, and S. B. Yamada. 1994. The keystone species concept: variation in interaction strength in a rocky intertidal habitat. Ecological Monographs 64:249-286.
Menge, B. A., and T. M. Farrell. 1989. Community structure and interaction webs in shallow marine hard-bottom communities: tests of an environmental stress model. Advances in Ecological Research 19: 189-262.
Menge, B. A., and J. Lubchenco. 1981. Community organization in temperate and tropical rocky intertidal habitats: prey refuges in relation to consumer pressure gradients. Ecological Monographs 51:429-450.
Menge, B. A., and J. P. Sutherland. 1976. Species diversity gradients: synthesis of the roles of predation, competition, and temporal heterogeneity. American Naturalist 110:351369.
Menge, B. A., and J. P. Sutherland. 1987. Community regulation: variation in disturbance, competition, and predation in relation to environmental stress and recruitment American Naturalist 130:730-757.
Mittelbach, G. G. 1984. Predation and resource partitioning in two sunfishes (Centrarchidae). Ecology 65:499-513.
Mittelbach, G. G., and P. L. Chesson. 1987. Predation risk: indirect effects on fish populations. Pages 315-332 in W. C. Kerfoot, and A. Sih, editors. Predation. Direct and indirect impacts on aquatic communities, University Press of New England, Hanover, New Hampshire, USA.
Muenchow, G. 1986. Ecological use of failure time analysis. Ecology 67:246-250.
Murdoch, W. W., and A. Oaten. 1975. Predation and population stability. Advances in Ecological Research 9:2-132.
Navarrete, S. A., and B. A. Menge. 1996. Keystone predation and interaction strength: interactive effects of predators on their main prey. Ecological Monographs 66:409-429.
Officer, C. B. 1983. Physics of estuarine circulation. Pages 15-41 in B. H. Ketchum, editor. Estuaries and enclosed seas. Ecosystems of the world 26. Elsevier, Amsterdam, The Netherlands.
Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist 100:65-76.
-----. 1971. A short-term experimental investigation of resource partitioning in a New Zealand rocky intertidal habitat. Ecology 52:1096-1106.
-----. 1976. Size-limited predation: an observational and experimental approach with the Mytilus-Pisaster interaction. Ecology 57:858-873.
-----. 1980. Food webs: linkage, interaction strength and community infrastructure. Journal of Animal Ecology 49: 667-685.
Paine, R. T., and S. A. Levin 1981. Intertidal landscapes: disturbance and the dynamics of pattern, Ecological Monographs 51:145-178.
Peterson C. H., and R. Black. 1994. An experimentalist's challenge: when artifacts of intervention interact with treatments. Marine Ecology Progress Series 11:289-297.
Petraitis, P. 1990. Direct and indirect effects of predation, herbivory and surface rugosity on mussel recruitment. Oecologia 83:405-413.
Pickard, G. L. 1961. Oceanographic features of inlets in the British Columbia mainland coast. Journal of Fisheries Research Board of Canada 18:907-999.
Pickard, G. L., and B. R. Stanton, 1980. Pacific fjords: a review of their water characteristics. Pages 1-51 in H. J. Freeland, D. M. Farmer, and C. C. Levings, editors. Fjord oceanography. Plenum, New York, New York, USA.
Pickrill, R. A. 1987. Circulation and sedimentation of suspended particulate matter in New Zealand fjords. Marine Geology 74:21-39.
-----. 1993. Sediment yields in Fiordland, New Zealand, New Zealand Journal of Hydrology 7:223-229.
Power, M. E. 1987. Predator avoidance by grazing fishes in temperate and tropical streams: importance of stream depth and prey size. Pages 333-351 in W. C. Kerfoot, and A. Sih, editors. Predation. Direct and indirect impacts on aquatic communities. University Press of New England, Hanover, New Hampshire, USA
Pyke, D. A, and J. N. Thompson 1986. Statistical analysis of survival and removal rate experiments. Ecology 67:240245.
Rice, W. R. 1990. A consensus combined p-value test and the family wide significance of component tests. Biometrics 46:303-308.
Riedl, R. 1971. Water movement: animals. Pages 1123-1149 in O. Kinne, editor. Marine ecology, Volume 1. Wiley-Interscience, London, UK.
Robles, C., R. Sherwood-Stephens, and M. Alvarado. 1995. Responses of a key intertidal predator to varying recruitment of its prey. Ecology 76:565-579.
Robles, C., D. Sweetham, and J. Eminike 1990. Lobster predation on mussels: shore level differences in prey vulnerability and predator preference. Ecology 71:1564-1577.
Rosenberg, R., and P. Moller. 1979. Salinity stratified benthic macrofaunal communities and long-term monitoring along the west-coast of Sweden. Journal of Experimental Marine Biology and Ecology 37:175-203.
SAS. 1995. JMP statistics and graphics guide. Version 3.1. SAS Institute, Cary, North Carolina, USA.
Schneider, S. H. 1992. The climatic response to greenhouse gases. Advances in Ecological Research 22: 1-32.
Seed, R. A. 1969. The ecology of Mytilus edulis (L). Lamellibranchiata on exposed rocky shores. 2. Growth and mortality. Oecologia 3:277-316.
Seed, R. A., and T. H. Suchanek. 1992. Population and community ecology of Mytilus. Pages 87-169 in E. Gosling, editor. The mussel Mytilus: ecology, physiology, genetics and culture. Elsevier, New York, New York, USA.
Sih, A. 1982. Foraging strategies and the avoidance of predation by an aquatic insect Notonecta hoffmanni. Ecology 63:786-796.
-----. 1987. Predators and prey lifestyles: an evolutionary and ecological overview. Pages 203-224 in W. C. Kerfoot. and A. Sih, editors. Predation. Direct and indirect impacts on aquatic communities. University Press of New England, Hanover, New Hampshire, USA
Smith, F., and J. D. Witman. 1999. Patterns of species diversity in subtidal landscapes: maintenance by physical processes and larval recruitment. Ecology, in press.
Sokal R. R., and F. J. Rohlf. 1981. Biometry. W. H. Freeman, San Francisco, California, USA.
Stanton, B. R., and G. L. Pickard. 1981. Physical oceanography of the New Zealand fiords. New Zealand Oceanographic Memoir 88.
Systat. 1992. Statistics, Version 5.2 edition. Systat, Evanston, Illinois, USA.
Tunnicliffe, V. 1981. High species diversity and abundance of the epibenthic community in an oxygen deficient basin. Nature 294:354-356.
Underwood, A. J. 1981. Techniques of analysis of variance in experimental marine biology and ecology. Oceanography and Marine Biology Annual Review 19:513-605.
-----. 1986. Physical factors and biological interactions: the necessity and nature of ecological experiments. Pages 372-390 in P. G. Moore and R. Seed, editors. The ecology of rocky coasts. Columbia University Press, New York, New York, USA.
Watanabe, J. M. 1984. The influence of recruitment, competition, and benthic predation on the spatial distributions of three species of kelp forest gastropods. Ecology 65:920936.
Werner, E. E, J. F. Gilliam, D. J. Hall, and G. G. Mittlebach. 1983. An experimental test of the effects of predation risk on habitat use in fish. Ecology 64:1540-1548.
Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397.
Winer, B. J., D. R. Brown, and K. M. Michels. 1991. Statistical principles in experimental design. Third edition. McGraw-Hill, New York, New York, USA.
Witman, J. D. 1985. Refuges, biological disturbance and rocky subtidal community structure in New England. Ecological Monographs 55:421-445.
-----. 1987. Subtidal coexistence: storms, grazing, mutualism and the zonation of kelps and mussels. Ecological Monographs 57:167-187.
Woodin, S. A. 1978. Refuges, disturbance and community structure: a marine soft bottom example. Ecology 59:274284.
Wootton, J. T 1994. Putting the pieces together: testing the independence of interactions among organisms. Ecology 75:1544-1551.
Yih, C. S. 1980. Stratified flows. Academic Press, New York, New York, USA.
Zaret, T. M., and J. S. Suffern. 1976. Vertical migration in zooplankton as a predator avoidance mechanism. Limnology and Oceanography 21:804-813.
[TABULAR DATA FOR APPENDIX OMITTED]
|Printer friendly Cite/link Email Feedback|
|Author:||Witman, Jon D.; Grange, Ken R.|
|Date:||Oct 1, 1998|
|Previous Article:||Conspecific reproductive success and breeding habitat selection: implications for the study of coloniality.|
|Next Article:||Experimental reduction of predators reverses the crash phase of small-rodent cycles.|