(Spatial scales of variation) I. Fu Abstract: The spatial scales of variation of three intertidal invertebrates were determined using a nested sampling design and optimal quadrat size for each species. Sampling was done at sites, which were kilometers apart; subsites within sites, which were roughly 100 meters apart; transects within subsites, which were about 10 meters apart; and quadrats, which were on the scale of meters apart. An analysis of variance revealed that Lottia limatula is patchy on the spatial scale of tens of meters, while Serpulorbis squamigerus is patchy on the spatial scale of hundreds of meters, with considerable spatial variation within the patches themselves. For Littorina planaxis, there were no significant differences at any of the three levels of nesting, indicating that the snail is fairly evenly distributed in the sites sampled. The index of departure from the Poisson distribution and the index of aggregation for S. squamigerus and L. planaxis indicate that both species tend to have clumped distributions, which agree with the natural histories of these organisms. The indices are more ambiguous for L. limatula, but the presence of a few high and many low counts, as well as aspects of the natural history of this organism, indicate a clumped distribution. Determination of the spatial scales of variation for an organism to be studied permits the formation of a sampling design that maximizes accuracy and precision while minimizing the total costs in terms of time, effort, and money. I. Fu (Spatial scales of variation) Introduction: One problem often encountered in sampling is that the patchiness of the distribution of an organism is often not known before a study is conducted. A common method to assess differences between sites is to take replicate samples, on a spatial scale of meters, within sites that themselves may be separated by kilometers. The problem with this sampling design is that it will not be able to detect variation on spatial scales between meters and kilometers. If an organism is patchy on the scale of hundreds of meters, all of the one-meter replicates may fall between two patches, resulting in a low mean abundance for that organism in that site. On the other hand, the replicates for the other site may all fall within one large patch, resulting in a high mean abundance at the second site. Using this sampling method, the two sites would incorrectly appear to have very different mean abundances, when they are actually fairly similar. One solution to this problem is to employ a nested sampling design, in which replicate samples are taken at different spatial scales so that variations at intermediate spatial scales can be distinguished. In this study, sampling for three intertidal invertebrates was done following a fully nested design. The species studied were Lottia limatula, a limpet found in the middle to low intertidal zone; Serpulorbis squamigerus, a tube snail found low in the intertidal zone or subtidally; and Littorina planaxis, a periwinkle abundant on the rocks of the high intertidal zone. Although numerous studies have been conducted on the limpet, little work has been done on its abundance and distribution. Some studies have determined the number of limpets at one or several particular sites in the Monterey Peninsula area (Ruth 1948, Hahn 1985), but the overall patterns of distribution for this well-studied limpet are not known. S. squamigerus occurs in concentrations of up to 650 m 2 in southern California (Pequegnat 1964) and has been seen singly or in clusters in central California (Morris et al. 1980). In the last decade or so, the abundance of the tube snail has increased dramatically in the Monterey area (C. Baxter, pers. comm.). Although the species is noted as gregarious by a number of studies (Holmes 1900, Morton 1965, MacGinitie & MacGinitie 1968), the larger spatial scales of variation for this organism are not known. L. planaris group together in cracks or other irregularities in the rock surface, and it has been I. Fu (Spatial scales of variation) theorized that this clumping behavior helps reduce desiccation (Miyamoto 1968). Like the other two species, the larger spatial scales of variation have not been studied. Use of a fully nested sampling design enabled the determination of the spatial scales of variation for these three invertebrates. This study illustrates the importance of doing such a nested sampling design to obtain accurate estimates of the abundance and distribution of organisms. Materials and Methods: Determination of optimal quadrat size This portion of the study was conducted in the rocky intertidal area at Hopkins Marine Station (HMS) in Pacific Grove, California (Fig. 1). Sampling procedures consisted of placing 20-meter transects in areas of comparable vertical height and then counting the number of desired organisms present in randomly spaced quadrats. Notes were also taken on the microhabitat of each quadrat. For L. limatula, two transects were located in the mid-intertidal zone of an area protected from wave action, with one crossing predominantly horizontal surfaces and the other passing primarily over vertical surfaces. A third transect was situated in a mid-intertidal area that was more exposed to wave action, and it encompassed both vertical and horizontal surfaces. For each transect, five 0.0625 m2 quadrats were sampled. The counts were repeated at the same locations using 0.25 m2 and 1.0 m2 quadrats. For S. squamigerus, one transect was run in the low-intertidal zone, and the numbers of tube snails were counted in 0.0625 m2 and 0.25 m2 quadrats. Finally, one transect was run in the high-intertidal zone for L. planaris, with 0.0625 m2 and 0.25 m¬ quadrats. The precision of the counts for each transect was calculated using the following formula: P =SE/X, where p is the precision of a set of counts, SE is the standard error of the counts for one transect, and Xis the average count per quadrat for that transect (Underwood 1981). Then, setting the desired precision at the lowest value obtained for the different quadrat sizes, the number of I. Fu (Spatial scales of variation) quadrats of each size that would need to be counted to obtain this level of precision was determined for each quadrat size and species. Based on the assumption that a 1.0 m2 quadrat takes 16 times the effort needed to examine a 0.0625 m2 quadrat, the cost of counting the number of quadrats needed to get the desired precision was calculated for each quadrat size using the following equation: C=nc, where Cr is the total cost of the sampling effort (time), n'is the number of quadrats of a particular size needed to obtain the desired precision level, and cis the cost in terms of time and effort needed to count a quadrat of a particular size (Underwood 1981). The quadrat size that resulted in the lowest total cost for each species was then used in the second part of the study. Determination of the spatial variation of three species Sampling methods for the second part of the experiment were similar to those used in the preliminary study, except that the transects were 15 rather than 20 meters in length and only the optimal quadrat size for each species was used. For all the species, the study involved sampling at the following four spatial scales: sites, which were kilometers apart; subsites within sites, which were separated by roughly 100 meters; transects within subsites, which were about 10 meters from each other; and quadrats, which were on the scale of meters apart. This was a fully nested sampling design, with a specific number of quadrats counted for each transect, three transects counted in each subsite, and two subsites examined for every site that was chosen (Fig. 2). For L. limatula, the four sites chosen were HMS, Carmel, Point Pinos, and the Great Tide Pool (Fig.1). Along each transect, 20 0.0625 m2 quadrats were counted. For S. squamigerus, three sites were chosen-HMS, the area between Fisherman's Wharf and the Coast Guard wharf, and Point Pinos (Fig. 1). Five 0.25 m2 quadrats were counted for each transect. For L. planaris, the three sites chosen were HMS, Carmel, and Point Pinos (Fig. 1), and eight 0.0625 m2 quadrats were counted per transect. I. Fu (Spatial scales of variation) Data Analysis Cochran’s tests were run on the data to detect heterogeneity of variances (Underwood 1981). For all three species, the Cochran’s test were significant (pæ0.05), indicating that the variances were too heterogeneous to meet the assumptions of an ANOVA. To increase the homogeneity of the transect variances, the limpet and tube snail counts were transformed by taking the square root of the count for each quadrat and then adding the result to the square root of the count plus one (Sokal & Rohlf 1995). L. planaris data was transformed by taking the square roots of the counts. After transformation, only the limpet data remained heterogeneous. Subsequent tests for these data were conducted at a-0.0l rather than 0.05. The transformed data were then analyzed using a three factor nested ANOVA. To determine whether the distributions of the species were random, regular, or aggregated, the variance to mean ratio, the index of departure from the Poisson curve, and the index of aggregation were calculated for each of the species (Hurlbert 1990). The index of departure was calculated using the following equation: Dp=1-2 min (qy0, qk0. were Dp is the index of departure, w is the largest number of individuals observed in any quadrat, k is the number of individuals per quadrat, qup is the number of quadrats expected to contain k individuals in a Poisson distribution, and qua is the number of quadrats containing k individuals in the observed distribution (Hurlbert 1990). The index of aggregation was calculated using the following formula: IM-(X/X-1) (1/X) (Var /X+X-1), where IM is the index of aggregation, Xis the total number of individuals that were counted, Xis the average count for all the quadrats sampled, and Var is the variance. IM measures how many times more likely it is that two randomly selected individuals will be from the same quadrat than if the X individuals in the population were distributed at random (Hurlbert 1990). (Spatial scales of variation) I. Fu Results: Mean limpet densities per transect ranged from 0.05-1.45 m2 (Fig. 3). There were significant (pe0.05) differences among transects within sites, but not at either of the larger spatial scales (Table la). A significant difference at the transect level indicates that, in at least one subsite within one site, there were significantly different mean limpet densities between transects. The largest differences among the transect averages were in Subsite 1 in Carmel, which most likely contributed to the significant result at the transect level (Fig. 3). There was some variation between the average limpet counts for the subsites at HMS, Carmel, and the Great Tide Pool, but these differences were not significant (p»0.05). Mean tube snail densities per transect ranged from 0-584 m22 (Fig. 4). There were significant differences among transects and subsites for S. squamigerus but not among sites (Table 1b). Large differences among transects and subsites at HMS most likely contributed to the significant differences at those levels of nesting (Fig. 4, Table 1). An apparently large difference between the average tube snail abundances at the Fisherman's Wharf and Point Pinos sites was not significant (p2O.05). For L. planaris, mean densities per transect ranged from 145.6 606.4 m2 (Fig. 5). There were no significant differences at any of the three spatial scales. Residual variance contributed 84.3% of the total variance for L. planaxis and 92.8% of the total variance for L. limatula. (Fig. 6). For the tube snail, the four sources of variation contributed roughly equal percentages of the total variance (Fig. 6). The variance to mean ratios and indices of aggregation calculated for the three species were all greater than one (Table 2). The indices of departure from the Poisson curve ranged from a high of 0.844 for S. squamigerus to a low of 0.125 for L. limatula, with an intermediate value of 0.632 obtained for L. planaxis (Table 2). These values indicate that the observed distribution for the limpet has a fairly large degree of overlap with a Poisson distribution of the same mean (Fig. 7a). while the observed distribution for the tube snail has a relatively small degree of overlap (Fig. 7b). I. Fu (Spatial scales of variation) The observed distribution for L. planaris has an intermediate degree of oversap with the corresponding Poisson distribution (Fig. 7c). Discussion: Because L. limatula had significant differences among transects (within subsites), it is apparently patchy on a scale of tens of meters. Optimal sampling efforts should thus concentrate on this spatial scale to detect changes among sites. In contrast, S. squamigerus varied significantly among both transects and subsites. This indicates that the tube snail is patchy on a scale of hundreds of meters and that there is a great deal of spatial variation within the patches themselves. For L. planaxis, there were no significant differences at any of the three spatial scales, indicating that it is spread out fairly evenly throughout the sites. The large contribution of residual variance to the total variances for L. limatula and L. planaxis indicate that there was considerable variation among quadrats (Fig. 6). One difficulty with a nested sampling design is that the degrees of freedom (and hence the power of the test) decrease at increasingly higher levels of nesting. For example, in this study, the mean abundances of tube snails at Fisherman's Wharf and Point Pinos appeared to be fairly different, yet the analysis of variance indicated this difference was not significant. This surprising result is due partly to the small power of the test at that level of nesting (3 and 4 degrees of freedom). Another problem with this sampling design is that a large variance at a lower level in the hierarchy tends to obscure variation at higher levels. For the tube snail, 25% of the total variance was contributed at the subsite level for this species, meaning that only a very large difference in mean abundances at the sites would have been detected as significant. The large Dp for S. squamigerus shows that very little of the tube snail's observed distribution conformed to a Poisson distribution. The HI indicates a clumped distribution, which is supported by the natural history of the organism. Because S. squamigerus is a sessile snail, the differences in mean abundances at the subsite and transect level most likely resulted from patterns of larval settlement. Based on the gregarious nature of the tube snail, the larvae are thought to (Spatial scales of variation) I. Fu settle on or near adults of the same species (Hadfield 1966). This continual addition of new individuals to areas near established populations could result in the larger spatial scale of patchiness that was observed for this species. Also, if some of the larvae settle in the existing large patches, then variation could occur within the patches at smaller spatial scales. A particle feeder, the tube snail spins delicate mucus webs to trap its food and therefore prefers more quiet water (Keen 1960). At HMS, the tube snails had a much higher mean density in the subsite that was more protected from wave action than in the subsite that was more exposed (Fig. 4). Thus, localized variation in water motion could also contribute to spatial patchiness. The observed distribution of L. planaris did not closely follow that of a Poisson distribution. An I of 1.55 for the organism suggests a clumped distribution, which is supported by previous studies on the natural history of this snail. Miyamoto (1964) theorized that L. planaxis tends to cluster in cracks in an attempt to reduce desiccation, and Bush (1964) has shown that groups of snails desiccate at a slower rate than single individuals in both laboratory and field conditions. If the snails are grouped together in cracks or irregularities in the rocks, their distribution will be clumped, with most of the variation occurring on the spatial scale of meters. Mating behavior in L. planaxis could also affect its patterns of distribution, for males can often be found sitting on the shells of females throughout the year, although reproduction in mass seems to be confined to spring and summer (Ricketts et al. 1952, Gibson 1964). Other possible factors include the abundance and distribution of the predatory snail Acanthina spirata, which releases a waterborne chemical stimulus that elicits an avoidance reaction in the L. planaxis; the tendency of the snails to follow mucus trails laid down by other snails (Peters 1964); and wave action, which appears to selectively knock larger snails off of rock faces (Bigler 1964). All or a combination of these factors may have contributed to the variation observed at a spatial scale smaller than tens of meters for L. planaxis. Limpet distribution did not depart strongly from a Poisson distribution (Dp-O. 13). This appears to contradict the l, which indicates a rather clumped distribution (I-3.6). The apparent conflict may be due to the overall low abundance of limpets. Iy is very sensitive to the Var/X I. Fu (Spatial scales of variation) ratio. Thus, the presence of outliers will greatly increase the variance and inflate M, especially if X is small. In contrast, Dp is less sensitive to departures from the Poisson distribution when the overall abundances are very low. In the case of L. limatula, which had most of the quadrat counts at zero with a few outliers, Dp will not weight the outliers as heavily, and the observed distribution could fit a Poisson distribution fairly well. Two of the quadrats had counts of seven or higher, showing that there were clumps in the limpet distribution. Aspects of the natural history of the limpet could result in a patchy distribution. For example, L. limatula exhibits an escape response to predatory sea stars, especially Pisaster ochraceus, by moving upward on vertical rocks or downstream on horizontal ones in response to a sea star scent in the water (Phillips 1974). Thus, if the distribution of sea stars is patchy on the scale of tens of meters, then the variation at that spatial scale observed in the limpets may have been due to this avoidance behavior. In terms of food availability, the limpets feed on microscopic algae, as well as the encrusting algae Hildenbrandia and Peyssonnelia (Kitting 1978, Eaton 1968). If the abundances of these algae vary on the scale of tens of meters, then they may have been partly responsible for the significant differences among transects. Patchy larval settlement and juvenile survival could also have contributed to the significant differences at that spatial scale. With limited time, money, and effort available for conducting research, it is imperative that any sampling done on an organism provide an accurate estimate of the abundance and distribution of the species of interest. Without knowing beforehand what the spatial scales of variation for the organism are, however, any sampling effort cannot be depended on for reliable results. This study illustrates the value of doing a preliminary study, using a nested sampling design, on an organism to determine the patchiness of that animal’s distribution before an extensive and costly sampling effort is made. Literature Cited Andrew, N. L. and B. D. Mapstone. 1987. Sampling and the description of spatial pattern in marine ecology. Oceanography and Marine Biology Annual Review. 5:39-90. Bigler, E. 1964. Attrition on the Littorina planaxis population. Final papers, Biol. I75H. library, Hopkins Marine Station of Stanford University, Pacific Grove, California. Bush, P. S. 1964. Effects of desiccation on Littorina planaris. Final papers, Biol. 175H, library, Hopkins Marine Station of Stanford University, Pacific Grove, California. Eaton, C. M. 1968. The activity and food of the file limpet, Acmaea limatula.. Veliger. 11:5-12. Gibson, D. G. 1964. Mating Behavior in Littorina planaxis Philippi. Veliger. Il (Suppl.):134-137. Hadfield, M. G. 1966. The reproductive biology of the California vermetid gastropods Serpulorbis squamigerus (Carpenter, 1857) and Petaloconchus montereyensis Dall, 1919. Doctoral thesis, Biological Sciences, Stanford University, Stanford, California 174 pp. Hahn. T. P. 1985. Effects of predation hy black oystercatchers (Haematopus bachmani audubon) on intertidal limpets. Masters thesis, Biological Sciences, Stanford University, Stanford, California 70 pp. Holmes, S. J. 1909.The early cleavage and formation of the mesoderm of Serpulorbis squamigérus Carpenter. Biological Bulletin. 1:115-121. Hurlbert, S. H. 1990. Spatial distribution of the montane unicorn. Oikos. 58:257-271. Keen, A. M. 1960. Vermetid gastropods and marine intertidal zonation. Veliger 3:1-2. Kitting, C. L. 1979. Foraging of individuals within the limpet species Acmaea (Notoacmea) scutum at Monterey Bay, California, and the consequences on their mid-intertidal algal foods. Doctoral thesis, Biological Sciences, Stanford University, Stanford, California 192 pp. MacGinitie, G. E., and N. MacGinitie. 1968. Natural history of marine animals 2nd ed. McGraw-Hill, New York, USA. Mivamoto, A. 1964. Clustering in Littorina planaxis. Final papers, Biol. 175H, library Hopkins Marine Station of Stanford University, Pacific Grove, California. Morris, R. H., Abbott, D. P., and E. C. Haderlie. 1980. Intertidal invertebrates of California. Stanford Universtiy Press, Stanford, California, USA. Morton, J. E. 1965. Form and function in the evolution of the Vermetidae. Bulletin of the British Museum (Natural History). 11:585-630. Pequegnat, W. E. 1964. The epifauna of a California siltstone reef. Ecology 45:272- 283. Peters, R. S. 1964. Function of the cephalic tentacles in Littorina planaris Philippi. Veliger. 11(Suppl.):143-148. Phillips, D. W. 1974. Distance chemoreception and avoidance of the predatory starfish Pisaster ochraceus by the gastropods Acmea (Collisella) limatula and Acmaea (Notoacmaea) scutum. Doctoral thesis, Biological Sciences, Stanford University, Stanford, California 156 pp. Ricketts, E. F., Calvin, J. C., and D. W. Phillips. 1985. Between Pacific Tides Sth ed. Stanford University Press, Stanford, California, USA. Ruth, F. S. 1948. Studies of the natural history of limpets of the family Acmaedde. Masters thesis. Department of Zoology, University of the Pacific, Stockton, Calfornia 153 pp. Sokal, R. R. and F. J. Rohlf. 1995. Biometry 3rd ed. W. H. Freeman and Company, New York, USA. Underwood, A. J. 1981. Techniques of analysis of variance in experimental marine biology and ecology. Oceanography and Marine Biology Annual Review. 25:513. 605. lable 1: Results of the analysis of variance for A) Lottia limatula B) Serpulorbis squamigerus and C) Littorina planaxis Source Mean Squares F-ratio P-value Sites 1.441 0.569 0.665 Subsites 2.534 1.437 0.267 Transects 16 1.763 2.225 0.004 Residual 456 0.793 Source Mean Squares F-ratio P-value Sites 863.799 2.495 0.230 Subsites 346.166 4.063 0.033 Transects 85.204 3.884 0.000 Residual 21.936 72 Source Mean Squares F-ratio P-value Sites 14.025 1.376 0.377 Subsites 10.192 2.253 0.135 Transects 4.523 1.597 0.100 Residual 126 2.832 Table 2: Index of departure from the Poisson distribution (Dp), index of aggregation (ID, and variance to mean ratios of the three species. Dp refers to the degree of overlap between the observed distribution and the Poisson distribution. M measures how many times more likely it is that two randomly selected individuals will be from the same quadrat than it would be if the individuals in the population were distributed at random. Species Variance: Mean Index of Departure Index of aggregation Lottia limatula 2.24 0.125 3.626 Serpulorbis squamigerus 0.844 5.93. Littorina planaxi 1.55 12.99 0.63. Figure legends: Fig. 1: Map of the Monterey Peninsula area, showing the four sites that were sampled in this study Fig. 2: Nested sampling design used in the study. Three or four sites were sampled per species, with two subsites nested within each site, three transects nested within each subsite, and a particular number of quadrats for each transect. The number of quadrats varied for the different species Fig. 3: Mean densities of Lottia limatula at the different spatial scales. Each vertical bar represents the mean limpet density for a particular transect. The arrows point to the mean limpet densities for subsites, and the mean limpet density for a site is seen by finding the halfway point between the two arrows for the subsites of that site. The error bars are the standard errors. Fig. 4: Mean densities of Serpulorbis squamigerus at the different spatial scales. Each vertical bar represents the mean limpet density for a particular transect. The arrows point to the mean limpet densities for subsites, and the mean limpet density for a site is seen by finding the halfway point between the two arrows for the subsites of that site. The error bars are the standard errors. Fig. S: Mean densities of Littorina planaxis at the different spatial scales. Each vertical bar represents the mean limpet density for a particular transect. The arrows point to the mean limpet densities for subsites, and the mean limpet density for a site is seen by finding the halfway point between the two arrows for the subsites of that site. The error bars are the standard errors. Fig. 6: Percentage of the total variance contributed by each level of nesting in the sampling design. Fig. 7: A) Graph of the observed distribution of Lottia limatula and the Poisson distribution for the same mean, showing a large degree of overlap between the two plots. B) Graph of the observed distribution of Serpulorbis squamigerus and the Poisson distribution for the same mean, showing very little overlap between the two plots. C Graph of the observed distribution of Littorina planaxis and the Poisson distribution for the same mean, showing some overlap between the two plots. Fig. 1: V Great Tide Pool Point Pinos Pacific Grove MONTEREY PENINSULA Carmel Hopkins Marine Station O Monterey Fisherman's Wharf N dotoaatia- * 1unoo ledun eAV 2 oiisans Liisn 2 oisgns Loiisans 2 eiisgns 1Siisns 2 eiisgns 1 eiisgns 82 389888828 lunoo jieus ogni 2eiAy 2 eusgns Leiisans 2 ousgns 1 Siisgns 2 ousgns 1eiisgns sttakatkakaovoa- 1unoo jeus 2beiAV 2 ousgns 1 S1sgn z ausgns Leiisgns 2 ousans 1 Sisgns 38aa- uonelen jeioi jo juesie Fig. 7: 0.8- LActual Distribution 0.6 — Random Distribution 0.4 0.2- 5 6 7 10 Count per quadrat 0.2 —Actual distribution — Random distribution 5 0.1 0.05 n nn e n o aaaaaaaaaaaaaaaaaa- Count per quadrat g0.09 - —Actual Distribution 0.08 + — Random Distribution 0.07+ 0.06 + 0.05 5 0.04 0.03 §0.02 30.01 O n ana 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76 80 Count per quadrat