REU: Kate Merrick  
Wolf Spiders in Humphreys Basin:
Exploring Population Density Patterns and Links to Aquatic Food Webs


Power Point Presentation

Introduced species often disrupt food webs in native communities, with resultant impacts on food webs in other linked communities. Stocking of Sierra Nevada lakes with non-native trout may have impacts that extend into terrestrial food webs, because the fish decrease emergence of insects from lakes, and many terrestrial animals prey on these insects. I studied several factors that may affect wolf spider and general spider abundance at alpine lakes, in order to determine if spiders are affected by introduced trout. Using a classification and regression tree I found that habitat type, classified as either low density or high density boulder, explained the most variability in wolf spider abundance, with date of sampling period next most important to explaining variability. Sampling period was first in importance for explaining variability in overall spider abundance, with habitat next most important. Presence or absence of fish only helped explain variability in overall spider abundance after the effects of sampling period, habitat, time of day, and weather were taken into consideration. Effects of trout are therefore still debatable, and may be indirect and tempered by other interspecific interactions, such as predation or competition, with other animals that are also influenced by trout presence or absence.

Introduced species often disrupt existing food webs, altering structure and function of native communities (Flecker and Townsend 1994, Simberloff and Stiling 1996). Food webs in distinct communities or even ecosystems are often linked to one other by the movement of nutrients or organisms (Polis et al. 1997, Nakano and Murakami 2001, Kato et al. 2003, Sanzone et al. 2003, Baxter et al. In Press). Therefore, impacts of introduced species may extend beyond ecosystem boundaries (Flecker and Townsend 1994, Knapp 1996). In particular, resource subsidies from one ecosystem to another often link aquatic and terrestrial food webs. For example, detritus and prey organisms from the ocean subsidize high populations of consumers on relatively unproductive Gulf of California islands (Polis et al. 1997). In riparian forest ecosystems, some fish consume terrestrial invertebrates that fall into streams, while insects emerging from streams reciprocally serve as prey for spiders and birds (Nakano and Murakami 2001, Kato et al. 2003, Baxter et al. In Press). Transfer of nutrients from streams to terrestrial food webs, via predation by spiders on emerging aquatic insects, may be paricularly important in desert habitats (Sanzone et al. 2003).
In California’s Sierra Nevada, introduced trout have already been shown to disrupt lake food webs. Of particular concern has been the widespread decline of the once common mountain yellow-legged frog, due to predation by trout (Drost and Fellers 1996, Knapp 1996, Knapp and Matthews 2000). Beyond the effect on this amphibian, trout introduction also has implications for at least one wholly terrestrial species. At lakes that have lost their mountain yellow-legged frog populations, mountain garter snakes that prey on frogs are also absent (Jennings et al. 1992, Matthews et al. 2002, Harper-Smith In Press).
Yet even the impacts of introduced fish on aquatic species may further affect terrestrial communities. In the Sierra Nevada, fewer insects emerge from fish-containing lakes than from fishless lakes (Knapp 1996, Knapp et al. 2001). Flecker and Townsend (1994) have similarly observed a decrease in insect densities in New Zealand streams containing introduced trout. In northern Japan as well, non-native rainbow trout have been observed to decrease the biomass of insects emerging from streams, with a resultant decrease in the density of riparian spiders (Kato et al. 2003, Baxter et al. In Press). Could the decrease in insects emerging from Sierra Nevada lakes also lead to a decrease in lakeshore spiders that prey on those insects? Or if the impacts of introduced trout on lake food webs do carry over into the terrestrial system, is the effect on spider populations less direct, and modulated by other factors?
Experiments that seek explanations for patterns in nature often work by holding equal among treatments all but the specific condition under scrutiny. Obviously the natural world is much less controlled, so that distinct units, such as the individual lakes in this study, can never be made equal in all respects other than a single differing condition such as presence or absence of fish. Therefore it is both interesting and important to explore several factors that vary among lakes in order to compare the relative importance of each within a multi-faceted explanation of patterns observed in nature. For example, if there is an effect of fish on spider abundance, how important is this effect relative to other factors that regulate spiders?
In this study I specifically looked for an effect of fish presence on spider abundance near alpine lakes in the Sierra Nevadas, but I also explored other factors that may help explain patterns of spider abundance. Furthermore, I compared relative explanatory power among these factors and fish presence/absence. Specifically, and in addition to the effect of fish, I explored the effects of sampling period, time of day, weather, lakeshore habitat type, relative abundance of birds, and lakeshore perimeter on the abundance of spiders. I considered effects both on a conspicuous species of wolf spider (Family Lycosidae, species undescribed) that is common throughout the study area, and on all spiders, regardless of species.

Study site: My study encompassed eight lakes in Humphreys Basin, in the John Muir Wilderness within the Sierra National Forest of Fresno County, California. These lakes are above treeline, with elevations ranging from 11,145’ to approximately 11,800’. Their shorelines include stretches of tundra, sand, and rock or boulder. Four of the studied lakes contain introduced trout and four are fishless. Lake surveys: I surveyed each lakeshore for spider abundance and other conditions of interest to this study at least three times between July 15 and August 4, 2004, with an effort to vary time of day among the three surveys at a given lake. During preliminary visual surveys of lakeshore spiders, I observed the focal lycosid much more frequently on rock or boulder than on sand or tundra. I therefore restricted my visual surveys for this study to the preferred habitats, in order to quantify spiders seen per area of rocky lakeshore. At each lake I counted spiders for six minutes at each of eight 10m x 2m transects, which were located in randomly selected rock or boulder areas along the shore. In order to minimize any effect of disruption, counts were never carried out at a given transect directly after setting up the transect. In addition to spider counts, each lake survey included documentation of date, time, temperature and cloud cover, habitat type at each transect, and at least one 10-minute point count of birds at the lakeshore. Data analysis: To facilitate comparisons, I grouped actual date, time, weather condition, habitat type, and relative bird abundance data into a few distinct categories each. To examine seasonal effects on spider abundance I reclassified dates into four categories: July 15-19, July 20-25, July 26-30, and July 31-August 4. To examine diurnal patterns I also reclassified time of day into four categories: earlier than 11:00am, 11:00am-1:59pm, 2:00pm-4:59pm, and 5:00pm and later. I divided weather conditions in two ways: temperature = 20° or > 20°, and either mostly sunny or mostly cloudy. Habitat classifications, originally recorded as low, medium, or high density boulder, or high density rocky (where there were no large boulders), were pooled into two categories: the original low density category, and a new high density category which encompassed the remaining three classifications, within which I observed little difference in spider abundance. Relative bird abundance, either high or low, was assigned to each lake based on patterns observed throughout the summer. I use JMP statistics software (version 5.0, SAS Institute Inc., 2002) to calculate mean number of lycosids and of all spiders per 20m2 transect for each value or category of the following factors: fish presence or absence, date, time of day, weather, habitat type, relative bird abundance, and lake perimeter. I also examined each factor’s relative power to explain patterns in spider abundance, using a classification and regression tree. Spider abundance data were averaged over transects at each lake to maintain the lake rather than the transect as the sampling unit.

Averaged over all factors considered, spider abundance per 20m2 of rock or boulder shoreline varied considerably among lakes (fig. 1). Mean number of lycosids per 20m2 ranged from 0.11 to 1.53, and mean number of spiders in general ranged from 1.51 to 5.11.
Averaged over all other factors, there was no significant relationship between fish presence or absence and mean abundance of either lycosids alone or of all spiders, although in both cases there was a trend toward higher abundance at fish-containing lakes (fig. 2).
Spider abundance averaged over all other factors varied significantly with habitat, with over three times as many lycosids, and nearly twice as many spiders in general, found in the pooled high density boulder habitats as compared to the low density boulder habitats (fig. 3).
Spider abundance also showed seasonal and diurnal patterns. Lycosid abundance peaked in the early-mid sampling period, between July 20 and July 25, while overall spider abundance continued to rise throughout the sampling season, and was highest during the latest sampling period, between July 31 and August 4 (fig. 4). Number of visible lycosids varied little throughout the day, but there was a trend toward higher counts during the period from11:00am to 1:59pm, and lowest counts after 5:00pm, with a similar but more exaggerated pattern in number of all visible spiders (fig. 5).
Abundance of all visible spiders, averaged over all other factors, was also correlated with weather, specifically with warm or cool temperature. More spiders in general were seen when the temperature was greater than 20°C, with a similar but not statistically significant trend in lycosids (fig. 6).
There was no significant correlation between lycosid or general spider abundance, averaged over all other factors, and lake perimeter. Neither was there a statistically significant correlation with relative bird abundance, though there was a trend toward lower lycosid abundance at lakes with relatively more birds (fig. 7). The four lakes in the higher bird abundance category were the same as the fishless lakes. Bird abundance was lower at three of the fish-containing lakes, and indeterminate at the fourth.
Spider abundance at lakes in Humphreys Basin is correlated with several of the above factors, but which factors actually explain the most variation in abundance? The classification and regression tree for mean number of lycosids (fig. 8), which explains 45% of variation in abundance, highlights habitat type as the most important factor for separating high from low lycosid abundance. Mean lycosid abundance per 20m2 in the pooled high density habitats was over three times higher than in the low density boulder habitats. Within the high density rock or boulder habitat, sampling period is the next most important factor in determining lycosid abundance. Mean abundance per transect in the early to early-mid season categories is about twice as high as in the late-mid to late season categories. Earlier in the season, time code best explains remaining variation in lycosid abundance, with a higher mean early in the day, while weather better explains variation later in the season, with a higher mean under warmer conditions. Fish presence, bird abundance, and perimeter did not emerge as important.
The classification and regression tree for mean number of all spiders (fig. 9) explains 55% of variation in abundance. It highlights sampling period as the most important dividing factor between low and high abundance, with higher mean abundance during the latest sampling period (July 31 – August 4) than during the previous three (July 15 – July 30). Earlier in the season, habitat had more power to explain variation in mean abundance, which was higher at high density habitats. Later in the season, lake perimeter was more important, with higher spider abundance at smaller lakes. Early in the season at high density habitats, weather was the next most important factor, with higher abundance under warmer conditions. Finally, early in the season at high density habitats under warm conditions, fish presence or absence helps explain some of the remaining variation, with higher mean spider abundance at lakes with fish. Bird abundance did not emerge as important.
These trees show that while many factors may be correlated with spider abundance, some factors explain more of the variation than others. In particular, variation in spider abundance due to fish is much smaller than variation due to other factors.

Research on community and ecosystem-level impacts of the intentional introduction of non-native species should be considered an essential contribution to the resource management decision-making process (Knapp et al. 2001). It seems particularly important to study impacts that are difficult to predict or to trace back to their causes, since these may be easily overlooked. Looking for such impacts while carrying out this study, the equal importance of understanding the factors other than an introduced species that may affect a focal organism became apparent, because there must be multiple factors behind the high variability in spider abundance that I observed among lakes. With a clearer understanding of these factors it may be possible either to isolate the actual contribution of the introduced species to the overall variability, or to recognize that the other factors are influential enough to overshadow any effect of the introduced species.
Because the classification and regression trees show that the latter is the case in my study, effects of introduced trout on spiders are still debatable. Lingering questions suggest possible directions for future study. For example, my census walk spider surveys were not designed to quantify absolute spider population density, though such a measure may be preferable for a study of fish effects. Activity level of spiders will determine what percentage of the entire population is exposed and visible to the observer, and so influences the results of a census walk (Powell et al. 1996). This is desired when examining factors such as temperature, which are assumed to affect spiders’ activity level rather than their population size. My results, which show a significant and positive effect of warmer temperature on observed number of spiders of all species combined, agree with other studies that have shown spider activity to increase with temperature (Edgar 1971, Buddle 2000).
However, an assessment of absolute population density would give a more reliable comparison of any fish effects among lakes, because fish would hypothetically have more influence on the size of spider populations than on the activity level of spiders. My visual surveys still provide a reasonable comparison among lakes, because they were time-limited and spiders were approximately equally easy to observe at all transects, so number of visible spiders is assumed to be correlated with actual population size (Edgar 1971). However, in the future a mark and recapture study could perhaps more accurately assess spider population density (Edgar 1971, Powell et al. 1996, Kreiter and Wise 2001). My data on the effects of habitat and season could be used to plan trapping efforts for a mark and recapture study. Such a study would require a reliable method for catching spiders unharmed, which is a challenge because pitfall traps cannot be sunk into boulders and the lycosids in particular are large and fast enough to make collection with an aspirator difficult. I would suggest that any future studies of fish effects continue to focus on the distinctive lycosid found in Humphreys Basin. Including a single-species focus within a study of fish effects may eliminate one source of variability encountered when counting all types of spiders, because each species may be influenced differently by factors other than fish presence/absence.
The classification and regression trees suggest that among these factors, habitat and season are among the most important in explaining variability in lycosid and general spider population densities. The findings regarding lycosid seasonal patterns are particularly interesting, because little is known already about the life history of this species, which has yet to be described scientifically.
The finding that the difference between low density and high density boulder habitat is especially important to patterns in lycosid and general spider density leads to interesting interpretations and questions. High density boulder habitats at the study lakes offered more cracks within and between rocks than did low density boulder habitats. Lycosids in particular were usually observed disappearing into such cracks, or cautiously emerging from them, which suggests they are under significant predation pressure. In the context of research on factors that structure communities, this interpretation places my study within the debate over the relative importance of predation versus competition (Sih et al. 1985). In general, both top-down and bottom-up effects, in addition to physical factors, are determinants of population abundance, but each may be relatively more important under different conditions (Polis et al. 1998). Because habitat type falls out first in the partition of lycosid variability, and second in the partition of general spider variability, top-down effects would seem a key determinant of spider abundance if microhabitat preference in spiders is indeed primarily a response to predation. This would agree with other studies that have found spider populations to be highly responsive to predation pressures (Schoener and Toft 1983, Spiller and Schoener 1988).
Potential spider predators include birds and mountain yellow-legged frogs. Birds, in particular gray-crowned rosy finches were observed in greater numbers at fishless lakes, and frogs were seen only at fishless lakes. In high density boulder habitats, under early to mid-summer warm conditions, the positive relationship between fish presence and spider abundance may be explained by this negative effect of fish on potential spider predators. Availability of resource subsidies has been shown to lead to higher populations of subsidized organisms, which consequently decrease populations of their prey (Polis et al. 1997). In this case, emerging aquatic insects at fishless lakes could be subsidizing birds and frogs, which could consequently decrease spider populations.
Alternatively, competition may be the basis for an indirect effect of fish on spiders: birds and frogs, in higher numbers at fishless lakes, could be competing with spiders for emerging insect prey. Shifting the focus toward bottom-up effects, this hypothesis is more in line with several studies that have shown spider abundance to depend on prey availability (Wagner and Wise 1997, Wise and Chen 1999, Kreiter and Wise 2001, Sanzone et al. 2003). However, it contrasts with findings such as those of Wenninger and Fagan (2000), which indicate that population densities of some wolf spiders are independent of prey availability, and are determined instead by physical factors such as moisture and temperature.
Physical factors certainly contributed to explaining variation in spider abundance in this case, but the less conclusive nature of my study of biological, trophically-mediated factors highlights a critical gap that could be filled by future research. Further examination of the trophic relationships of Humphreys Basin spiders is needed, in order to assess their position within currently understood terrestrial and aquatic food webs, such as those contstructed by Harper-Smith et al. (In Press). Research on population dynamics should be considered in the context of food webs not only to insure that important trophically-mediated indirect effects on a population’s characteristics are taken into account, but also to give the research greater ecological and practical relevance (Cohen et al. 1993, Baxter et al. In Press). Clearly linking lycosids and all spiders to other species in their communities would hopefully provide a more mechanistic explanation of the factors that determine their population abundances. Effects of each factor could be followed link by link, through and between systems, facilitating examination of possible indirect effects of introduced trout or other organisms.

About Us Science Fridays Staff Corie Cann: Bio Home Kate Merrick References About Me