Field research was conducted during June 1998 on Bodega Head, a small peninsula along the California coast about 80 km north of San Francisco. The peninsula rises up to 80 m in elevation and is composed of Mesozoic granitic rock partly overlain by Pleistocene deposits (Barbour et al. 1973). Soils are acidic, sandy and well-drained. Mean annual temperature is 11.7 °C, mean annual precipitation, entirely as rain, is 785 mm, and winds come predominantly from the west and north-west (Bodega Marine Laboratory, archived data available at www.bml.ucdavis.edu). Most of the rainfall (84%) occurs from November through March, and this largely determines the growing season for grassland herbs, which extends from c. November until c. June.
The natural vegetation on Bodega Head is primarily coastal prairie and coastal scrub, two community types characteristic of the coast of northern California (Heady et al. 1995). The coastal prairie at Bodega Head is dominated by native and non-native grasses, with scattered to locally dense stands of the shrubs Baccharis pilularis (nomenclature follows Hickman 1993) and Lupinus arboreus (bush lupine, referred to hereafter as Lupinus or lupine). Shrubs of Lupinus create bare, nitrogen-enriched microsites in the grassland when they die (Maron & Jefferies 1999), and the percentage of species that are non-native is higher in microsites where lupines have recently died than in adjacent grassland (Maron & Connors 1996). We restricted our field surveys to areas where there was no known previous construction or agriculture. However, almost all of Bodega Head was probably grazed by sheep at times up until the mid-1960s, when use of the area was restricted to research and walking.
To test the hypothesis that degree of invasion differs between microhabitat types in the coastal prairie community on Bodega Head, we measured the cover of each vascular plant species in relation to distance from the ocean cliff edge, presence of dead remains of shrubs of Lupinus, topographic position in gullies, and patches of very shallow soil. We chose these factors because communications with Bodega Marine Reserve staff, preliminary inspection of the grassland and previous work on soil salinity (Barbour et al. 1973) and Lupinus (Maron & Connors 1996) at Bodega Head suggested that they might show the strongest relationships with invasion.
To test the effects of distance from the cliff edge, we located 50 × 50 cm plots at different distances from the cliff edge in two level sites, one behind the largest cove on Bodega Head (‘cove site’) and one on the nearest point (‘exposed site’). The cliff edge faced south-west at both sites. Transects at the exposed site paralleled a transect measured for soil salinity by Barbour et al. (1973), who found large differences in salinity over distances of tens of metres along their transect during the late spring. At each site, plots were placed at fixed distances (cove: 0, 2, 4, 10, 20, 30 and 40 m; exposed: 16, 36, 56 and 76 m) along six transects that started at random points along the cliff edge and ran perpendicularly inland. We chose the minimum and maximum distances at each site to avoid areas with stands of shrubs. We also incorporated a test of larger-scale distance from the ocean into our test of shallow vs. deep soils (see below).
To test the effects of past growth of Lupinus, we centred a plot on the remains of each of the 10 dead shrubs nearest to the plots at 30 and 40 m along the transects at the cove site, and compared plots with dead shrubs with the 10 nearest plots along transects. Dead shrubs consisted of partly broken canopies of bare, woody stems and main branches, which had little effect on light availability (Peter Alpert and Jonathon Blanchard, unpublished data). Work on lupine demography (Peter Alpert and Nathan Abare, unpublished data) suggested that these shrubs had died during the previous growing season.
To test the effects of topographic position in gullies, we randomly located eight 50 × 50 cm plots in each of three positions (bottom – slope < 5°; side – slope > 10°; shoulder – slope 5–10°) in the gully nearest the cove and the nearest gully of comparable size and aspect that was on the inland side of Bodega Head, near Bodega Bay. The first gully was selected so as to be relatively near other surveys and thus help make comparison between surveys more valid. The second gully was selected to replicate the first gully in topography but to potentially differ from it in microclimate and thus help test the generality of effect of topography in different gullies on Bodega Head.
To test the effects of patches of very shallow soil, we located 50 × 50 cm plots in four areas, each containing a cluster of one to three large shallow patches (> 5 m in diameter, < 25 cm in depth), as identified by probing the soil with a sharpened steel rod. These patches occurred on or near the tops of the small ridges and hills south of the cove, and appeared to be caused by the surface topography of the underlying bedrock. We systematically selected the two areas with shallow patches that were within 50 m of the cliff edge and that were nearest to the cove (‘ocean area 1’ and ‘ocean area 2’), and the two areas nearest to these that were similar in size, elevation and aspect but 200–300 m from the cliff edge (‘inland area 1’ and ‘inland area 2’). These areas lay near a longer transect measured for soil salinity by Barbour et al. (1973), who found large differences over the distance between the cove and inland areas. We located plots along random radii from the centre of each patch of shallow soil in each area, placing a ‘shallow’ plot along each radius at the point closest to halfway between the centre and edge of the patch where the soil was < 15 cm deep, and a ‘deep’ plot along each radius at the first place beyond the edge of the patch where the soil was > 60 cm deep. Replicates were distributed between patches in proportion to patch size: we measured 18 replicates at ocean 1, 12 at ocean 2, 12 at inland 1 and 8 at inland 2.
Cover of each vascular plant species in each plot was estimated visually according to a system of 14 cover classes (0–1, 1–5, 5–10, 10–15, 15–20, 20–25, 25–30, 30–40, 40–50, 50–60, 60–70, 70–80, 80–90 and 90–100% cover). In selected plots, the accuracy of these estimates was tested by comparing them with measurements of species occurrence at 100 points per plot. Visual estimates were generally within one cover class of point occurrence values and showed no tendency to be higher or lower. As June is probably the month when most of the grassland herbs set most of their seed, species abundance during June is likely to be related to seed production and plant fitness.
To test the hypothesis that microhabitats that differ in invasion also differ in resource availability or in soil characteristics likely to influence resource availability, we measured soil water content, depth, pH and conductivity (as a measure of salinity) in plots where we measured plant species abundances. Because we had limited time and resources, we were able to measure only a subset of plots for water content, pH and conductivity, and attempted to allocate measurements so as to provide the most information possible. We omitted soil measurements in plots with dead shrubs of Lupinus, based on the assumption that dead shrubs were not likely to affect soil depth, moisture, pH or salinity. We omitted measurements of water content, pH and conductivity at four intermediate distances along the cove site transects (2, 4, 10 and 30 m), and in randomly selected replicates in the test of shallow patches (nine replicates were measured at ocean 1, none at ocean 2, eight at inland 1, and five at inland 2). We were not able to measure soil nutrients and relied on previous studies (Maron & Connors 1996; Maron & Jefferies 1999) to demonstrate that soil nitrogen availability was relatively high under lupines.
We measured depth with a sharpened steel rod and recorded the mean of three measurements per plot. For the other soil measurements, we first collected and pooled three to five 10-cm-deep soil cores per plot. Cores were collected over as short a time interval as possible and by order of sets of replicates within each test (e.g. first replicate of each distance from cliff edge, second replicate, etc.). Each pooled sample was hand-cleaned of plant fragments and pebbles, weighed, dried at 60 °C for 72 h, and re-weighed to determine its water content. Although a single measure of water content is not a direct or absolute measure of water availability to plants, it is likely to provide a reasonable measure of relative water availability when soil texture and organic matter content are fairly homogeneous, which appeared to be the case in the grassland (Kolb 1999). We then sieved each dried sample to remove particles > 2 mm and suspended a subsample of 25 g in 50 mL of distilled, de-ionized water. We stirred the suspension four times at 20–30-minute intervals, vacuum-filtered it (Rhoades 1982), and measured it for electrical conductivity with an automatic, temperature-compensated conductivity meter (Hach CO150), and for pH with a standard glass electrode.
Like the measurements of soil water content, measurements of soil conductivity provided only a single-time comparison between plots. Both absolute levels of salinity and differences between microhabitats in salinity have been shown to vary seasonally at Bodega Head (Barbour et al. 1973). Levels and differences tend to peak at the end of the dry season in October and November and to be least during January and February, two of the rainiest months. As the main growing season for herbs is from November through June, we probably measured salinity during the month of the growing season when levels and differences between habitats are likely to be greatest. Our results were thus likely to overstate rather than underestimate differences between microhabitats in soil salinity.
To test the hypothesis that effects of resource availabilities on invasibility are mediated by their effects on the ability of native species to compete with non-native species, we conducted two competition experiments between the second most common native species in our field plots, the perennial grass Hordeum brachyantherum (referred to as Hordeum), and the most common non-native species, the annual grass Lolium multiflorum (referred to as Lolium), at different levels of two resources that appeared to vary between microhabitat types with different degrees of invasion, nitrogen and water. We chose the second most common native species rather than the first, which was Bromus carinatus, because the abundance of Hordeum showed a stronger relationship to patterns of invasion (Kolb 1999).
In our first experiment, intended to mimic competition on bare ground or in heavily invaded microhabitats where Hordeum was absent, the species began competing as seedlings. In a second experiment, intended to at least partly mimic competition in less heavily invaded microhabitats where adult plants of Hordeum were present, we included treatments in which 10-week-old plants of Hordeum competed with seedlings of Lolium as well as treatments in which both species competed as seedlings. As 10-week-old plants were much younger than most established plants of Hordeum would be in the field situation, results should understate the effect of life stage of Hordeum on competition with Lolium.
We collected seeds from Bodega Head in June 1998 and conducted experiments in a greenhouse at the University of Massachusetts during March to August 1999. Seeds were germinated in trays of sand. Approximately 10 days after germination, seedlings were transplanted into pots (6.4 cm in diameter by 25 cm deep) filled with acid-washed sand. For the first experiment, each species was grown in three competition treatments crossed with four resource treatments. The competition treatments were: plants grown singly (i.e. no competition, one plant per pot); monoculture (intraspecific competition, four plants of the same species per pot); and species grown together (mixed intra- and interspecific competition, two plants of each species per pot). The resource treatments were: high N (50 mg N L−1 nutrient solution); medium N (5 mg N L−1); low N (1 mg N L−1); and high N but low water (see below).
We selected the nitrogen levels so that the high and low levels would bracket the range of levels likely to occur in the grassland. We based the selection on the effects of a range of nutrient solutions on growth of a native species from the grassland, Fragaria chiloensis (Alpert 1991); the highest level caused plants of this species to grow much larger than they do in the grassland, and the lowest level supported survival but no net growth in biomass. Nitrogen was supplied as Ca(NO3)2 in a modified Hoagland's nutrient solution (Alpert 1991); concentrations of Ca2+ and of total ions were held constant in the different N treatments by appropriate additions of CaSO4. All pots were watered with nutrient solution every 10 days or when any plant showed incipient wilting; enough solution was added at each watering to flush the soil and thus help avoid any build-up of nutrients. No water was added at any other time in the low water treatment but other pots were watered with tap water whenever the soil surface of any pot became dry.
We also included treatments at medium N and low water and at low N and low water. However, plants in these treatments did not grow enough to deplete the soil moisture, meaning that water levels were higher than in the high N, low water treatment. We therefore omitted these other low water treatments from the data analysis.
The singly grown plants and the plants grown in monoculture or species mixture were arranged in two separate randomized block designs to avoid edge effects between treatments with different plant densities. Each array had 10 blocks, each with one replicate of each treatment. Pots in the first array were spaced 8 cm apart to minimize shading. Pots in the second array were placed next to each other and surrounded by a border of edge pots also each planted with four plants.
After 9 weeks (plants grown singly) or 10 weeks (other treatments, in which plants grew more slowly), plants were harvested, divided into shoots and roots, dried to constant mass at 60 °C, and weighed. As plants were grown in sand, it was relatively easy to separate their roots from the soil and from each other. A small proportion of roots became detached during harvest and were not included in estimates of biomass.
For the second experiment, we used only the competition treatment in which species were grown together and grew plants at two nitrogen levels (high and medium, as above) crossed with two life stages of Hordeum (seedlings or 10-week-old plants). The older plants had been germinated at the same time as the seedlings for the first experiment and grown in the same greenhouse. Treatments were arrayed as in the second array in the first experiment, but with 11 blocks, and harvested after 9 weeks in August 1999.
Before statistical analysis of the data, cover classes were converted to their midpoints, proportional data were transformed to the arcsine of the square root, and other data were log-transformed as needed to improve homoscedasticity and normality. We present only untransformed data in figures to best represent the actual values found in the field and the greenhouse. In cases where soil depth exceeded the length of the soil probe, effects of independent variables on soil depth were tested with Kruskal–Wallis tests. We tested effects of independent variables on the cover and species richness of natives and non-natives and on soil characteristics with analysis of variance (anova) in systat 9.0 or sas 6.12. We used one-way anova and posthoc Tukey tests to analyse the effects of distance from the ocean and of Lupinus, two-way anova and orthogonal contrasts (single degree of freedom tests, using the hypothesis function in systat) to test effects of topographic position in gullies and of gully, and a mixed anova model with area (i.e. cluster of shallow soil patches) nested within distance of area from the ocean to test effects of shallow vs. deep soil, area, and distance of area from the ocean.
To analyse the first competition experiment, we used three-way anovas to test effects of competition, species and resource availability (water or N) on total plant dry biomass in the monoculture and mixed species treatments, and two-way anovas to test effects of species and resource availability on the total mass of singly grown plants. In the mixed species treatment, we used means of the two plants of the same species in each pot in analyses. In the monoculture treatment, we used means of the mass of two randomly chosen plants per pot. Pots in which any plants died were excluded from analysis. To analyse the second competition experiment, we used a two-way anova to test the effects of N and life stage of Hordeum on the relative performance of Hordeum (mass of Hordeum per pot divided by total plant mass per pot).