Sampling spatial and temporal variation in soil nitrogen availability.

There are few studies in natural ecosystems on how spatial maps of soil attributes change within a growing season. In part, this is due to methodological difficulties associated with sampling the same spatial locations repeatedly over time. We describe the use of ion exchange membrane spikes, a relatively nondestructive way to measure how soil resources at a given point in space fluctuate over time. We used this method to examine spatial patterns of soil ammonium (NH+4) and nitrate (NO-3) availability in a mid-successional coastal dune for four periods of time during the growing season. For a single point in time, we also measured soil NH+4 and NO-3 concentrations from soil cores collected from the mid-successional dune and from an early and a late successional dune. Soil nitrogen concentrations were low and highly variable in dunes of all ages. Mean NH+4 and NO-3 concentrations increased with the age of the dune, whereas coefficients of variation for NH+4 and NO-3 concentrations decreased with the age of the dune. Soil NO-3 concentration showed strong spatial structure, but soil NH+4 concentration was not spatially structured. Plant-available NH+4 and NO-3 showed relatively little spatial structure: only NO-3 availability in the second sampling period had significant patch structure. Spatial maps of NH+4 and NO-3 availability changed greatly over time, and there were few significant correlations among soil nitrogen availability at different points in time. NO-3 availability in the second sampling period was highly correlated (r = 0.90) with the initial soil NO-3 concentrations, providing some evidence that patches of plant-available NO-3 may reappear at the same spatial locations at irregular points in time.

REINFORCING SELECTION IS EFFECTIVE UNDER A RELATIVELY BROAD SET OF CONDITIONS IN A MOSAIC HYBRID ZONE.

Many hybrid zones have a mosaic structure, yet we know of no theoretical work that examines the impact of mosaicism on the outcome of evolution. We developed a computer simulation model designed to test whether the outcome of reinforcing selection differs in a mosaic and a clinal hybrid zone. Our model was a one-dimensional stepping-stone model. The mosaic and clinal hybrid zones that we modeled were, respectively, a mosaic maintained by differential fitness of the interacting taxa in patchy habitats and a tension zone. We modeled changes in gene frequency at two biallelic loci, A and B. Hybrids at the A locus were selected against. An allele at the B locus caused assortative mating at the A locus, which promoted reinforcement; there was a selective cost to this allele. In a mosaic hybrid zone, spatial variation in the fitness of A-locus homozygotes in different patches caused gene and genotype frequencies at the A and B loci to differ greatly from those in a tension zone. Compared to a tension zone, a mosaic hybrid zone had a broader region in which hybrids could be formed and, thus, a broader region in which the assortative-mating allele provided a net selective advantage (via decreased production of the less fit A-locus hybrids). This caused the assortative-mating allele to be favored under a broader set of conditions in a mosaic hybrid zone than in a tension zone. In mosaic and tension hybrid zones, both low and high levels of migration could prevent the establishment of the allele that promoted reinforcement, but the allele could establish under a wider range of migration rates in a mosaic than in a tension zone. In a tension zone, both low and high levels of selection against A-locus hybrids could prevent the establishment of the assortative-mating allele. In a mosaic hybrid zone, the assortative-mating allele established under lower levels of selection against hybrids than in a tension zone, and high levels of selection did not impede the establishment of this allele. Overall, our work illustrates how the structure of a hybrid zone can alter the outcome of an important evolutionary process, in this case, reinforcement.

CONSPECIFIC SPERM PRECEDENCE IS AN EFFECTIVE BARRIER TO HYBRIDIZATION BETWEEN CLOSELY RELATED SPECIES.

Conspecific sperm precedence is widespread in animals, appears to evolve rapidly, and is thought to have the potential to prevent hybridization between closely related species. However, to date no study has tested the isolating potential of such a barrier in mixed populations of two taxa under conditions in which other potential barriers to gene flow are controlled for or are prevented from operating. We tested the isolating potential of conspecific sperm precedence in the ground crickets Allonemobius fasciatus and A. socius in population cage experiments in which the frequency of the two species was varied. Despite the observation of abundant interspecific matings, the proportions of hybrid progeny were low and differed statistically from the proportions expected in the absence of conspecific sperm precedence. The results demonstrate that conspecific sperm precedence can severely limit gene flow between closely related species, even when one species is less abundant than the other.

Long-term suppression of insect herbivores increases the production and growth of Solidago altissima rhizomes.

Although insect herbivores have many well documented effects on plant performance, there are few studies that assess the impact of above-ground herbivory on below-ground plant growth. For a seven year period in which no large-scale herbivore outbreaks occurred, a broad spectrum insecticide was utilized to suppress herbivorous insects in a natural community dominated by Solidago altissima. Ramet heights, rhizome lengths, rhizome biomass, and the number of daughter rhizomes all were lower in the control plots than in the insecticidetreated plots. These effects should lead to a decrease in the fitness of genets in the control plots relative to the fitness of genets in the insecticide-treated plots. We also found that ramets in the control plots appear to have compensated for herbivory: the ratio of rhizome length to rhizome biomass was greatest in the control plots, which indicates that clones moved farther per unit biomass in these plots than in the insecticide-treated plots. Clonal growth models show that this shift in allocation patterns greatly reduced the magnitude of treatment differences in long-term clonal displacements.Previous work has shown, and this study verified, that clonal growth in S. altissima is well represented by random-walk and diffusion models. Therefore, we used these models to examine possible treatment differences in rates of clonal expansion. Although rhizome lengths were greater in the insecticide-treated plots, results from the models suggest that our treatments had little impact on the short- and long-term displacement of S. altissima ramets from a point of origin. This occurred because S. altissima ramets backtrack often, and thus, treatment differences in net displacements are less pronounced than treatment differences in rhizome lengths.

Patterns of Solidago altissima ramet growth and mortality: the role of below-ground ramet connections.

For the rhizomatous perennial, Solidago altissima, I identified clonal fragments in the field, mapped ramet spatial locations, and documented patterns of ramet recruitment, growth, and mortality. Parent ramet size influenced the size and number of daughter ramets produced, and small ramets had lower survivorship and fecundity than large ramets. Similarly, small rhizomes tended to develop into small ramets, and ramets that survived to produce daughter ramets had longer parent-daughter rhizome connections than ramets that did not survive. In addition, most ramets that died during the growing season were connected to (genetically identical) ramets that persisted. There were large size inequalities among rhizomes, ramets, and clonal fragments. Inequalities in the size of ramets increased during the early part of the growing season, then decreased at the end of the season; similar patterns were observed for the growth of clonal fragments. In both instances, the decrease in size inequality could be attributed to the mortality of small individuals (ramets or clonal fragments). I found little evidence that ramet size hierarchies were structured by intraspecific competition. For example, path analyses and randomization tests indicated that size variation among S. altissima ramets was influenced little by the size of their near neighbors (but was influenced by parent size and rhizome size). In addition, within-season variation for the relative size and growth rate of individual ramets led to poor correlations between early and final ramet size; this result suggests that there was no stable hierarchy of dominant and suppressed ramets. I discuss implications of my results for contrasting interpretations of clonal plant population dynamics.

GROWTH IN MEDEOLA VIRGINIANA CLONES. II. STOCHASTIC SIMULATION OF VEGETATIVE SPREAD.

Medeola virginiana, a forest understory herb, has been hypothesized to minimize competition among ramets by positioning ramets regularly in space (Bell, 1974). We have constructed a computer model that simulates vegetative spread in Medeola by incorporating the variation in growth form found in natural populations. This stochastic model, based on data in Cook (1988), includes variation in the number and type of daughter ramets, the distance between parent and daughter ramets, and the angle of divergence between the lines of growth of the parent and daughter rhizomes. Over a wide range of initial ramet densities, initial degrees of ramet dispersion, and intensities of ramet mortality, our simulations indicate that the growth pattern of M. virginiana leads to aggregated distributions of ramets with overlapping leaf canopies. The observed aggregation of ramets may result from the tendency for directional but variable growth in clones. Such variable growth calls into question adaptive interpretations based on deterministic models that utilize the mean values of growth parameters but ignore the highly variable nature of clonal growth.