John Bishir

THE FREQUENCY OF LETHAL ALLELES IN FOREST TREE POPULATIONS

In summary, by extending the Gross and Charnov (1980) model to allow for cannibalism, we have shown that the equilibrium frequency of stream males should be increased by heterocannibalistic behavior. We have also provided some empirical support for the hypothesis that the frequency ofstream males in Miyabe charr may be a mixed ESS in which the lifetime fitness of stream and lake males is equal, and that it may be slightly affected by cannibalism.

Analysis of failure time in clonally propagated plant populations

A problem originating in forest tree breeding concerns the number of clones needed in clonally propagated plantings to manage risk of failure due to an unforeseen catastrophic event. In this paper, we present a model for and analysis of time to failure for clonally propagated populations, assuming that in each year there is a chance for attack by an insect or pathogen. We develop the probability distribution of the number of years until population failure, T. A surprising finding is that in some circumstances increasing the number of clones can increase, rather than decrease, the chance of population failure. This suggests that laws, such as those current in the European Community, mandating minimum numbers of clones to be used in reforestation, may not achieve their intended effects, and that further investigation is needed to clarify the situation.

Limit theorems and a general framework for risk analysis in clonal forestry

Use of clonally propagated plantings in reforestation offers management advantages of phenotypic uniformity and high yields. Disadvantages include low genetic diversity and the possibility that the clone or clones chosen are particularly susceptible to attack by an insect or pathogen unforeseen as a problem at the time of clonal selection. In this paper, we continue consideration of the problem of choosing an optimal number of clones to minimize the risk of plantation failure. We present an analysis in which risk of failure for a plantation is represented by the probability that the proportion, S, of ramets that survive until harvest is less than or equal to a prescribed value. Our approach includes most earlier treatments as special cases. We show that the proportion S converges in distribution and, furthermore, that, under general conditions, a moderate number of clones, usually no more than 20 to 40 and often fewer, provides equivalent or better protection against catastrophic loss than does a large number of clones.

On-Bark Behavior of Dendroctonus frontalis: A Markov Chain Analysis

Tree-killing species of the Scolytidae (Coleoptera) must locate suitable hosts at least once per generation for successful reproduction. The process used to select hosts is complex, involving a sequence of steps and many possible outcomes. Because more beetles land on bark (host-find) than bore galleries (host-recognize), postlanding behaviors appear to be important in determining whether a potential host is selected. We applied Markov chain analysis to on-bark behaviors of the southern pine beetle, Dendroctonus frontalis Zimmermann, that were described by previous investigators. Predictions obtained from our analysis agree well with earlier descriptions and provide additional information not heretofore apparent. By developing chains for each sex, sexually dichotomous fighting behaviors were revealed. Testable predictions were generated for outcomes of complex interactions that occurred when tree resistance and predator density were varied. Markov chain analysis also provides a framework for future host selection studies. Use of this type of analysis requires collecting data over the entire behavioral sequence of interest, with concentration on estimating the transition probabilities among states. This approach provides results for the many possible outcomes derived from a sequence of interrelated activities. We expect that such an integrated treatment will lead to a greater understanding of important facets of scolytid host selection behavior.

Density-dependent dynamics in size-structured plant populations

Year-to-year changes in numbers of individuals in a plant community occur in response to a confluence of factors—nutrient and water availability, climatic conditions, and the type and status of individual plants. Since effects of these factors are conditioned by plant density, we wish to examine the dynamical changes in the composition of a plant community that can be explained, or at least simulated, by a model driven by density alone. We develop a model of plants classified by size or life-state in which each density vector is generated from that of the preceding time period by an appropriate matrix multiplication. Matrix models have been generally discussed by Caswell. In our formulation, matrix entries representing survival and reproduction rates are determined in each time period as implicit or explicit functions of density. This temporal plasticity in life history “parameters” produces an interaction between individual plant histories and community development. We develop a joint analysis of how plants may allocate energy among reproduction, maintenance, and growth so as to maximize “fitness”, and of the long-term dynamics of communities that contain such plants. Density affects plant access to water and to required nutrients, as well as to energy. We shall, however, focus on the latter. In this paper, we emphasize intra- or interspecific competition among perennials, though we sketch out an approach for annuals. Our discussion concerns the model itself, together with some indication, both theoretical and via computer simulations, of the possible dynamical behaviors of the system. (For instance, a species of pioneer type in which all individuals are of the same size will tend to maximize fitness at high density by devoting all energy to maintenance, and none to reproduction. At low density the optimal allocation typically is reversed, and involves either reproduction only, or a mixture of reproduction and maintenance.) Subsequent papers will enlarge on the implications of the model, and will deal more fully with populations of annuals.

Evolutionary effects of density-dependent selection in plants

The evolution of traits that affect genotypic responses to density regulated resources can be strongly affected by population dynamics in ways that are unpredictable from individual viability or reproduction potentials. Genotypes that are most efficient in utilizing energy may not always displace less efficient ones, and the evolution of energy allocation strategies may not always favour reproductive fitness because of their effects on destabilizing population growth rates. Furthermore, genetic polymorphisms in single loci that affect such traits can be maintained in populations with stable, periodic changes in population size and gene frequencies in the absence of heterozygote superiority. In fact, in the models investigated in this paper, the polymorphism is maintained, even in the absence of equilibrium genotypic frequencies.

A generalized linear learning model

Abstract The learning model of Bush and Mosteller is combined with a learning “machine,” described by Mitchie, to arrive at a generalized linear learning model. This model considers the interaction of many simple Bush-Mosteller choice situations. Subject to certain restrictions, it is proved that an organism acting according to the model will, with probability one, learn to make correct choices in each choice situation with which he is confronted. Certain implications for psychological learning theory are considered and the results of some computer simulation runs are presented.

Predicting County-Level Southern Pine Beetle Outbreaks From Neighborhood Patterns

ABSTRACT The southern pine beetle (Dendroctonus frontalis Zimmermann, Coleoptera: Curculionidae) is the most destructive insect in southern forests. States have kept county-level records on the locations of beetle outbreaks for the past 50 yr. This study determined how accurately patterns of county-level infestations in preceding years could predict infestation o ccurrence in the current year and if there were emergent patterns that correlated strongly with beetle outbreaks. A variety of methods were tested as infestation predictors, including quantification of either the exact locations of infested grid cells during one or two preceding years, or the neighborhood infestation intensity (number of infested cells in a neighborhood) in these years. The methods had similar predictive abilities, but the simpler methods performed somewhat better than the more complex ones. The factors most correlated with infestations in future years were infestation in the current year and the number of surrounding counties that were infested. Infestation history helped to predict the probability of future infestations in a region, but county-level patterns alone left much of the year-to-year variability unexplained.