A sequence of changes in vital rates observed as populations approach maximal levels has been used as the basis for a “paradigm” for population analysis. Previous work indicates that early survival decreases first, followed by lower reproductive rates; ultimately, adult female survival may decrease. “Sensitivity” of population growth rates, as measured by partial derivatives of an approximation to the Lotka‐Leslie model, appears to follow the same sequence, suggesting that population regulation may follow that sequence. This may imply some evolutionary significance in the sequence. Thus, it may be possible to assess population status by measuring the vital rates, as shown in a number of examples reported here. Measuring vital rates in the field is subject to a variety of biases; hence, an analysis should include direct estimates of population trend. In the absence of complete data, suitable trend data might be used to estimate missing rates. Bootstrapping provides a simple way to obtain confidence intervals, and the delta method can be used to obtain components of variance and thus improve sampling. Various methods for studying trend are given, with examples and simple statistical tests.

Field experiments in ecological and environmental research usually do not meet the criteria for modern experimental design. Subsampling is often mistakenly substituted for true replication, and sample sizes are too small for adequate power in tests of significance. In many cases, field—study objectives may be better served by various kinds of sampling procedures, even though the resulting inferences will be weaker than those obtainable through controlled experimentation. The present paper provides a classification and description of methods for designing environmental studies, with emphasis on techniques as yet little used in ecology. Eight categories of techniques for field studies are defined in terms of the nature of control exerted by the observer, by the presence or absence of a perturbation, and by the domain of study. The first two categories include classical experimental approaches, replicated and unreplicated. Sampling for modelling provides efficient designs for estimating parameters in a specified model. Intervention analysis measures the effect of a known perturbation in a time series. Observational studies contrast selected groups from a population, while analytical sampling provides comparisons over the entire population. Descriptive survey sampling estimates means or totals over an entire population, while sampling for pattern deals with spatial patterns over a selected region. We propose that the statistical concept of a "superpopulation" may be useful in ecology, and that it may be desirable to approach community and ecosystem studies in a sampling framework, with experimentation used for a fairly narrow range of subsidiary investigations. Much more attention to processes for drawing inferences about cause and effect is needed, in any case.

No abstract

Such fluctuations may well be associated with density-independent factors.

No abstract