Introduction

The overall variability in observations from an ecological study generally is composed of multiple components of random error. The statistical theory of variance component estimation is well established (Searle et al. 1992; see also Chapters 7, 9). What is not well established is the routine use of preliminary surveys to collect information on sampling costs, expected response levels, and the magnitude of error sources (Box 8.1). Preliminary surveys should be an integrated component of every monitoring program which has as its objective more than simply long-term employment for those involved. In ecological studies, natural variation is typically too large and the sampling techniques too imprecise to leave study design to chance.

Sample-size calculations require knowledge of both the nature and magnitude of error sources. Add to this financial limitations, and the only prospect for an efficient monitoring program is often design optimization. Most monitoring programs will have sample sizes that are multidimensional – for example, a number of samples within a site and a number of sites within the landscape. Optimal allocation based on cost functions and variance component estimates can be used to determine the best allocation of survey effort. Design optimization is also useful in identifying discrepancies between desired study performance and budget that must be reconciled if a study is to be effective. All this, however, begins with variance component estimation.