| Abstract: | There is widespread application of indicators to the assessment of environmental condition of streams. These indicators are intended for use by managers in making various comparative and absolute assessments and often have a role in resource allocation and performance assessment. Therefore, the problem of formally defining confidence in the results is important but difficult because the sampling strategies used are commonly based on a compromise between the requirements of statistical rigour and the pragmatic issues of access and resources. It is rare to see this compromise explicitly considered and consequently there is seldom quantification of the uncertainty that could affect the confidence a manager has in an indicator. In this paper, we present a method for quantitatively assessing the tradeoffs between sampling density and uncertainty in meeting various monitoring objectives. Assessments using judgement‐based representative reaches are shown to be unreliable; instead a sampling approach is recommended based on the random selection of measuring sites. A detailed dataset was collected along two streams in Victoria, Australia, and the effect of sampling density was assessed by subsampling from this dataset with precision related to the number of sites assessed per reach length and the intensity of the sampling at each site. The sampling scheme to achieve a given precision is shown to depend on the monitoring objective. In particular, three objectives were considered: (1) making a baseline assessment of current condition; (2) change detection; and (3) detection of a critical threshold in condition. Change detection is shown to be more demanding than assessing baseline condition with additional sampling effort required to achieve the same precision. Sampling to detect a critical threshold depends on nominating acceptable values of Type I and II error and the size of the effect to be detected. Copyright © 2006 John Wiley & Sons, Ltd. |
