Design tradeoffs for trend assessment in aquatic biological monitoring programs

Assessments of long-term (multiyear) temporal trends in biological monitoring programs are generally undertaken without an adequate understanding of the temporal variability of biological communities. When the sources and levels of variability are unknown, managers cannot make informed choices in sampling design to achieve monitoring goals in a cost-effective manner. We evaluated different trend sampling designs by estimating components of both short- and long-term variability in biological indicators of water quality in streams. Invertebrate samples were collected from 32 sites—9 urban, 6 agricultural, and 17 relatively undisturbed (reference) streams—distributed throughout the United States. Between 5 and 12 yearly samples were collected at each site during the period 1993–2008, plus 2 samples within a 10-week index period during either 2007 or 2008. These data allowed calculation of four sources of variance for invertebrate indicators: among sites, among years within sites, interaction among sites and years (site-specific annual variation), and among samples collected within an index period at a site (residual). When estimates of these variance components are known, changes to sampling design can be made to improve trend detection. Design modifications that result in the ability to detect the smallest trend with the fewest samples are, from most to least effective: (1) increasing the number of years in the sampling period (duration of the monitoring program), (2) decreasing the interval between samples, and (3) increasing the number of repeat-visit samples per year (within an index period). This order of improvement in trend detection, which achieves the greatest gain for the fewest samples, is the same whether trends are assessed at an individual site or an average trend of multiple sites. In multiple-site surveys, increasing the number of sites has an effect similar to that of decreasing the sampling interval; the benefit of adding sites is greater when a new set of different sites is selected for each sampling effort than when the same sites are sampled each time. Understanding variance components of the ecological attributes of interest can lead to more cost-effective monitoring designs to detect trends.