Biological assessment of degraded urban streams

Walsh (1997) showed that, for quantitatively sampled communities from degraded urban streams, the power of multivariate analyses to detect differences between streams was maximised for minimum processing effort by a conditional subsampling method of a fixed percentage (10% for Melbourne metropolitan streams), followed by a subsample to a fixed count of individuals (300 individuals for Melbourne). This processing effort is considerably greater than routinely used by rapid biological assessment techniques.  It is likely, that because of the more uneven nature of community composition in degraded streams (communities tend to be numerically dominated by a few hardy taxa), that a precise estimate of community composition requires greater processing effort than for more even, less degraded communities.

Software is available for simulated subsampling and the effects of subsampling are pursued further in an associated discussion document.

Walsh et al. (2001, using rapid bioassessment methods but with the Walsh, 1997, conditional subsampling method) concluded that community compositional patterns among metropolitan Melbourne sites were uniformly depauperate, so that compositional patterns related to environmental gradients among metropolitan sites were not detected.  In contrast, Walsh et al. (in review) found a strong gradient among metropolitan communities that was well explained by sediment zinc and sediment lead concentrations.  The gradient was, at best, weakly detectable using data from separate habitats and separate seasons, but became distinct as samples were combined.  The taxa that were the strongest discriminators were all (but one) taxa that occurred in very low abundances in one or a few low-metals sites, and not at all in high-metals sites.  The 10% or 300-count conditional subsampling method was therefore apparently insufficient to detect community patterns among degraded streams using rapid bio-assessment methods, because it was insufficient to adequately sample these non-abundant taxa.

Walsh et al. (in review) also found that the metals-related gradient was only detectable when the data set included the large set of less degraded, hinterland sites (i.e. when the analysis was conducted with just urban sites, the gradient was not detected).  This was because the taxa that discriminated among high-metals and low-metals sites, although non-abundant and uncommon among metropolitan sites, were common or abundant among hinterland sites.  The presence of those taxa in low-metals sites was enough to make hinterland communities more similar to low-metals communities than to the high-metals communities.  In the absence of hinterland communities, those taxa only produced weak multivariate patterns.

This finding illustrates the importance of using a large regional dataset with best possible reference sites, when seeking to detect differences among severely degraded communities.  This contradicts the premise behind the urban AUSRIVAS project that sought to build a predictive model based on less-than-pristine hinterland sites.

Further reading

• Walsh, C. J. (1997). A multivariate method for determining optimal subsample size in the analysis of macroinvertebrate samples. Marine and Freshwater Research 48(3), 241-8.
• Walsh, C.J., Sharpe, A.K., Breen, P.F. & Sonneman, J.A. (2001). Effects of urbanization on streams of the Melbourne region, Victoria, Australia. I. Benthic macroinvertebrate communities. Freshwater Biology 46, 535-551. (abstract)
• Walsh, C.J., Sharpe, A.K., Breen, P.F. & Sonneman, J.A. (in review). Detecting gradients among degraded urban streams using macroinvertebrates: the importance of sampling effort, rare taxa, and reference sites.

More information

Dr Chris Walsh,
CRC for Freshwater Ecology, Water Studies Centre,
Monash University
PO Box 23 Clayton Vic. 3800, Australia