CADDIS Volume 3: Examples & Applications
Analytical techniques used:
Type of evidence supported :
Stressor-Response from the Field
We would like determine whether water quality variables in Long Creek, Maine (U.S. EPA 2007) are associated with three observed changes in the aquatic invertebrate community relative to the reference stream: a decrease in Ephemeroptera, Plecoptera and Trichoptera (EPT) richness; an increase in percent non-insect taxa; and a shift towards increased pollution tolerance, estimated using Hilsenhoff's Biotic Index (HBI) (Hilsenhoff 1987, 1988).
In this example, we present analyses relevant to two candidate causes, ionic strength (measured using specific conductivity), and zinc. If specific conductivity (or zinc) is not associated with the biological responses in the expected direction, this evidence would weaken the argument for ionic strength (or zinc) being a cause of the observed biological changes. Conversely, if specific conductivity (or zinc) is associated with the biological responses in the expected direction, this evidence would somewhat support the argument that ionic strength (or zinc) is the cause of the observed changes. These associations can provide only weak support for a causal argument because other stressors may be correlated with increased conductivity (or zinc), and are not controlled for in this analysis. For this reason, it is important to conduct this analysis for as many of the candidate causes as possible.
Biological and water chemistry data from 8 sites along Long Creek and a similar but unimpaired reference stream, are used in this example.
Biological metrics were calculated from macroinvertebrate rockbag samples deployed throughout the study area beginning August 5-6, 1999, for a period of 32 days, following standard Maine Department of Environmental Protection (MEDEP) protocol (Davies and Tsomides 2002).
Water chemistry measurements of conductivity and zinc were made from baseflow water samples collected by MEDEP on three days in August 2000. Methods and analyses are described in MEDEP (2002). Here, the analysts assume that the differences in the collection dates for biological samples (1999) and for water chemistry samples (2000) did not affect observed relationships. Ideally, additional data would be collected as a follow-up to validate this assumption (see the discussion under Getting Started: Matching Data).
Analysis and results
The data were analyzed using scatter plots (Figure 1). The project team interpreted the scatter plots by looking for linear and curvilinear trends in the data. Because only one data point from each site was available, the plots were not used to make judgments about individual sites or stream reaches. Instead, the plots were used to characterize trends across the two watersheds.
The visual interpretation of the scatterplots was supplemented with correlation coefficients (Table 1). Correlation coefficients were not evaluated for significance because of the small sample size and pseudo-replication of sites. Rather, consistent correlations of relatively large magnitude for all three biological responses were considered by the analysts to provide some support for ionic strength as a candidate cause. When evaluating this evidence, it is worth noting again that both analyses hinge on the assumption that samples of water chemistry taken in August 2000 are similar to exposures experienced by organisms in August 1999.
Tools for generating both scatter plots and correlation coefficients are available in CADStat.
How do I score this evidence?
Associations between specific conductivity and all three biological responses were apparent and in the expected direction. We would score this evidence as + for each of the biological responses.
There were no clear associations between zinc and any of the three biological responses. We would score this evidence as - for each of the biological responses.