Spatial and annual variation to define the normal range of biological endpoints: an example with biomarkers in perch

A signal from a biological endpoint can be considered evidence of environmental impact only if it is known that similar signals are unlikely to occur naturally. Using multiple reference sites, the normal range can be estimated. This can be defined as the span that includes 95% of the reference sites. If an investigated site is outside this range, it is interpreted as evidence of impact. The choice of reference sites is thus important for the outcome of the study. Here, biomarker levels in perch (Perca fluviatilis) at a potentially impacted site were compared to different types of reference data. The potentially impacted site was located close to a densely populated area. Four reference sites were located in relatively undisturbed areas on the Swedish Baltic coast. Furthermore, historical data from one of the reference sites were included. The present study showed that multiple reference sites are needed to avoid the risk of interpreting natural variation as impact. The number of reference sites needed depends on the desired level of statistical power. An alternative to multiple reference sites may be to use historical data to estimate the natural variation. However, historical data can include temporal variation due to factors that may not be relevant for the hypothesis that is tested, e.g., climatic variation or changed background levels of pollution.