Before/After, Control/Impact sampling design

Because they are easily studied, rocky shores provide a means of assessing changes which may reflect effects in other marine communities. This principle has been used by OPRU to monitor the impacts of operations at the Sullom Voe oil terminal in Shetland. More recently several publications have appeared which have developed a solid scientific framework for impact assessment (e.g. Stewart-Oaten et al., 1986; Underwood 1991; 1992; 1995). When new activities begin within or close to SACs, any potential impacts on conservation features should be assessed.

The abundance of almost any species will fluctuate over time. Abundances will also fluctuate from place to place. Perturbations of human origin may be judged to have had an impact when they alter the scale of these fluctuations over space and time. In many cases an impact results in changes in the mean abundance of species. For example, the average number of dogwhelks close to an effluent outflow might be low compared with other areas and to the same area before the installation of the outflow. It is also possible that an impact could alter the magnitude (variance) of fluctuations without affecting the mean abundance (Underwood, 1991). In order to identify the presence of an impact with a high degree of statistical certainty, it is necessary to establish a link between the presence of the factor causing the impact and the observed fluctuations in abundance of target species. The abundance should have changed after the appearance of the supposed impact, in a way which is not predicted by changes at other similar sites.

The BACI (Before/After, Control/Impact) design was proposed by Bernstein and Zalinski (1983) and Stewart-Oaten et al (1986) as a means of assessing impacts. Measures of species abundance would be taken at two sites (the site of the putative impact and a similar, control, site where no impact was expected) on several occasions both before and after the onset of the putative impact. Repeated measurements allow a determination of whether observed changes at the site of the putative impact are part of the pre-existing cycle of change. The control site is intended to show whether any observed disruption in this cycle is part of a wider effect not due to the putative impact. The use of a single control led to criticism of this method (Underwood, 1992). A natural change at the control site which was coincidentally similar to that caused by the impact at the other site could lead to the impact going undetected. Alternatively, a change at the control site from before to after the onset of the putative impact while the other site remained unchanged, could result in an impact being diagnosed where there was none. The solution, proposed by Underwood (1992), is to use several randomly-selected control sites. Thus the effects of global trends can be separated from those of natural fluctuations within individual sites. Further details on the design and analysis of this type of sampling programme can be found in Underwood (1991; 1992).

Next Section                     References