Evaluating disturbance impacts on fauna

Proponents of a development should be able to demonstrate that any impacts on fauna or fauna assemblages in uncleared and adjacent areas have been kept to predetermined minimal levels. Ideally, appropriate triggers are provided with prescribed remedial actions when these triggers are reached. For some developments there is a ministerial condition which requires that potential impacts on the fauna in adjacent areas are monitored to ensure that they remain below an acceptable level, or ensure that management strategies put in place to protect a conservation significant species or ecosystem are effective in achieving their purpose.

Spatial and temporal variability in vertebrate fauna assemblages can be significant which adds a level of complexity to establishing effective evaluation protocols (How 1998, Stewart et al. 2001, How and Cooper 2002, Petchey et al. 2002, Thompson et al. 2003, Maron et al. 2005, Thompson and Thompson 2005). To detect change, particularly change that is a result of a specific disturbance, it is important that substantial and robust baseline data are collected prior to the disturbance. This will often mean more extensive pre-development surveys are undertaken than might otherwise be the case.

For quantitative evaluations there are six broad designs:

  • Single data point, single site, before and after;
  • Multiple data points (e.g. time series), single site, before and after;
  • Single data point at control and impact sites, before and after;
  • Multiple data points (e.g. time series) at single control and impact sites, before and after;
  • Single data point at multiple control and impact sites, before and after; and
  • Multiple data points (e.g. time series) at multiple control and impact sites, before and after.

A single data point, at a single site, collected before and after design is very common in evaluating the impacts of a development on the fauna assemblage, but many are seriously flawed because there is no way of distinguishing a change due to an impact compared with natural spatial and temporal variations in the fauna assemblage. They are, however, useful when a criterion is provided. For example, a criterion might be the number of adult Malleefowl in the 200ha remnant vegetation adjacent to a mine site will not fall below 25 individuals. This criterion is based on an implied relationship between the potential impact (e.g. mining) and Malleefowl numbers, which may or may not be true. However, once accepted by the proponent as an evaluation criterion, then there is an obligation to collect adequate data periodically to evaluate performance against the criterion.

Quasi-experimental approaches are normally based on time series data for single ‘impact’ and ‘control’ sites. In this situation, the evaluation is attempting to identify changes from a ‘normal’ or ‘established’ pattern prior to the event. In this situation, there has been an attempt to address temporal variations in the data set, but the design does not address spatial variability and ignores other potential reasons for differences between before and after scores.

Experimental designs attempt to control extraneous effects and demonstrate a cause and effect relationship. This is most often done by randomly assigning survey sites in control and impact categories and excluding all other variables that are likely to influence changes in the fauna. In most circumstances this is not feasible; as impact sites are predetermined and controlling other variables that might affect changes in fauna assemblages (e.g. drought, predators) are difficult and often not practical.

Issues with single site and quasi-experimental designs

To illustrate potential issues with single site and quasi-experimental designs, we will consider hypothetical monitoring data for a population of Malleefowl (L. ocellata) in a 500ha remnant plot of vegetation adjacent to a mine site in the mid-west of Western Australia. These hypothetical Malleefowl monitoring data were collected annually. In the figures below (frames B-D and E-H) the dotted vertical line in 2018 indicates the beginning of the disturbance impact associated with the mine. In frame H there were two disturbances (2018 and 2023). The solid line in frame A indicates the temporal variations in Malleefowl numbers without a disturbance. The solid line in frames B-H and the small dots in frame H represent data collected from an ‘impact’ site. The dashed line in frames C and D represents data that were not collected, but what the data would have looked like had it been collected. The dashed line in frames E, F and G represent data from a ‘control’ site. Frame H is time series data for two impact sites with the disturbance occurring at two different times (dotted vertical lines).

Evaluation 1 Evaluation 2

Frame A indicates the results of Malleefowl monitoring data between 2009 and 2028 had the area remained undisturbed. What is evident is that the population fluctuated annually, with some evidence of a cycle.

Frame B would suggest there was an overall decline in the number of Malleefowl after the disturbance. However, this decline would have been difficult to detect until at least three years after the decline in the number of Malleefowl. It would also be difficult to be sure that the decline was not due to natural temporal variations and not a result of the disturbance until at least four years after the disturbance event. This time lag might be too late to reverse the loss.

In frames C and D, the monitoring data prior to 2018 (dashed line) are unknown. It would therefore be difficult to know whether the initial increase after the disturbance in 2018 (frame C and D) was the result of natural temporal variations in Malleefowl numbers or the result of a disturbance. Any conclusions drawn from these data in frame C about potential impacts 3-6 years after the disturbance would be different from those drawn 7-9 years after the disturbance. The initial conclusion might be the disturbance impact has caused a reduction in Malleefowl numbers; however, subsequent years of monitoring (2023-2026) might suggest Malleefowl had accommodated this disturbance and their numbers had increased, when the real situation is that data represents natural temporal variations in the population and are not the result of a disturbance impact.

The data post disturbance in frame D indicates Malleefowl numbers have declined. However, there had been a progressive decline in the pre-disturbance data, so the mine has not impacted on Malleefowl. The single data point in 2017 prior to the disturbance in frames C and D is typical of fauna evaluation programs in Western Australia. Seldom are sufficient data available to indicate pre-event natural variations.

Frames E-G show control and impact data for an extended period before and after the disturbance impact. Frame E shows the decline in the number of Malleefowl after the disturbance which contrasts with the number of birds in the control site. This could indicate the disturbance was impacting on Malleefowl numbers, presuming there were adequate data to show a statistical difference.

Frame F shows the progressive decline in the population of Malleefowl pre- and post-disturbance suggesting any impacts of the disturbance on Malleefowl numbers are either small or non-existent. A substantial number of samples would be required to demonstrate a difference between the control and impact sites post disturbance.

Frame G indicates the population in the control site has increased and the population in the site impacted by the disturbance has decreased providing the strongest case to demonstrate a disturbance impact on the Malleefowl population.

In frame H, there is no control site and two monitoring sites are impacted by the disturbance at different times (vertical dotted lines). In both monitoring sites the population of Malleefowl has declined after each impact disturbance (2018 and 2023), suggesting a link between the disturbance and a reduction in Malleefowl numbers. What is evident from these examples is that a single data point prior to a disturbance and a single group of data (e.g. no control) provided the least useful information, and can result in a misinterpretation of the real situation, as it is difficult to separate natural temporal variations from ‘impacts’. Comparable control and impact data provided a better appreciation of the situation, and time series data before the impact of a disturbance occurs is useful in determining temporal variations.

The best possible design for evaluating impacts on fauna is the Before-After-Control-Impact-Paired-Series (BACIPS). In this situation we are able to compare impact and control sites before the disturbance and have sufficient data to understand natural variations (i.e. an estimate of variance). This measure of variance enables the proponent to define confidence limits, which are used to determine whether differences are significant or due to chance occurrences. The larger the time series data set before the disturbance the more robust the understanding of natural temporal variations in undisturbed conditions.

A primary concern for most fauna impact assessments is finding comparable impact and control sites. Ideally, there are multiple impact and control sites, and variability (i.e. temporal and spatial) among all sites is low. However, there are circumstances where it is impossible to have more than one impact and one control site (e.g. when measuring impacts on stream fauna by a discharge and sampling above and below the discharge). In this circumstance, Stewart-Oaten et al. (1986) suggested time series data be used, with sampling times being used as replicates (i.e. pseudo-replication). In the stream discharge situation, data from before and after, above and below the discharge would be used and the variance would come from multiple sampling (Stewart-Oaten et al. 1986). Underwood (1991, 1994) provided multiple examples to demonstrate that a single control site to contrast with a single impact site was in most circumstances unsatisfactory.

Underwood (1994) described some of the problems associated with the more widely used procedures to detect the effect of environmental disturbances due to human activities. The first, and often the most difficult to demonstrate, is causality when a difference is detected. He also suggested issues linked with temporal and spatial variability were of concern. Stewart-Oaten et al. (1986) discussed the need for temporal replication, and suggested comparative data sets would be improved by randomising the sampling periods. As discussed in another post, fauna assemblages are rarely evenly distributed across a habitat type, and relative abundance also shows temporal variations providing an interaction between temporal and spatial variability (Stewart-Oaten et al. 1986, Hewitt et al. 2001). However, although desirable, control and impact sites can rarely be randomly assigned in a field situation. In addition, there is the situation where a relatively stable environment is impacted on resulting in a series of pulses post disturbance (Underwood 1991, 1994). In such a circumstance, the timing of the post-event sampling will largely determine the result when a before-after analysis approach is adopted.

Although best practice would require a before-after-control-impact-paired series (BACIPS) design be used this is often not achievable. In many circumstances, the evaluation is focussed on potential impacts on fauna in areas adjacent to the development, with the consequence that ‘impact’ sites must be located adjacent to the development. This restricts the range of potential impact sites. If the intention is to obtain paired control and impact sites, then similar fauna habitats must be located in both impact and control areas. It could be anticipated the sites used would also be representative of the general habitat in the area and this would further restrict the choice of sampling sites. An example of this problem occurs where a proponent wants to mine a Banded Iron Formation (BIF) in a sand plain and there are no adjacent BIFs that can be used as control sites. The proponent therefore has the choice of having no controls or assigning the control sites to different habitat types in adjacent areas.

If the potential impacts of the fauna were variables such as noise, dust, vibrations, vehicle movements, etc, then it may be reasonable to presume that these impacts were similar in all fauna habitat types, and as a consequence, it is not necessary to measure potential impacts in all habitat types.


Hewitt, J. E., S. E. Thrush, and V. J. Cummings. 2001. Assessing environmental impacts: effects of spatial and temporal variability at likely impact scales. Ecological Applications 11:1502-1516.

How, R. A. 1998. Long-term sampling of a herpetofauna assemblage on an isolated urban bushland remnant, Bold Park, Perth. Journal of the Royal Society of Western Australia 81:143-148.

How, R. A. and N. K. Cooper. 2002. Terrestrial small mammals of the Abydos Plain in the north-eastern Pilbara, Western Australia. Journal of the Royal Society of Western Australia 85:71-82.

Maron, M., A. Lill, D. M. Watson, and R. Mac Nally. 2005. Temporal variation in bird assemblages: How representative is a one-year snapshot? Austral Ecology 30:383-394.

Petchey, O. L., T. Casey, L. Jiang, P. T. McPherson, and J. Price. 2002. Species richness, environmental fluctuations, and temporal change in total community biomass. Oikos 99:231-240.

Stewart-Oaten, A., W. W. Murdoch, and K. S. Parker. 1986. Environmental impact assessment: “Pseudoreplication” in time? Ecology 67:929-940.

Stewart, M., G. Brent, D. Giurco, and J. Petrie. 2001. Decision making for sustainability: The case of minerals development in Australia. In the 26th Annual Minerals Council of Australia Environmental Workshop. Minerals Council of Australia, Adelaide.

Thompson, G. G., S. A. Thompson, P. C. Withers, and E. R. Pianka. 2003. Diversity and abundance of pit-trapped reptiles in Australian arid and mesic habitats: Biodiversity for environmental impact assessments. Pacific Conservation Biology 9:120-135.

Thompson, S. A. and G. G. Thompson. 2005. Temporal variation in reptile assemblages in the Goldfields of Western Australia. Journal of the Royal Society of Western Australia 88:25-36.

Underwood, A. J. 1991. Beyond BACI: Experimental designs for detecting human environmental impacts on temporal variations in natural populations. Australian Journal of Marine and Freshwater Research 42:569-587.

Underwood, A. J. 1994. On Beyond BACI: sampling designs that might reliably detect environmental disturbances. Ecological Applications 4:3-15.

Print Friendly


Got something to say?