Capturing expert uncertainty in spatial cumulative impact assessments

What scientific problem does our paper deal with?

There are many ways in which human activities can negatively affect marine environments. We use the oceans to support industry (e.g. fishing, shipping, tourism) as well as for recreation (e.g. boating, fishing, diving, beach activities). The things we do on land can also impact marine systems, often in ways that can be difficult to see and control (e.g. land-use change, coastal development, run-off of nutrients and pollutants). On top of all these localised impacts, global climate change is affecting marine environments by increasing water temperatures, causing ocean acidification and increasing the frequency and severity of storms and heat waves.

Marine ecosystems are often exposed to many human activities at the same time – they do not occur in isolation. For example, there may be recreational activities occurring in the same area as commercial fishing and shipping, with this area also affected by climate change. We know that each of these activities has the potential to cause stress to marine ecosystems, and that this impact is likely to be made worse in areas where multiple human activities occur together. We also know that different marine ecosystems will be affected in different ways by the same human activity. For example, we know that seagrass meadows are vulnerable to nutrient inputs and physical disturbance, but that deeper seabed habitats are less affected by these activities. Unfortunately, we do not have enough field and lab data to test for the impact of each different human activity on all of the marine ecosystems, or what their combined effect (cumulative impact) will be. This lack of data is a real problem for ecologists in general (and for natural resource managers). It is expensive and time consuming to collect field data from multiple ecosystems and to carry out large-scale experiments that can accurately represent how entire marine systems respond to multiple human impacts.

Because of this lack of data, scientists have developed methods to try and assess cumulative impacts to marine ecosystems based-on the best-available evidence, which involves using expert opinion. The approach combines data on where different human activities occur and how often (in the form of maps), with expert opinion about the effect of each of these activities on the marine ecosystems they occur within. The experts (such as scientists) are asked to provide scores for the impact of human activities on different marine ecosystems, as it is likely some ecosystems will be more vulnerable to certain activities than others. This process allows us to extract useful information from the experts, who have a wealth of knowledge locked up in their brains! However, the experts can be uncertain about the impact of human activities on marine ecosystems…….

How does our work help to solve the problem?

We were interested in measuring how much expert uncertainty can affect cumulative impact assessments (hint: a lot!) and whether there was a straight-forward way to deal with this source of uncertainty. Ultimately, we wanted to support marine management decisions by developing a method that is transparent about the effect of expert uncertainty and can highlight the most certain results.

We did this by asking the experts to tell us what they thought was the ‘most-likely’ impact of each human activity on specific marine ecosystems in Spencer Gulf (South Australia). However (crucially), we also asked the experts to tell us how certain they were about the impacts. This meant that we didn’t just get a single score for the impact of each human activity, but instead we got a range of scores that reflected what the expects thought was plausible, as well as reflecting their level of uncertainty. Sounds simple right? Well, it is! You can read more about our expert survey methods in our open access paper from last year, led by University of Adelaide researcher Dr Zoe Doubleday.

In our latest paper, we further developed these methods, by using the expert score ranges, and spatial data on 32 different human activities occurring in Spencer Gulf, South Australia (see figure below) to generate cumulative impact maps that account for expert’s uncertainty.

Figure to the left shows some examples of the 32 maps we made of the footprint and intensity of human activities in Spencer Gulf (South Australia).

Because we got a range of impact scores from the experts, we were able to produce a range of maps that reflected the potential cumulative impact risks to the Gulf’s marine ecosystems. Our results illustrate how expert uncertainty causes uncertainty in the results of cumulative impact

assessments. We were also able to look for results that were consistent across all of these maps (i.e. were not affected by uncertainty) and we highlighted these as the most certain results of our assessment.

Figure to the left: We had a range of expert scores (representing expert uncertainty about how each human activity impacts marine ecosystems). This meant we could carry out multiple cumulative impact assessments by varying the impact scores within the range the experts gave. The results highlight which results stay the same (i.e. are certain), after we allowed for the expert uncertainty.

We know that it is important to continue to use expert knowledge to inform impact assessments (because of the lack of data from marine systems and the need to make management decisions based on the available knowledge). We hope that the method we have developed can be used by others to support more transparent and reliable cumulative impact assessments for marine (and other) ecosystems.

Article details: Alice R. Jones, Zoë A. Doubleday, Thomas A. A. Prowse, Kathryn H. Wiltshire, Marty R. Deveney, Tim Ward, Sally L. Scrivens, Phillip Cassey, Laura G. O’Connell & Bronwyn M. Gillanders. Capturing expert uncertainty in spatial cumulative impact assessments. Scientific Reports (2018) 8:1469. DOI:10.1038/s41598-018-19354-6.


Data image -,_that_is_the_question_-_journal.pbio.1001779.g001.png