It’s the responsibility of the Australian Government’s Department of the Environment to make sure that development proposals don’t negatively impact on threatened species listed under the Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act). How do they do this? Part of the job is simply checking where listed species occur in relation to planned developments. It’s important, therefore, to have reliable species distribution maps and a lot of effort goes into maintaining and updating these records. My project, as part of the ANU Summer Scholars Scheme, was to investigate the accuracy of aspects of the Department’s mapping procedures.
Since 1998, the Environmental Resources Information Network (ERIN), a branch within the Department, has compiled and
maintained digital distribution maps of species listed under the EPBC Act. If a species’ population is not indicated on the distribution map, then the species may not be considered during the development assessment process. On the other hand, if a species is mapped in a location where it does not occur, then time and effort may be lost by the developers and the Department. The accuracy of these distribution maps is therefore critical.
In order to ensure that the species distribution is accurate, the ERIN species team has developed and implemented a tool
that automatically updates the existing distribution maps. The automated species-mapping tool incorporates into maps reported sightings of species from external organisations (which include state and territory environment departments, herbariums, museums and some citizen science sightings). Each individual record is checked by an algorithm that ascertains the quality of the data and statistically checks to exclude outliers.
One of the main processes involved here is evaluating each record against 20 bioclimatic surfaces. Developed by Professor Michael Hutchinson from the ANU Fenner School of Environment and Society, these surfaces consist of spatial modelling of climate and terrain. Two key outcomes are outlier identification and surface malfunction. A record is considered as an outlier when it fails to meet the attributes of five or more surfaces. However, when more than 10% of the species records have been flagged as outliers, then the respective surface is considered to have malfunctioned and is then excluded from the process. This helps ensure anomalies in the surfaces do not affect the outlier calculation.
Surface malfunction and false outliers
The surface with the most malfunctions recorded against it was‘Precipitation of Driest Period’. This surface malfunctioned in almost half (46.20%) of all species. The ArcGIS spatial map demonstrated that the cause of malfunction was due to 99.74% of points having a value of ‘0’. The ‘Highest Period Moisture Index’ surface had the next highest rate, which malfunctioned in almost a quarter (24.45%) of all species. Similarly, nearly all records (98.50) belonged to the same value of ‘1’.
“My project, as part of the ANU Summer Scholars Scheme, was to investigate the accuracy of aspects of the Department’s mapping procedures.”
Overall, the results validate that the records which caused surface malfunctions are largely located in spatial areas that contain only a single value. An initial proposal was to omit these two surfaces from the automated species mapping tool due to their perceived lack of usefulness. However, these surfaces are considered helpful even though they consistently malfunctioned because they still identified outliers and may have assisted in records reaching the outlier threshold of failing five or more surfaces.
Individual case studies were also selected to assess the veracity of new species records that had been excluded in the existing distribution. The most common cause for false outliers was records with incorrect coordinates. These records were discovered on the basis that the locality metadata did not match the spatial information contained within the coordinates. In most instances, these records were horizontally or vertically away from the actual locality being within the known distribution. When such errors are discovered in the species data, the source supplier may be notified by ERIN. However, in the past, these agencies and organisations have not necessarily updated their information. Despite the knowledge that certain records are inaccurate, the lack of time, resources and staff were reasons that prevented the scientific source from having the most up-to-date information.
ERIN sits at the interface between scientific research with policy and decision making. Science underpins both the source data of species sightings collected by external environmental organisations and the bioclimatic surfaces used to assess them. The value of accurate species distribution maps is worth the additional effort required for collaboration between the department and external agencies to ensure that incorrect sources, once identified, are fixed. In addition, the malfunction of bioclimatic surfaces could be addressed through further collaboration between the Department and the ANU – either by developing more detailed surfaces or improving the outlier detection algorithm.
I am grateful for the opportunity I had in conducting research for the Australian Government’s Department of the Environment. Despite the challenges of working within the realm of the natural environment (my background is in architecture and the built environment), the experience pushed me outside of my comfort zone. Specifically, I have developed a greater knowledge of the processes and systems available to the government in not only creating maps of the country’s fauna and flora but in how such information assists in protecting Australia’s biodiversity and threatened species.