Sensitive species data obscuring

Modified on Tue, 14 Apr at 2:17 PM

Author: Zachary Amir (TERN)

This article can be read in conjunction with:

TABLE OF CONTENTS

1. Background: threatened versus sensitive species

2. How WildObs identifies sensitive species

3. What gets obscured – and what doesn't

4. How obscured records appear in a download

5. Alignment with the Atlas of Living Australia

6. The RASD framework and future policy direction

7. Requesting access to full sensitive species data

1. Background: threatened versus sensitive species

The terms "threatened" and "sensitive" are not interchangeable in the context of biodiversity data, even though they frequently overlap.

Threatened species are those formally assessed against criteria under legislation – most commonly the Commonwealth Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) – and assigned a conservation status category (e.g., Vulnerable, Endangered, Critically Endangered, Extinct in the Wild).
Sensitive species are those for which the exact location of occurrence records is considered too risky to make publicly available. Sensitivity may arise from threats such as poaching, collection, or disturbance, and does not require a formal threat listing. Sensitive species listings often apply to specific populations rather than the full distribution of a species.

Not every species that is threatened is sensitive, and vice versa. WildObs currently uses the EPBC Act threat classifications to identify species whose location information requires protection in public data downloads. We are actively working toward a more nuanced framework (see section 6. The RASD framework and future policy direction below).

2. How WildObs identifies sensitive species

When a data package is downloaded from the WildObs Database, the download function checks all detected species against the WildObs maintained taxonomy dataset that is bundled within the WildObsR R package. This dataset contains verified taxonomy and species traits for all species recorded in the WildObs Database, including their state-specific EPBC threat classification.

The sensitivity check works as follows:

All species in the dataset are matched against the verified taxonomy dataset using their scientific name (`scientificName`).
Any species with an EPBC category other than "not listed" – that is, any species classified as "Vulnerable", "Endangered", "Critically Endangered", "Conservation Dependent" or "Extinct in the Wild" – is flagged as sensitive.

Critically, this flagging is state-specific. The data package’s geographic footprint is first identified (that is, which Australian state(s) the camera deployments fall within), and only species listed as sensitive in that state are flagged. A species listed as Endangered in Queensland but not in New South Wales will only trigger obscuring for surveys conducted in Queensland.

3. What gets obscured – and what doesn't

The guiding principle behind the obscuring approach adopted by WildObs is to protect sensitive location information without degrading the scientific value of the broader survey. A single sensitive species detection should not render an entire dataset unusable.

For more information on the Camtrap DP data package and description of the associated data resources, visit WildObs data standards.

What is not obscured:

Camera deployment locations (`deployments` data resource). Coordinates of camera trap locations are retained in full. Rounding or generalising deployment coordinates would reduce the utility of all other species detections at those same sites – the vast majority of which are not sensitive.
Project-level species lists (`datapackage.json`). Species names, including sensitive species, remain present in the taxonomic metadata of the data package. This allows data users to know that a sensitive species was recorded in a survey, while precise detection events or locations cannot be traced.

What is obscured:

Identifier fields `observationID`, `eventID`, `mediaID`, and `deploymentID` within the `observations` data resource. These four identifier fields link a specific observation back to its camera location, media file, and detection event. They are replaced with a standardised placeholder for each sensitive species record.

The placeholder takes the form `obscured_for_[epbc_category]_species` – for example, `obscured_for_endangered_species` – so that data users can clearly see that a record has been intentionally modified and understand the reason why.

4. How obscured records appear in a download

In practice, a download containing sensitive species data will look like this:

The `deployments` & `media` data resources are complete and unmodified – all camera locations are present with full coordinates, and all media file paths remain present.
The `observations` data resource contains all detection events, including those for sensitive species. However, for sensitive species records, the four linking fields (`observationID`, `eventID`, `mediaID`, `deploymentID`) will display the standardised placeholder rather than the actual identifiers.
All non-sensitive species records in the same dataset are completely unaffected.

This means it is not possible to determine at which specific camera, on which specific date, or from which specific image a sensitive species was detected. The chain of evidence linking "this species was here at this time" is intentionally broken.

For example, in a survey with 200,000 records across 50 cameras, if two Koala detections are categorised as sensitive (only applicable in some states), only those two records are modified. This approach means that the other 199,998 records remain fully usable for research.

5. Alignment with the Atlas of Living Australia

WildObs partners closely with the Atlas of Living Australia (ALA) and actively aligns with their sensitive species policies. It is worth being explicit about how our approaches are similar and how they differ, because both achieve the same conservation goal through different technical methods.

The ALA generalises sensitive species data based on state and territory sensitive and restricted access species lists. Depending on the sensitivity, records are withheld entirely or their geographic coordinates are generalised to 1 km or 10 km resolution by removing decimal places from the latitude and longitude.

The WildObs approach is better suited to the structure of camera trap data packages. Rather than rounding coordinates, WildObs retains full deployment coordinates but severs the relational links between a sensitive species observation and its spatial, temporal, and media context.

The effect is equivalent: a data user cannot determine where, when, or on what occasion a sensitive species was detected.

Both WildObs and the ALA use state- and territory-specific species lists to determine what requires protection, and both provide clear indicators to data users when a record has been modified. The ALA's sensitive species lists are publicly accessible, and WildObs draws on the same classifications to maintain consistency.

When WildObs data (contributed under an open license) is shared with ALA, detection histories are collapsed to presence-only data. Unobscured location information is shared directly with ALA, and sensitive records are handled in accordance with ALA's approach. This coordination ensures that obscured identifiers in WildObs shared data are appropriately and consistently managed in ALA's occurrence data.

For more information on data sharing options, visit Data sharing terms.

6. The RASD framework and future policy direction

WildObs is actively transitioning our sensitive species policy from EPBC Act classifications to the National Framework for the Sharing of Restricted Access Species Data in Australia (RASD).

The RASD framework is an initiative supported by the Australian Research Data Commons (ARDC) and the Atlas of Living Australia, developed collaboratively with Australian Government, state and territory conservation agencies, museums, and herbaria. It provides consistent best-practice guidance for sharing restricted-access species data between trusted parties and a nationally consistent method for modifying restricted-access data for public release.

Key reasons why RASD represents an improvement over EPBC-only classifications:

EPBC classifications are federal and do not always capture the full picture of state-level sensitivity. RASD is designed to integrate both federal and state/territory listings into a single, authoritative framework.
RASD covers a broader set of restricted access data types, including location-related, identification, and attribute data, providing more nuanced guidance than a simple threat-level threshold.
The RASD framework includes the Restricted Access Species Data Service (RASDS), which provides a single point of contact for approved users to request access to sensitive data from custodians.

The RASD framework and its associated species lists are still being finalised at the time of writing, with state and territory agencies in the process of review. WildObs aims to integrate the completed RASD-aligned species lists as soon as they are available, in coordination with the ALA.

Beyond national and state frameworks, WildObs recognises that formal listings do not capture every situation where data providers have reasons to protect species location information. A landowner may be monitoring a locally significant dingo population, or a conservation manager may be tracking an unlisted but vulnerable colony at a specific site – neither of which would trigger EPBC or RASD classifications.

WildObs is working toward a mechanism that allows data providers to nominate species for sensitivity protection during the data submission process or through review of data submission quality control reports. This will allow provider-defined sensitivity flags to be stored alongside the dataset and applied consistently during any public download. We see this as an important opportunity to build trust with data providers and ensure that the people closest to the data have a meaningful voice in how it is shared.

7. Requesting access to full sensitive species data

Public downloads from the WildObs Database via WildObsR apply sensitive species obscuring by default. Researchers or conservation practitioners with a legitimate need for unobscured sensitive species data have two potential avenues:

Via ALA. The ALA handles sensitive species access requests, creating a documented record of who accesses withheld information. Users submit a request to access data via the RASDS, including information about their role and the purpose of the access. If approved by data custodians, data access is facilitated by authentication using the relevant account.
Via WildObs directly. We are developing a streamlined data access request tool that will allow researchers to apply for access to full (unobscured) data under partial data-sharing agreements. All requests will be subject to data provider approval, and sensitive species access is expected to be handled through the same mechanism. Details will be published here when available.

If you have questions about how your survey data is handled or believe a species has been incorrectly classified as sensitive for your region, please contact the WildObs team.

WildObs is committed to making Australia's camera trap data as open and useful as possible, while fulfilling our obligations to protect species at risk. If you have feedback on these policies, we welcome your input.