ClimateWatch data quality control process
by Linden Ashcroft, 22 May 2014
Submitting your ClimateWatch sightings is only the first step in developing a national phenology database, even though it’s the most important. Before the ClimateWatch sightings are made available for research, they go through a quality control process involving the ClimateWatch team and our expert scientists.
Sightings are submitted by registered ClimateWatchers who all have access to species field guides online and through the smartphone app. This information aims to help people correctly identify the species they are observing.
2. Initial quality check
The data are extracted from our system on a species by species basis. This process occurs approximately twice a year. Once extracted, the data are manually analysed for:
- Location errors. This includes latitude and longitude coordinates that are clearly incorrect (e.g. 90°S, 180°E, locations in the ocean), and coordinates that have been submitted in degrees-minute-seconds format rather than decimal degrees. Observations with northern latitude coordinates or degree coordinates are corrected where possible. Observations with coordinates in the ocean or out of Australia are deleted. Some sightings are also submitted for remote locations (e.g. central Australia). In these cases, the details of the observer and region are checked to see if they are reasonable sightings. If not, they are removed.
- Image errors. This means images that don’t match the species being recorded, repeat images from the same user, sightings from different users that have the same image, or submitted images that have been sourced from the Internet. All sightings with these errors are removed.
- Duplicate sightings. Technical issues in an earlier version of ClimateWatch means that some duplicate sightings are present in the system. All duplicate sightings are removed to ensure there is only of each observation.
This process is currently conducted manually, and so remaining errors do occasionally make it through to the species quality check.
3. Species quality check
Next, the data are uploaded to the Atlas of Living Australia. From here, sightings are analysed by the ClimateWatch species advisors for more technical errors. Sightings that might be flagged by our advisors include a misidentification of species or a phenophase that is out of season. A sighting will be deemed incorrect if the image does not match the recorded species, or the location of the sighting is well out of the expected species distribution.
Suspicious sightings are flagged and the sightings are updated or deleted as suggested by the individual species advisors. In some cases, the details of the observer are also checked to see if the suspicious sighting is reliable. ClimateWatchers that are registered as students taking observations for an assignment are generally considered unreliable. Individuals who have a long record of ClimateWatching are considered reliable, as are groups who are being supervised (e.g. Scout groups or guided trails).
4. Future quality control steps
Future quality control of the ClimateWatch dataset will include contacting the original observer to query any suspicious sightings, and automatic species distribution checks within the website and app to minimise species misidentification.
If you want to volunteer with ClimateWatch and help out with this process, we would love to hear from you! Whether it is technical support or checking submitted images, it would be great to have you on the team.