Using Cacophony Project technology to find the South Island Kōkako

Every year there are a number of compelling reports of encounters with the elusive South Island Kōkako. This precious bird with a beautiful haunting song was once declared extinct but hope remains of finding it alive and bringing it back from the brink of extinction.

The South Island Kōkako Charitable Trust is on a mission to find any remaining birds and help with the recovery if possible. The first step is to get 100% verifiable identification via sound or photos. It seems most likely that the distinctive song will be the key to this identification.

Currently general purpose sound recorders are used to try and record Kōkako. These just record for a period of time and then people have to manually listen to the whole recording, potentially for hundreds of hours, to try to find a Kōkako. After talking to the South Island Kōkako Trust we think we could modify The Cacophony Bird Monitor to make it much more likely to find a Kōkako. The monitor could be modified to play a Kōkako sound and then listen for a response at several predetermined times during the day. This could dramatically increase the chances that someone reviewing the recordings will find a responding Kōkako.

The nice thing about The Cacophony Project's Bird Monitor is that recordings captured are GPS and time stamped, and then uploaded to cloud storage where they can be accessed over the web. This makes it easier for groups of people to review recordings and for them to be automatically analysed any time in the future.

The obvious next step would be to use Artificial Intelligence (AI) to try to find audio recordings that are most likely to be Kōkako. This is a much harder problem particularly given that we have been relying on recordings of North Island Kōkako - recordings of their South Island cousins being almost non-existent. To work around this lack of data to train from, one approach could be to train an AI to look for bird sounds that are sufficiently different from other known species. This could then be used to guide human detection efforts.

Both of these projects are outside The Cacophony Project's current goals around removing introduced predators but our technology is open source and largely applicable to the search for the Kōkako. If you or anyone you know wants to use their technical skills to help this project, please get in touch.

Below is a table with a short description of the parts of The Cacophony Project that are relevant to Kōkako project and a description of how they could be used.

Existing ComponentCacophonyKōkakoComment
Low power robust sound recorder (smartphone)XXStandard hardware we already use would work for the Kōkako
Sensitive heat cameraX  
Flexible power source (battery/solar/mains)XXWould likely use solar powered version to last a long as you need
Flexible trapping/experiment systemX  
Digital lure toolsXXCould be adapted to play North Island Kōkako call as a lure using a separate speaker. One approach could be to play the call several times around dawn and dusk, and periodically through the day, and then record audio for the following minutes?
Flexible tools for getting data to the cloud (app)XXUse SD card storage and then the Cacophony Project's tools for uploading to the cloud once the SD card is retrieved.
Software running on devices
App for sound recording and auto uploadingXXThe cell phone app could be modified for the best chance of recording Kōkako
App for wireless data upload and camera set upXXIf you have cell phone reception this part is not needed
Software to run the heat cameraX  
Software to run the flexible trapping/lure systemsX  
Web Software
Secure multi-user web interfaceXXAllows anyone from the team to see all data gathered
Robust data managementXXBacked up and secure
Ad hoc data viewing tools on webXXManually listen to recordings
Artificial Intelligence data tagging toolXXTag recordings for AI training
Artificial Intelligence animal recognitionXXTry to train a model to detect Kōkako-like sounds
Reports and graphsX  
Device managementXX 
Experiment creationXXSpecify parameters for lure playback (harder on cell phone)
Trap controlX