Real-time Curlew acoustic monitoring

Destruction of nests and chicks as a result of agricultural activities is a significant factor in the decline of many farmland and grassland bird species globally. Mowing, ploughing, harvesting, and slurry or muck spreading during the breeding season can destroy nests, cause nest abandonment or result in chick mortality. Eurasian Curlew are in severe decline in Western Europe and in many contexts agricultural activities are a contributor to these declines, as well as high rates of nest and chick predation associated with increased populations of generalist predators. There are various conservation groups working on the remaining isolated fragments of lowland Curlew populations in the UK - these groups have a strong focus on finding nests as quickly as possible so they can be protected both from agricultural activities and predators - it is not unusual for groups to report finding nests and the nest being predated by the time they return with the portable electric fencing equipment to protect the nest from mammalian predators. Particularly in cases where birds are nesting in silage fields if the nest isn't found, there is a chance that the nest will be spread with slurry or muck causing abandonment, or destroyed by the first cut of the season if predators haven't found the nest.

Over the last few years we have gathered a training dataset of Curlew vocalisations and Simon Gillings at the British Trust for Ornithology has trained a neural network to identify the specific alarm call made by adult Curlews when their chicks are in danger. So far, we have used Audiomoth recorders to gather data by recording onto an SD card for data to be downloaded and analysed at the end of the breeding season. This allows us to assess the extent to which breeding densities, productivity and hatching dates can be ascertained from the patterns of alarm calling and display calling across the breeding season. There is significant interest now within Curlew conservation groups in the development of a real-time Curlew vocalisation sensor - if Curlew conservation groups had networks of real-time sensors across the areas Curlew were nesting in providing up-to-date information on where Curlew were vocalising, it would be possible to target fieldwork effort more efficiently, and locate nests faster and ultimately protect more nests.

As such, the likely next step will be to deploy the Curlew call classifier model on a Raspberry Pi device that can run the model in near-real-time in the field, and send notifications over the mobile phone network to fieldworkers or researchers when it 'hears' a Curlew vocalisation. In time, with further development, such a system could provide benefits to conservation teams working on the ground and farmers working to conserve Curlew in managed grassland systems.

Curlew nest in managed grassland in the Yorkshire Dales National Park.

Audiomoth recorder deployed in Curlew habitat in the Cairngorms National Park, Scotland.