Conservation by numbers

New Zealand has amazing birdlife: nocturnal parrots, birds that can’t fly, unique and beguiling birdsong, and varieties that turn up 50 years of being thought extinct. Unfortunately, many native species require wildlife management programmes and their clever camouflage often makes them hard to monitor.

Stephen Marsland, wearing a blue hoodie, leans against a tree with green vegetation behind him.

This is why Professor Stephen Marsland from the School of Mathematics and Statistics is co-leading a large, interdisciplinary project using data science, mathematics, and new technology to help New Zealand birds survive.

“The aim of this project is to take recordings of birds collected in the field and identify all those that are calling, using mathematical and computational methods such as machine learning,” says Stephen.

The project involves collaboration with mathematicians, ecologists, statisticians, electronic specialists, engineers, and iwi from across New Zealand.

“Ecology produces a lot of data. Acoustic recorders, which are the main tools we use to monitor the bush, produce gigabytes of data and there are a few thousand of them in New Zealand alone. Add in the trail cameras, trap records, tracking tunnels, and chew cards and you start to understand how much data is already being recorded. Turning all this data into useful information that is accessible to everybody is the role of data science.”

“As a mathematician, it is tempting to stay away from applications but conservation and ecology are very important to me. Working out ways to use data effectively has become key to seeing what works and what doesn’t work in conservation.”

Monitoring the birdsong of native birds does come with its challenges however, none more so than background noise.

“Acoustic recorders are designed to capture all sound within a broad area. This means that they pick up all the sounds of the environment – wind, rain, streams, cars, other animals – as well as the birds, which can be up to 500 metres away. Extracting those calls from the general soundscape and then recognising them reliably is a really interesting problem we have to solve.”

Stephen and his team of researchers use their knowledge of the different species and statistical methods to turn the call rates into estimates of the numbers of birds present.

“The software that we are writing will be freely available to everybody who wants to use it, from community groups through to the Department of Conservation and other researchers,” he says. “We’ve used it to detect the presence of kiwi in the Ruahine ranges, to estimate the abundance of little spotted kiwi in Zealandia, to find bittern in wetlands, and to recognise the two New Zealand bat species.”

“The great scientist Lord Kelvin observed that you can’t improve things if you can’t measure them. We want to improve conditions for New Zealand’s wildlife, and to do that we need to know what works and what doesn’t. By monitoring the abundance of the birds and seeing how that changes over time, we are able to help wildlife managers and other conservationists to choose effective interventions.”