Using acoustic monitoring to understand human impacts on Amazonian wildlife

Posted by Oliver Metcalf

It was dawn on my first morning in the Amazon. 24 hours earlier, I had been walking through the snow in Manchester to catch a bus to the airport, now I was stood 50m up on a scaffolding tower, soaring above the rain forest canopy, listening to the forest come alive. We had been so keen to get here that we hadn’t slept, briefly dropped our bags at the accommodation and then headed straight out on dirt roads, so that my first sight of Brazil should be deep in the heart of pristine rain forest. Moments earlier, there had only been the occasional mournful whistle of a tinamou and the otherworldly, nightmare-ish growl of howler monkeys, but now the dawn chorus had started and songs and calls came from all directions.

The Tapajos National Forest at dawn – the sun glowing red from the smoke of surrounding small wildfires (Photo: Oliver Metcalf)

Alex Lees, my PhD Director of Studies, rattled off the species names like a football commentator during a particularly frenzied passage of play, recording each species on eBird as he pointed them out – cryptic forest-falcon, black-tailed trogon, rufous-winged antwren, three-striped flycatcher - before long my head was spinning and I couldn’t keep up. Zoning out of trying to identify individual species, I took a minute to feel this new environment, to indulge the surreal feeling of being so far from home, in the world’s greatest rain forest. But something jarred with the image my eyes were providing of pristine wilderness, there in the background of sensory overload was the acrid, lingering smell of smoke. 

Above: Undisturbed forest looks dark because of an intact canopy layer. Below: Burned forest lets in more light as dying trees open up gaps in the canopy. Many species in the Amazon are photophobic and won't cope well with these brighter conditions (Photos: Oliver Metcalf)

The ecological and environmental impact of deforestation in the Amazon are well known and well publicised. Many of us have watched news reports of falling canopy giants accompanied by the soundtrack of screaming chainsaws, or learnt in geography classes about the harm slash-and-burn agriculture can cause. However, the high visibility of this environmental destruction has led to the Brazilian government taking action, and the rate of deforestation has declined since their 2004 peak (albeit with some recent increases). Various sustainable forestry initiatives have been used to slow the rate of deforestation– including extractive reserves which permit selective logging. These allow large areas to be set aside ostensibly for conservation purposes whilst still generating income from extractive industries.

Delights and dangers of Amazonian fieldwork. Above, the lovely Wing-barred Piprites Piprites chloris (Photo: Nárgila Moura); Below, a Fer-de-lance Bothrops atrox sharing the trail with us (Photo: Oliver Metcalf)

Large swathes of Amazonia are now subject to disturbance from selective logging and fire, yet our understanding of the impacts of such disturbance on biodiversity is still limited (Barlow et al., 2016). One of the most harmful consequences of selective logging is an increase in forest susceptibility to wildfire, exacerbating an already volatile situation caused by drought and climate change. In fact, during the El Nino events of 2015-2016, there were so many wildfires in the Santarem region (where my study site is located), the area was responsible for driving a large rise in atmospheric carbon dioxide levels. Fire can have devastating and immediate impacts on tropical forests. During my PhD, I will be working with the Sustainable Amazon Network, a transdisciplinary research consortium that studies the effects of disturbance on all levels of the rain forest environment and its biodiversity, as well as the best means to avoid and mitigate these impacts. I will be working in a protected area that allows selective logging, the Tapajos National Forest (FLONA) in Para state, Brazil, to establish the impact of logging and fire on the biodiversity of the park, large areas inside and outside of the reserve were burnt in 2015-2016. Consequently, my study design will focus on establishing the impacts of disturbance on biodiversity across a degradation gradient – comparing areas that are variously undisturbed, logged, burned or logged and burned.

Layard's Woodcreeper Lepidocolaptes layardi (Photo: Alex Lees)

Technical innovations in measuring biodiversity in the rain forest

Of course, comparing ‘biodiversity’ in such a diverse and challenging region is easier said than done. Trekking from point count to point count, transect to transect can be extremely hard work, time consuming and even dangerous, with a preponderance of poisonous snakes, bullet ants and wasps. Even if everything goes to plan, observing wildlife can be extremely difficult, frogs hide in hollow logs, birds feed in canopies 50m above and mammals run for cover well before a surveyor can clap eyes on them. This means that the best way to survey a wide range of species in the Amazon is to listen for the sounds they make. Even then, some species will only call at dawn and dusk–particularly birds such as forest falcons and tinamous, whilst others will wait for the dead of night or the heat of the day before making noise. This means that any attempt to survey all species using point counts or transects will inevitably miss a range of species, unless repeated a huge number of times across the day.

Some of the wildlife we hope to record; Above: an Amazonian pygmy owl Glaucidium hardyii, Below: a white-cheeked spider monkey Ateles marginatus (Photos: Oliver Metcalf)

Instead, I will be using passive acoustic recording devices from Frontier Labs to record continuously for a week at each of 40 transects spanning the degradation gradient, once in the wet and once in the dry season. These devices allow a huge amount of data to be collected with comparatively little time required in the field, and have previously been used to survey a wide range of species, as well as anthropogenic disturbance events such as chainsaws and gunshots. The difference here is that I will aim to detect the vocalisations of all the species within each transect, including birds, mammals, frogs and insects (although many of the insects will be morphotyped rather than identified to a species level). This will allow a greater understanding of how entire forest communities are affected by disturbance, rather than particular species within it.

The widespread Forest Elaenia Myiopagis gaimardii should feature on the 'tapes' (Photo: Nárgila Moura)

Of course, analysing the huge amounts of data collected (over 20 terabytes) is a challenge in itself, and it would be impossible to manually listen to all of the sound data that will be collected. I will instead be using Tadarida, a program that utilises machine-learning algorithms, to automatically detect vocalisations. This is a painstaking process in which I must iteratively train the program to recognise each of the call types present in the data. To start with, I will focus my attention on the nocturnal and crepuscular communities, as these are the least studied and hardest to survey using traditional techniques – and the simpler soundscapes will give the automated recognition software the best chance of performing well. Even still, I will have my work cut out for many months to come! After the training process, the program will look through the entire dataset for further similar calls allowing the creation of a database of the presence or absence of a wide range of taxa, equivalent to many hundreds of repeat transect surveys.

A spectrogram of a small section of the data so far recorded – most of the marks below 6kHz are bird calls, whilst the dark line in the centre is an insect being very noisy!

This will allow us to understand, at a broader scale than ever before, how communities respond to the disturbance gradient, and gain a through understanding of how the precious, unique biodiversity of the Amazon is being impacted by ongoing human disturbance.
 
Oliver's PhD at MMU is supervised by Dr Alexander Lees, Professor Jos Barlow (University of Lancaster) and Stu. The PhD is a component of Rede Amazônia Sustentável (RAS), a multidisciplinary project on sustainable land use in the Amazon.