Listening to the Mountain: Using Audio Recorders to Monitor Forest Birds in Kenya
Tropical forest ecosystems are some of the most biodiverse on the planet. However, they face a number of threats including human encroachment and climate change and there is need for urgent and sustained action if we are to protect these ecosystems for future generations. To act before it is too late, we must monitor as many of these ecosystems as possible to detect absence of species in areas they previously occupied.
Bioacoustic approaches have recently been used successfully to monitor a variety of ecosystems. Bioacoustic techniques involve the use of audio recordings obtained in the natural environment to determine the presence or absence of species. These methods rely on the fact that several species produce characteristic calls which allow us to identify them using these recordings only.
For over two years, we have been conducting preliminary studies at the Dedan Kimathi University of Technology (DeKUT) conservancy in Nyeri, Kenya, to test the feasibility of using acoustic recorders to monitor birds and by extension monitor the state of the ecosystem since birds serve as indicator species. This work has been done as part of the Kenya Bioacoustics Project. We have obtained thousands of recordings, hundreds of which have been annotated by ornithologists from the National Museums of Kenya led by Dr Peter Njoroge, and developed methods to automatically detect species present in those recordings. This approach has worked particularly well for detecting the Hartlaub’s Turaco (Turaco hartlaubi, Gĩkũyũ: Ngũgũ) which is a species of conservation interest as it is a mountain specialist found only in East Africa whose range is shrinking as a result of climate change and forest fragmentation. Anecdotal evidence suggests that the bird is moving to higher altitudes as it moves to cooler regions as the climate warms. As such, monitoring the Hartlaub’s Turaco and other species within the Mt Kenya National Park will help determine if species are indeed moving to higher altitudes.
Over the weekend of 17th to 19th November 2017, we got our first opportunity to test our methods within the Mt Kenya National Park thanks to a generous grant by the African Bird Club. The team included David Muchiri, the DeKUT conservancy coordinator, Dominic Chesire an ornithologist from the National Museums of Kenya, Dickson Muriithi from the Kenya Wildlife Service and me. We spent two days mounting acoustic recorders and conducting bird species point counts within the park in the area near the Naro Moru gate.
The plan was to conduct ten minute fixed width point counts at ten locations, each separated by at least 200 meters and to mount acoustic recorders at each point. These recorders were generously provided by Prof Alex Rogers from Oxford and are an early version of the AudioMoth devices based on the ARM Mbed processor. The gadgets are low power devices powered by 3 AAA batteries. We programmed them to obtain one minute recordings every five minutes between 5.30am and 11am to coincide with the approximate time the point counts were conducted. At each point count location, we would record the birds species seen and heard there and also estimated their distance from the point.
Day One — 18th November 2017
Having spent the night of the 17th of November at a hotel in Naro Moru, we set off for the Mt Kenya National Park at 5am. The park gate is approximately 20km from Naro Moru and we arrived at around 6am. After paying the park entry fees and getting a ranger to accompany us, we set off at around 6.30am. We drove further into the park to find an appropriate place to get into the study area, parked by the side of the road and set off on foot, not sure what was in store. Immediately we branched off from the road, we were in thick bamboo forest! We walked for approximately 100m into the forest before selecting our first location for a point count. The study was on!
After the first point count, we set off in search of the next point. The plan was to walk into the forest perpendicular to the road for 2 km placing recorders every 200m. However after the second point count location, our path was blocked by the Tĩgĩthi river, the banks were just too steep to cross safely and we set off along the bank searching for a place to cross. It was also at this point that we discovered that we had failed to adequately mark our route so far, we were only marking the GPS coordinates of the count locations. After walking through the thick bush we noticed that we would need to leave some sort of trail to stand a chance of finding the points on the next day, we had no choice but to tie pieces of paper every point we made a turn.
After walking for quite some time without a chance to cross, we decided to conduct our third count and then cross the river! I learned a crucial lesson that fieldwork calls for flexibility. Luckily we were able to find a place to wade across the river and then proceed with the other points. Just like the first obstacle we met, other challenges lay in store including the unforgiving gradient and the slippery ground. By some miracle, we were able to complete the data collection on day one by 11am! What followed was a two hour hike back to the road between the park gate and the Meteorological station where we had parked our car.
Day Two — 19th November 2017
As on the first day, we set off at 5am from our hotel and began our hike up the mountain at around 6.30am. The plan was to repeat the point counts at the previous day’s locations. This time we had to retrieve our recorder locations and this proved quite difficult in some cases as all we had to rely on were some pieces of paper we had left on our trail and the GPS coordinates taken using the GPS Essentials App. We also had to mark our path more carefully using red ribbons tied on vegetation every 20 meters or so. Luckily we found all the recorders and we were done with the final point count at around 12.30pm. Since we had clearly marked our route, our return journey took approximately one hour.
The study area is dominated by mountain bamboo forest (Arundinaria alpina, Gĩkũyũ: Mũrangi) and podo trees (Podocarpus latifolius, Gĩkũyũ: Mũthengera). Other species that are common in the area include giant ferns (Pteridium aquilinum, Gĩkũyũ: Rũthirũ) and stinging nettles (Urtica massaica, Gĩkũyũ: Thabai — which caused us serious problems during the hike up and down the mountain). We also observed leopard and buffalo tracks on several occasions serving as a reminder that we were truly in the wild.
We plan to leave the recorders in the park for a least one month and hopefully the batteries will last the entire duration. We will repeat the point counts another two times to allow us to make some predictions about the birds present in the area. Once the study is complete we will compare the bird species detected by both the point counts and by listening to a subset of the recordings. We will also develop machine learning methods to detect bird species from the recordings.
I would like to thank Dedan Kimathi University of Technology for access to the conservancy and support during the project, the Kenya Education Network (KENET) and the African Bird Club for financial support, Prof. Alex Rogers and the Open Acoustic Devices team for giving me their recorders to test, Mr. David Muchiri, the conservancy coordinator and all the wardens at the Dedan Kimathi University Wildlife conservancy especially Rashid, Kimathi and Francis for support during fieldwork. I would also like to thank Dr. Peter Njoroge from the National Museums of Kenya for his help with the project.