San Diego Zoo Global Institute of Conservation Research
July brought us some logistical challenges! Colleague Diego had set up 20 SM4 audio recorders in the Maijuna-Kichwa Regional Conservation Area two months previously. He had been rained on copiously, and enjoyed fast-flowing, deep streams to access the forest. The recorders had spend six weeks listening for gunshots, and animals, recording 24 hours a day with 512GB SD cards and a large car battery each. So with some excitement, I went up the Sucusari River to collect those recorders and their paired camera traps. Unfortunately the water level dropped much earlier than expected, and the river dried to a trickle in some places. River sections that took Diego and the team an hour, now took half a day hauling the canoes and clearing logs with the chainsaw. It was clear that the top half of the stream was not even accessible. With great effort, and a LOT of hiking through a very try 'rainforest', we extracted half the recorders and camera traps. Half the recorders are still in the forest, and we have to wait another couple of months to bring them out. On the plus side, the car batteries were still fresh - the units used far less power than I expected. This means I'll get another months worth of data upriver until the cards fill up.
One of the main applications we have is to listen for gunshots in this attractive reserve, and I completed some important range tests while bringing in these first recorders. Conditions were perfect - low wind and no rain, and even before analysis, it is clear that the under these conditions the range is close to 2km... given that our sample points are typically about 2km apart in our camera trap and acoustic surveys - this means that we have almost complete coverage in perfect conditions. We will have more 'real word' tests when we combine our audio data with spatial hunting data from our GPS tracked hunters who are registering their hunts in a wider range of conditions.
I now have to repeat my expedition in October, to get the units we failed to retrieve before. Then we face the not insignificant task of analyzing about 8 weeks worth of data from 20 SM4+ recorders, all recording 24 hours per day. We plan to use Wildlife Acoustic's 'Kaleidoscope software to analyze maybe 25,000 hours of recordings, first for gunshots, but then for peccaries, woolly monkeys and a range of key species. Beyond that we will have a resource available with which could survey birds, amphibians or insects. But first, I'm looking forward to getting back out to the Sucusari to recover our recorders.
This month we completed the installation of 20 SM4 recorders at 2km intervals in a large array in the Maijuna Regional Conservation Area (MRCA), on the Rio Napo in the Peruvian Amazon. The recorders are in trees to keep them safe from curious people, and hopefully to maximize the range we get from the units. This was backbreaking work for field assistant Diego, who had to climb all the trees! The SM4s are connected to car batteries and we hope to have at least six weeks constant recording from each unit, in which we should be able to find recordings of a wide range on animals. We hope to find these with the help of Wildlife Acoustics Kaleidoscope software, but a camera trap close to each recorder will give us a clue as to where to find sample recordings of noisy animals like peccaries. The main reason for deploying the recorders, however, is to record gunshots in the area. the MRCA is an extractive reserve, so there is legal hunting by residents, managed by the community themselves, and informed by our research. We track much of this hunting with GPS trackers on the guns of hunters, and can pinpoint the location of kills. With this we can calibrate the range of our recorders under varying conditions through several weeks. We will also be able to detect hunting by unknown individuals who may be hunting in the area without permission, directly informing the community management of the area. We hope audio recorders will prove to be a highly efficient way of monitoring hunting that can be expanded through forests across the globe.
Kaleidoscope is was able to find virtually all our test gunshots — even most of those as far as 2km away. This is another step towards the autonomous monitoring of hunting across forested areas. Now we have the small matter of 4TB of recordings to process - to find the real-world gunshots, and especially those of our GPS tracked hunters. This means building 'recognisers' that Kaleidoscope can use to sort through the data more quickly and accurately, ignoring (for now) the hundreds of thousands of bird, insect and mammal noises that we have also recorded in our 2-month 24-hour recordings. Bottom line though - IT WORKS! We have a complete record of the hunting for two months on the Sucusari River basin.
Meanwhile, the recorders have not been idle, we have employed them on an exploratory project to find arguably the world's rarest bird. Confined to a few small patches of unusual white-sand forests, most of the bird's habit has been consumed by the demand for sand in the city of Iquitos. Now confined to The Allpahuayo Mishana National Reserve, we don't know where or how many there are- it could be as few at 20 pairs. We are using the SM4s and Kaleidoscope with the help of the local bird experts to try to find the bird and perhaps determine patch occupancy. So far we have not heard it, but have discovered another rare bird that is new to the Reserve - the barred tinamou- hardly ever seen or heard because it sings at 4am in the morning! The SM4s will soon go back out to listen for gunshots - this time on the Yavari river, but it goes to show how valuable these units are once you have them in the projects armory.
This is what we are looking for - Kaleidoscope sorts gunshots from our range tests into 'clusters' that we can identify and use to make classifiers