Cacophony Project: How Innovative Technologies are Helping to Restore our Dawn Chorus
Thank you very much and Kia ora koutou Kō Shaun Ryan tōku ingoa Nō Ōtautau ahau Nō Karitani me Airangi Me Ingarani ōku tīpuna Kō Ngai Tahu tōku iwi Kō Aoraki tōku maunga Kō Aparima tōku awa Kō Grant raua kō Craig ōku teina Kei Ōtautahi tōku kainga ināianei Ka nui te mihi kia katou katoa Nō reira tena koutou tena tatou (screen freezes, loss of sound) ...what's new with the Cacophony Project but I thought I'd start with just a little bit of a background on the project So the Cacophony Project was started by my brother Grant quite a few years ago now and Grant's a bit of a serial entrepreneur he has a mechanical engineering degree and a PHD in ecological economics whatever that is after the Christchurch earthquakes he moved over to Akaroa and ended up buying himself a pest infected house and like a lot of Kiwis he started trapping and kept trying to you know do a better and better job and after sort of a year or so he thought he noticed that there was more bird song at home and his engineer's mind kicked in he's like I wonder if I could measure that and at the same time he was becoming more and more disillusioned with the impact with how effective the traps were and how effective the cameras were capturing what was actually happening so he started the Cacophony Project to create technology to help with the New Zealand help the predator problem help the New Zealand birds and and it's a not-for-profit that's creating open source technology which means anyone can use the technology at no cost and what happened is after they developed their first products a bird monitor and a thermal camera they started getting people asking for these products and they had to then manufacture them and sell them and support them and he really wanted someone else to do that and anyone could do it because it's open source it's free to access the technology but no one was and I was at a bit of a loose end and he said bro why don't you do that so I've started a separate sister company called 2040 which is manufacturing selling and supporting the technology that Cacophony does it's a for-profit social enterprise with the mission of saving New Zealand's birds and we call it 2040 with the hope that with the exponential growth of technology that we could get to New Zealand's aim of being Predator Free 10 years before the government's ambitious goal So the Cacophony strategy can sort of be captured by the products that we've released so far there's a we've started off with the monitoring side of things because we want to be able to measure the impact of whatever we're doing and understand what's happening with the predators before we start creating technology to address the problems so the first product we have is a bird monitor which is a reasonably simple product it's designed to permanently record birdsong several times during the day it wakes up and will just do one minute of recording it uploads the recordings up into the Cloud and initially we weren't doing anything with those recordings apart from just storing them and making them available so you could listen to them which was pretty uninteresting but within about 18 months ago we released something called the Cacophony Index which is a measure of how much birdsong is in the recording there's an algorithm that one of our open source contributors developed and the little graph you can see here sort of shows how the Cacophony Index varies over the day and it's what you'd expect during night time it's very low and then you see this peak for dawn chorus and then it's elevated during the day and you get another peak at dusk so that's now available for all the recordings and we've got other open source contributors who are working on bird recognition algorithms they currently have a morepork algorithm working pretty effectively and we're we're looking to get some other algorithms going I just want to check that we're not having any technical problems here it's still going all right okay it's all good it's all good okay I'll continue yeah so so that's the bird monitor so eventually we hope to be able to tell you exactly what birds are in those recordings and you can get reports on how those bird numbers are changing so the important thing with that is to get the bird monitors out there start collecting the data so you can then see how those things are changing One of the other products they developed was a thermal camera and this is a picture of it here and Grant didn't really start off wanting to develop a thermal camera but he found that when he was putting trail cameras in front of the traps it would often obviously miss stuff there'd be the bait would go missing without capturing it or you'd just see the tail of an animal and so we developed a thermal camera to get something that was more sensitive than the existing trail cameras in fact we've had Lincoln University measure how effective these are compared to trail cameras and chew cards and they get about three times the number of possums compared to what a trail camera does and what these cameras do is they upload all of the videos they record to the cloud and we run machine vision algorithms over them so we automatically identify whether it's a possum or a cat or a rat or a stoat or a bird so it takes a lot of the labour out of the monitoring process and you can get some sort of nice reports on the activity how many visits there have been for predators on the different cameras and then so the cameras have let us do the monitoring job more effectively but they've also let us understand predator behavior and I'll show you some examples of that and then we've also got a trap and I'll talk about the the trap as well which is we're currently testing at the moment with some exciting results so this is where we are with our various projects so each of these products that we're trying to create we want to first of all create a viable model for how we might be able to solve the problem then build a prototype then try and make it the best tool that's on the market then try and see if you make it the lowest cost and and get it onto the market so we've got the the bird monitor and the predator monitor both are on the market the rapid lure trap experiment tool is really a combination of a camera that has a speaker plugged into it so you can play sounds to act as a lure so you can see what's happening with the trap and experiment with the trap design and that's that's available now and we've got various people that are playing around with that this model I'll talk a little bit it's just a spreadsheet but it's a model of how predators predator numbers change as a response to trapping because we wanted to see what things have the biggest impact and that model we've sort of created and it's available for anyone to download and have have a play with critique if you want to and that and then the multi predator suppression trap is currently we've got a prototype and we're just working well we're getting some really good results Grant's quite ambitious with his rankings of these but he's saying he thinks it's the best tool on the market in terms of the results it's getting I'll leave you to decide for yourselves when you see the videos and then we've got some ideas on how to do the last predator re-invasion elimination tool and we've recently come up with this thing called an active fence that we're prototyping so some more detail with the thermal camera as I said the thing with thermal cameras it's about more effective monitoring trying to take a lot of the labour out of that and trying to have a better understanding of what's happening there and also getting a better understanding of predator behavior so we're using Living Springs near Christchurch as a test bed they've got lots of all sorts of different types of predators are happy for us to see what we can do there and we did this experiment where we did some traditional monitoring with tracking tunnels and chew cards and we've developed a protocol for using the thermal cameras over the same sort of area that we've published so that's something new and you can see with this particular experiment the tracking tunnels and chew cards got about 63 predators over this time or 63 animals and our cameras we had three cameras that we swept across the area it's doing exactly the same area they detected 978 animals and so you can just see that you get so much more information with these cameras so it's really exciting you have a much better understanding of what's happening I'll give you a little example this is an experiment one of our customers is the Eastern Bay Songbird Project and they put one of our cameras here in front of the traps in front of a whole cornucopia of traps in front of this tree here and this is what they saw so this is they've followed the sort of manufacturer's instructions in terms of what lures they should be putting in what baits they should be putting in and this is the video sped up sort of 20 times that's obviously a cat going around and having a look a possum up a tree there's quite a few cats in here I believe there's a rat more cats same cat yeah same cat multiple times there's a rat climbing all over the traps it's kind of frustrating Just go in It might have been the hedgehog Here we've got a possum You can see they get close the recurring theme is none of them are interacting with the traps The possum's getting really close There we are and so we've got 52 predators and no catches and that's that's really disappointing but that's what we see all the time with these cameras and we think it's new information because other cameras don't capture as much of the information and in fact what we believe was seen was less than a one percent interaction rate so you need a hundred predators to go to go past before you catch one which is amazing but it also is a real opportunity because if you can improve that you can make a difference and that's kind of what Grant wanted to model is he he wanted to model if you've got a certain number of predators in an area and a certain number of traps different interaction rates you could model with the traps auto resetting or not you could model what the kill rate is for when the animal actually interacts with the trap and a few other things you want to sort of see if with the interaction rates we're seeing does it mimic what you see in reality and the model seems to be fairly predictive here you can see it doesn't eliminate the predators it just suppresses them which is what I think you see with trapping as you see a reduction in numbers but you don't get down to zero and the interesting thing here is this estimated that the catches per trap per year would be about 3.2 catches per year so you wouldn't so that doesn't sound very many to get only three catches per trap per year but that that's what the model was predicting and we went to Trap NZ and pulled out the data across 60 projects to see how many catches per trap per year are reported on those projects and the best one was six but on average it was sort of around that sort of one to two mark of so each trap only catches one or two predators per trap per year so what we did with the model is we then changed instead if we could get an interaction rate of nine percent what would that look like and it's this red line you get elimination really quickly and so what we want to do is try and test that theory and see well could we create a trap that has a high interaction rate and would it start catching a whole lot of predators and get rid of them so this information that we learned in the modelling we did went in to inform us on what we should do with the trap design and this is the current version of our what we call our high catch rate trap what we did is we looked at the interaction rates that different traps had and we saw that the live capture ones that you put on the ground have the highest interaction rate I think it was about five percent and this sort of makes sense because it's on the ground so it's easier for the predator to get to it's open but of course you have a lot of work to do with live capture traps they have to be checked every day but we think there's other ways to address that so what we've done is we've sort of created an uber version of the live capture trap that is as open as possible and so it has this area we've got a couple of designs but the triangular one it has three open sides to it and a cage at the back and so it's designed to be as open as possible to make it as easy as possible for animals to go into that and then we have blinds that snap up when the animal goes in and the idea is once the animal is in the trap and the blinds go up it will go into the cage at the back so that was the theory so we wanted to test it so in fact the new version we're trying up the trap has two cages and we're testing an auto reset mechanism so the blinds will go back down and we could potentially catch another animal in the second cage we'll talk about what's coming up soon but here's a video of our sort of prototyping and where where this trap has got to so this is an early prototype of the trap so you can see it just the blind snaps shut from a sensor that detects movement within the trap region There's a bird we're not trying to catch birds but of course rats in there hedgehog in there So this is at Living Springs Slowly with the trap's getting a little more refined mechanically we've got to put set up a little bit of hazing to guide the predators into the trap as well so the first test we did is we just wanted to see would predators even go into it so we haven't activated the trap we've just put it out there and we want to see well will the animals go into it you know for a possum like this it's pretty easy for them to climb over it there's a cat it's taking a bit of a sniff makes its way in this is all very encouraging and sort of like okay then oh there's a rat will a rat be able to get over the little lip that we have for the blinds and sure enough it could go into the trap We've got another cat wandering through There's a hedgehog and it could climb over and get into it so that gave us some confidence so now we wanted to say is it fast enough to actually catch these animals so some of these animals have cat like reflexes so a hedgehog yep no surprise there this is a possum and here I think the blinds might have actually hit the possum on the way out but it did catch it there's the cat yep did catch it and the hedgehog The other question is would our theory that the animals would then move into the back cage would that be correct and so here's a test in that so here's a rat that goes straight into the cage at the back which is on the left of the video here did catch it and this one's the possum just sort of going through quietly into the cage here is a cat you can sort of see it running around and it runs into the cage quite quickly obviously upset and so those tests were done sort of at the end of last year and we've gone ahead and built about 10 traps and are testing these at the moment around where we're based in Christchurch we've got Living Springs we're doing a test at Kaitorete Spit working with both Predator Free Banks Peninsula and DOC and we're also testing them in Akaroa and we're catching a whole lot of stuff it's actually pretty exciting because we've got thermal cameras in front of all of these we can sort of log in each morning and see what we've caught and I've actually created a little compilation of some of the catches just from the last few weeks so here's a possum just in front of the trap boom gets caught jumps in the back an animal at the back gets caught and jumps in the trap We have another animal at the back hedgehog gets caught Here we miss the blinds just going up and the possums moving from the trap into the cage we have I think it's a hedgehog going into the trap gets caught I think this is a cat I believe Just looking in the trap inching into it should go in boom there's the blinds go up and the cats gone into the bag There's a possum putting it's foot into the trap boom straight into the back cage Same trap next night almost an action replay Possum goes into the bag Here's a hedgehog just inching it's way snap it's caught Here we have I think it's a possum boom it's caught goes into the trap at the back cage at the back Here's a hedgehog inching into the trap snap it's caught Possum at the back caught runs into the cage straight away Here we have mother and baby possum I think The baby is looking at the trap inches in gets caught and the mother takes off ----------------------- Yeah so there we have it It's been quite exciting watching all these catches happening Grant has one trap that caught 13 predators 13 possums over two and a half weeks and he'd had conventional traps out there for the last year and caught about two and given that the average trap catches one to two predators a year catching 13 in a period of two weeks is really encouraging and we're measuring the interaction rate so how often do we see an animal and how often does it get caught by the trap and we're seeing the interaction rates as high as 50 percent for some animals so we're well above that nine percent threshold that we modelled should be working and so now we want to try and test an area to see whether we can actually get rid of possums or predators in an area and the first one we may work with is Predator Free Banks Peninsula on the Kaitorete Spit but I've got some ideas there but I'll talk about that a little bit First of all I want to take a little deviation and talk about something closer to home here we've been working with the people at Shakespear Park they have had a stoat incursion They contacted us asking if they could rent a camera because they weren't sure they wanted to spend the money on the camera so they rented it for a month They'd had stoat dogs in there and the trail cameras trying to find these stoats that some visitors had seen and within a couple of days this is what they saw so they're very quickly able to find where the stoat is and find evidence of the stoat and now they're trying to catch the thing and struggling to catch it and well I'm just talking to them at the moment about sending a trap up there next week to see whether that trap can catch it They have been actively trying to catch the stoats for some time so watch the space to see what happens with that They've since gone ahead and bought that camera so that was Auckland Council's first camera One of the things you might have seen with the camera is we tried a we had a bit of hazing with the camera with the trap to try and guide the animals into it and we thought what if we extended this idea literally and how long you know could you make this hazing and have animals follow it along and so we sort of come up with this idea of what we call an active fence which it's really cheap it's not a predator free fence by any stretch the imagination but we wanted to know if an animal hits that will they go along it and go through some gaps or will they just jump over it or climb under it which is going to be pretty easy to do It's just a hessian fence with some you know stakes to hold it up and so we did an experiment out at Kaitorete Spit which is where this photo was taken where we put this fence up left some gaps put our cameras out to see what would happen and were pleasantly surprised to see that the animals for the most part would go through those gaps and this is another variation with the gap underneath but this is I think on the Spit so here's a stoat hits the fence and then goes through the gap I think that was a rat against sped up it might have been a hedgehog and that's a possum so with what what we've seen is almost all these animals would go along there so what we've now done is we've now gone along and extended it a bit further and put some traps in those gaps and we're busy catching a whole lot of stuff out at Kaitorete so that's it's a pretty simple idea is just making use of hazing it sort of shows the importance of it for making the effective catch area of a trap much much larger and again even that little gap animals would squeeze their way through the gap you see cats going through and the hedgehogs it's probably possums as well There's even one where the cat comes along and has a look over the fence and then decides to go back through the gap and it's just easier to go through the gap than it is to jump to expend energy to go after it so if there's one take away it's just use more hazing up around your traps and you might make them more effective so what we've got next just quickly to wrap up is we've got work on all of our projects so the low cost monitoring we want to be able to prove that that works at scale with the thermal cameras we're trying at the moment experimenting and blurring at a distance so we are doing playing sounds and we've also got a Phd student who is experimenting with recording sounds and listening to see whether predators are there could that be sort of proof that predators are there or proof of absence as well and we may integrate some of that record some of those sounds to use to act as lures if we see them working we want to continue to work with that (sound lost) to check what sort of animal it is in the traps so we'll only activate it if it's a target animal so we won't be catching the birds and then we also want to integrate a auto kill mechanism as well and then the sort of longer term one is to do a projectile elimination so we're talking about having a paintball gun with poison in the paintball that will target the predator so we'd get them all they wouldn't have to go into a trap we'd just hit them it's a longer term project So that's what we've been up to and what we're planning on doing I'm happy to take any questions that there might be ------------------------------------------- Question 1 (Matt): Great great presentation thanks a lot man Which were the species that you've been looking at through thermal imaging that show you know more reluctance than others to actually go into that trap and what percentage you said something like 50 percent of the animals that approach the trap go in are there any any of the species in particular that show real reluctance to to enter the the trap Shaun: Yeah we it's still too early days to have so good sort of statistically significant numbers that 50 percent was off a relatively small but it was for possums I think with hedgehogs it was maybe around the 20 percent but I think we have also seen numbers sort of down as low as 10 percent as well for maybe rats at the moment we're not really focusing too much on rats because the cages we're using we need to put additional wrapping around to so the rats can't escape and we have just haven't seen that many mustelids to make a call upon what's happening with those hopefully the experiment with Council will help us and we haven't seen many rabbits we've seen rabbits on our cameras near the traps haven't seen any of them go into the traps Matt: One of the things that we tend to find with those large DOC200 boxes is that rats seem to spend a lot of time looking at it and even hopping up on top of the box but not going in they just you know compared with those rat motels which are dark and with the low ceiling provide a really nice sort of an alluring sort of shelter for them those DOC 200s you know just somehow don't sort of tick the boxes and it you know might come down to they're highly suspicious animals aren't they Shaun: yeah yeah and I mean I've got to say our focus probably has been more on the larger animals but that sort of information is really good because maybe we need to make something you know a separate entrance for the rats or something like that that is more alluring to them but we have seen rats they will move along the hazing and go through the through the trap space so so we think with the hazing potentially it could work if it's open enough then it's not like just like walking through an open area Matt: yeah that hazing looks like a great idea it's very similar to you know the the leaders that we use with our fyke nets for you know catching uh catching fish that you know you're far more likely it's a brilliant idea it's a way of sort of concentrating your predators it's great Shaun: yeah yeah really really simple and inexpensive as well Matt: yeah thanks man Shaun: yeah no worries Any other questions there One thing we have done in the absence of questions I'll just tell you we've been working with the Otago Regional Council to detect wallabies automatically they suspect that they're moving from South Canterbury into North Otago across the dams and the bridges and so if we expanded the machine vision to automatically detect wallabies and have put a camera on the Aviemore Dam and a couple weeks ago we caught our first wallaby crossing over the dam so it sort of confirmed their suspicions and we're continuing to work on that Another little feature we're working on is the ability to get an email alert when an animal pops up on the camera so you can almost in real time find out oh I've got a predator in front of this camera it's amazing (sound is lost momentarily) Please contact us at 2040 if you've got any questions about any of the technology Matt: Thanks very much appreciate it Shaun: Thank you.
2021-06-01 17:56
