A major area of research that has interested me since graduate school is developing a better understanding of the real-world behavior of animals. In the case of bats, I had some success analyzing recordings that included multiple bats to estimate how many bats were actually recorded. This suggested it might be possible to get an estimate of bat population sizes by recording the bats at foraging locations and just analyzing the recordings. This would be better for the bats because it wouldn't require capturing them in nets and stressing them out, and it would be better for the researchers because catching large numbers of bats is really difficult!
The initial work I did with this project involved the Mark I recording system, which is very expensive, thus limiting the number of sites where I could get recordings, since I only had one system. There is also the problem that the recordings took a significant amount of time to analyze, which meant I had no information on what was going on in a timely fashion. I thought there was a lot of promise in developing ways to examine bat behavior without disturbing the animals (which occupied a couple of chapters in my PhD dissertation). When I came to CSU, I wanted to expand on these ideas, but I needed a way to do so that could fit within my time and budgetary constraints.
In 2010, I was reading an article in Make magazine called "Kitty Twitty". In this project the author developed a cat toy using an Arduino that was hooked up to the internet so that that it would send a tweet every time the cat played with it. I started with the same basic premise, but attached a motion sensor to the device. The initial goal was to develop a system that would detect bats entering or leaving a roost and would give a count in that fashion. The initial "Batty Twitty" system worked pretty well, but it only detected motion, which wouldn't be useful for examining bat behavior beyond entering or leaving a roost. I needed something that detected echolocation, which would allow me to examine bat behavior in the field. So I took the output from a bat detector and used it to build Batty Twitty v 2.0:
Batty Twitty 2.0 - Connection to the bat detector is through the yellow wire.
This system waited until an ultrasonic signal was detected by the bat detector, then it would try to count the number of separate pulses (corresponding to how many echolocation sounds were received) in a 30-second time period. By getting an estimate of the number of echolocation calls, you have at least some idea of how much bat activity occurred in a given location. This estimate would be sent out as part of the tweet:
Based on my preliminary results, the system seemed to show a lot of promise, but it was limited by two factors. The bat detector I was using (a Petersson D100) was still fairly expensive (although much cheaper than using the full recording system) and the need to have a hardwired internet connection for sending tweets meant that the system wasn't really suitable for the field. However, based on these results, I was able to get a grant from the Clayton State College of Arts & Sciences, which I used to purchase equipment to update the system including several more Arduinos and materials I would need to overcome these two limitations.
The first problem was getting away from the wired connection and using a wireless system. I was able to get XBEE wireless radios, which can connect to the Arduino and send messages over a distance of several hundred meters. This allowed me set up each Arduino as a monitoring station and have them transmit their information to a base station connected to a computer. The computer could then display real-time updates on bat activity. There are times where the wireless signals do not successfully arrive at the base station, meaning that I could potentially lose data, so I outfitted each Arduino with an SD card so that they can record the data locally in addition to transmitting it to the base station. The image below the Arduino MEGA which is used by the base station with the SD card and XBEE radio.
The second problem was the cost of the bat detector, which I solved by finding a cheaper alternative. A retired engineer by the name of Tony Messina developed something he calls the Simple Bat Detector (SBD). His design has been available since 1997 and he offers kits and circuit boards for it very cheaply. I tested the output of this detector and it worked very well, even compared to the much more expensive D100. At this point I felt like I had something that was going to allow me to get a lot of data in a very short while. Once I knew patterns of bat activity, I could use that to focus my efforts with the more expensive recording system so that I could get the maximum results from my time in the field. In the course of working on this system, I realized that I could take advantage of it for determining real-time bat activity patterns if I could find a way to get the information from the different monitoring stations as it was received. For this, I hooked the based station up to a computer and wrote a program in Processing that can take the data from all monitoring stations and display the data as it is received:
The output includes a time-stamp to indicate when the data was received from the monitoring station, the number of calls that were detected with the size of the colored circle corresponding to the number of calls as well, so that you have a graphical representation that is easy to interpret as it is happening. From this, I would be able to see not just that bats were active, but where they were active, and when, so that I could determine patterns of behavior much more quickly than would be possible if I had to wait for another program to do the analysis. Based on these results I'm currently working on refining the system to produce a set of bat monitoring stations that can be deployed and record data all night while giving real-time updates on bat activity over a fairly large area. With the addition of more monitoring stations, the area for monitoring could be expanded further.