Today marks the start of day 10 of our trip—just 46 more days to go! We are now near the tiny town of Freedom, Oklahoma, camping at Alabaster Caverns and driving to the Selman Bat Cave each night for recordings. We are in a record heat wave, but in good spirits. We try to get out of the heat during the hottest part of the day and stay cool otherwise thanks to a shady campsite and those magical cooling towels (Best. Investment. Ever.)
By now we have found our routine and, as one of our cave hosts mentioned, we are like a well-oiled machine. We arrive to each cave at 8:15 and immediately get all our gear set up: Cassi gets her environmental station organized and starts her recordings promptly at 8:30. She then takes recordings every 2 minutes until the bats emerge, and notes the exact time of the first few bats emerging.
Steph helps me get the FLIR camera set up, which includes running extension cords from our large power bank in the car. She also then helps babysit me and pass me equipment as I climb down into the canyon to set up additional cameras and microphones.
As soon as emergence is over, we snap into action and pack everything up. Although we arrive back at camp after midnight, our night is not over! Our image extraction algorithms take hours to process, so we get everything set up so it will run overnight. This means we set up a power station inside the car, download the videos from the camera, export the videos to a different format, determine the average time it takes 100 bats to cross our field of view (our frame sampling interval for counting), then set up our frame extraction algorithm to run.
Once everyone is awake, well fed, and well-caffeinated, we go to work. While I start processing and analyzing our acoustic files, Cassi and Steph start the bat counting using ImageJ. They work together to determine noise tolerances and thresholds, then they independently run a script to automatically count each bat in each video frame (usually 15,000 frames or more). They then compare their results, and adjust parameters until they both get the same number. Finally, they take 100 random frames and hand-count the bats to compare that number to what ImageJ counted. We usually get between 2-6% error, with ImageJ giving us an underestimate of the population.
Since we are interested in how flight density affects acoustics, for all our work the first step is determining how many bats are leaving at each time in the emergence. An added benefit is we can give the cave managers an accurate population estimate for their cave!
It’s been really fun for me to watch Steph and Cassi work together. First of all, these girls are such hard workers. Even though we arrive back from the cave exhausted, they jump on the first step of image extraction without any complaining. Also, watching them troubleshoot any problems together is also wonderful. One of the biggest joys as an instructor is when you get to watch your students solve their own problems without asking for your help. When you can watch them really think and work through problems, you know true learning is taking place.