Katherine Avery
Artificial neural networks (ANNs) are an effective tool for automatically locating bird roosts using radar. NEXt generation weather RADars (NEXRAD) are designed to collect data on weather, but they also pick up information on moving, airborne objects, including birds. NEXRAD helps ecologists to detect bird roost location, but this data is tedious to process manually. Therefore, ANNs detected the roosts automatically from a NEXRAD image dataset of purple martin and tree swallow roosts in the eastern U.S. Four types of radar field images, including reflectivity, velocity, Rho HV, and Zdr, were useful for finding roosts. The ANN achieved an accuracy, true positive rate, and true negative rate of around 80 percent each, showing that this method has potential as a tool for roost detection. Convolutional neural networks (CNNs), a type of ANN, were found to perform better than the traditional ANNs, achieving an accuracy, true positive rate, and true negative rate of over 90 percent each.
Even if you’re intimidated by reaching out to people, do it anyway because they often surprise you. I sent Dr. McGovern a cold email during first semester of my freshman year and asked if I could join her lab. I hadn’t met her before, but I saw her on the computer science faculty page, and I thought that her work looked interesting. Unexpectedly, she said yes, and I started doing research with her right away.
Awards: Goldwater Scholarship Honorable Mention; National Center for Women & Information Technology (NCWIT) Collegiate Award Honorable Mention;
Conferences: 98th Annual Meeting of the American Meteorological Society; Austin, TX; 7-11 January 2018; funded by the University of Oklahoma; National Conference of Undergraduate Research; Edmond, OK; 4-7 April 2018; funded by NSF (through LSAMP) Undergraduate Research Day; Norman, OK; 7 April 2018; Curiosity to Creativity Symposium; Norman, OK; 25 April 2018.
Avery, K., 2018: Automated Detection of Bird Roosts Using NEXRAD Radar Data and Artificial Neural Networks. Honor’s thesis, The University of Oklahoma, Norman, OK.
Chilson, C., K. Avery, A. McGovern, E. Bridge, D. Sheldon, and J. Kelly, 2018: Automated Detection of Bird Roosts Using NEXRAD Radar Data and Convolutional Neural Networks. Remote Sensing and Ecology and Conservation, submitted.