A Comparison of Current Anuran Monitoring Methods with Emphasis on the Accuracy of Automatic Vocalization Detection Software

Currently, a variety of methods are available to monitor anurans, and little standardization of methods exists. New methods to monitor anurans have become available over the past twenty years, including PVC pipe arrays used for tree frog capture and Automated Digital Recording Systems (ADRS) used to remotely monitor calling activity. In addition to ADRS, machine-learning computer software, automated vocalization recognition software (AVRS), has been developed to automatically detect vocalizations within digital sound recordings. The use of a combination of ADRS and AVRS shows the promise to reduce the number of people, time, and resources needed for an effective call survey program. However, little research exists that uses the described tools for wildlife monitoring, especially for anuran monitoring. In the study, there were two problems addressed relating to AVRS. The first was the poorly understood relationship between auditory survey methods and physical survey methods. I tested this problem by using current auditory monitoring methods, ADRS and the AVRS Song Scope© (v.3.1), alongside more traditional physical monitoring methods that included drift fences, a PVC pipe array, and visual encounter transects. No significant relationship between physical and auditory community population measures was found. Auditory methods were also effective in the detection of call characteristic differences between urban and rural locations, further suggesting an influence of noise pollution. The second problem addressed was the call identification errors found in auditory survey methods. I examined the influence of treatments including the ADRS location, listener group, species, and season on the error rates of the AVRS Song Scope© (v.3.1) and groups of human listeners. Computer error rates were higher than human listeners, yet less affected by the treatments. Both studies suggested that AVRS was a viable method to monitor anuran populations, but the choice of methods should be dependent upon the species of interest and the objectives of the study