PS 74-135
Using TuneR to develop detectors to aid in the analysis of calling secretive marsh birds in data collected from SongMeter remote acoustic recorders

Thursday, August 13, 2015
Exhibit Hall, Baltimore Convention Center
Timothy Raymond Freiday, Entomology and Wildlife Ecology, University of Delaware, Del Haven, NJ
W. Gregory Shriver, Entomology & Wildlife Ecology, University of Delaware, Newark, DE
Sasha Hafner, University of Southern Denmark

Secretive marsh birds (SMBs) are a guild facing numerous threats, and reliable information on population status and trends are crucial in aiding management decisions.  Many estimates of density are derived from point count data that account for imperfect detection.  However, little is known about the effect that observer presence has on detection probability of SMBs.  It is our hope to utilize data obtained from SongMeter remote recording units to quantify this observer effect on SMB detection probability.  SongMeter units were deployed at 32 point count survey locations in 2014 prior to, during and after conducting surveys at those points using the North American Marsh Bird Monitoring protocol.  We will use detectors developed in the package TuneR in programming language R to quantify differences in call rates of SMBs at point count locations that may be due to observer presence at those sites.  If there is an observer effect on call rates then we will quantify how this changes the detection probability of these species.  This information will help to increase the accuracy of density estimates of SMBs derived from point count surveys.


We have currently developed a detector in TuneR which is 76% effective at identifying the kek calls of Clapper and King Rails, and which produces very few false positives.  The imperfection in the detector is the result of distant calls which are not as prominent on the sonogram, and thus not picked up by the detector.  As such, we may only be able to use calls given closer than 100m for analysis.  It is our hope to refine this detector in order to maximize detections and minimize false positives.  We will also develop detectors for different call types and species in order to maximize the amount of information obtained from our SongMeter data.  Once we have fully developed our suite of detectors, we will analyze our acoustic data and make inferences on how detection probability of SMBs is affected by observer presence at point count survey points.