Acoustic detection of Red-capped Manakin (Ceratopipra mentalis) in Sarapiquí, Costa Rica.

Abstract

Most of the vast biodiversity of the Caribbean foothills of Costa Rica is poorly studied. Passive acoustic monitoring is useful for collecting large datasets of wildlife and apply them for conservation initiatives, but handling these large datasets can be impractical if automated detection algorithms are not applied to reduce labeling time. The use of Pattern Machine (PM) and Random Forest Models (RFM) helps to identify single species though a dataset, simplifying the work. We sampled a private reserve in the Caribbean foothills of Costa Rica with 10 Audiomoth recorders. We used PM and RFM to label the presence of the Red-capped Manakin (Ceratopipra mentalis) in the recordings based on a portion of its distinctive vocalization. The RFM scored 0.86 of accuracy and 0.82 of precision to detect the species within the recordings. Three sites showed the highestnumber of detections. The acoustic activity of the bird got three different peaks during the day. The peak of detections took place in March. Based on the results, we recommended several conservation actions to the reserve on how to reduce the impact of tourism in the trails and how to rehabilitate the landscape with assistance of the bird and its activity as seed disperser.