Preliminary analysis of acoustic detection of the Red-throated Caracara in northern Costa Rica

Abstract

The noisy Red-Throated Caracara (Ibycter americanus) is a species whose population has inexplicably declined across much of its range and is now rare in the Pacific and Caribbean slopes of Costa Rica. Advances in automatic acoustic detection have transformed bird ecology, allowing researchers to analyze bird populations using pattern matching algorithms, machine learning, and random forest models. Although these studies are limited in the country, it represents an area with great interdisciplinary potential for technological advances. This study focused on the use of Pattern Matching to detect the presence of the Red-Throated Caracara in northern Costa Rica using a large number of sound recordings and its validation with metrics such as Accuracy, Precision Negative predictive value, Sensitivity, Specificity and Unweighted average recall. The results showed a moderate performance of the model by obtaining accuracy and precision values of 0.71 compared to the values obtained in other investigations in which the reported model was used. Therefore, we suggest exploring new techniques and methods to improve the detection of the species, considering the particular acoustic structure, the repertoire of sounds of the species and similarities with vocalizations of other species. This similarity could indicate a supposed anti-predator defense behavior by “imitating” the sounds of other species with which it shares habitat. To optimize this acoustic detection, we recommend using complementary techniques such as noise filters that improve the quality and precision of the data.