Análise comparativa da eficiência de gravadores em dois ambientes amazônicos para monitoramento de aves

Abstract

Passive Acoustic Monitoring (PAM) is a fundamental tool for biodiversity studies, but its efficiency critically depends on the effective detection distance, a parameter modulated by the vegetation structure. This study empirically determined this distance for birds in two contrasting environments in Viruá National Park, Roraima: campina (open formation) and dense rainforest (i.e., terra firma). Using autonomous recording units (ARUs, AudioMoth) and playback recordings of 20 bird species, we assessed the probability of detection at distances of 50 to 300 m from the sound source. A Generalized Linear Model (binomial GLM) revealed a dichotomous pattern. In the meadow, detection remained viable at all distances, with a gradual decline up to 300 m. In the dense forest, the detection probability declined sharply, becoming statistically insignificant beyond 200 m, with most detections concentrated in the first 50 m. An assessment by species confirmed this contrast, with drastically reduced ranges in the forest environment. The results demonstrate that there is no universal detection radius and that the sampling design of PAM studies must be calibrated by phytophysiognomy. A larger spacing between ARUs (≥600 m) is recommended in meadows, while in dense forest a smaller spacing (<150 m) is viable for adequate acoustic sampling. This study provides empirical parameters that increase the effectiveness, comparability and ecological robustness of bioacoustic inventories.