Large-scale passive-acoustics-based population estimation of African forest elephants

Abstract

African forest elephants (Loxodonta cyclotis) live in tropical rainforests in Central Africa and often use low-frequency vocalizations for long-distance communication and coordination of group activities. There is great interest in monitoring population size in this species; however, the dense rainforest canopy severely limits visibility, making it difficult to estimate abundance using traditional methods such as aerial surveys. Passive acoustic monitoring offers an alternative approach of estimating its abundance in a low visibility environment. The work we present here can be divided into three steps. First, we apply an automatic elephant call detector, which enables the processing of large-scale acoustic signals in a reasonable amount of time. Second, we apply a density estimation method we designed for a single microphone. Because microphones are often positioned far apart in order to cover a large area in the rainforest, meaning that the same call will not produce multiple arrivals on different recording units. Lastly, we examine results from our historic data across five years in six locations in central Africa, which includes over 1000 days of sound stream. We will address the feasibility of long-term population monitoring and also the potential impact of human activity on elephant calling behavior.