Predicting the probability of large carnivore occurrence: A strategy to promote crocodile and human coexistence

Abstract

Preserving large carnivores that perceive humans as prey brings conservation values into direct conflict with human security. Informing when and where humans and large carnivores occupy the same space may reduce attack frequency and promote coexistence. Here, we demonstrate a methodology to better understand the spatiotemporal relationship between a population of large carnivores and humans. The carnivore of study was the estuarine crocodile Crocodylus porosus, a large semi-aquatic predator responsible for 705 recorded human attacks over the last 20 years. Crocodiles were captured every August over 3 years and individuals greater than 2.5m in length were implanted with an acoustic transmitter (n=84). The transmitter emitted a coded pulse detected when in proximity to underwater hydrophones deployed throughout the river. The telemetry data informed which previously captured crocodiles were present during subsequent trapping episodes and adult population size was estimated using a closed-population model. Over 3 years, 24 of the tagged crocodiles were detected 269 times moving through a shallow-water area where humans frequently entered the water. The tagged crocodile presence was extrapolated to the population level to provide a probability of adult crocodile presence across a range of temporal scales. The results showed that between September and December, the probability of crocodile presence within the human entry zone was 0.97±0.01 during darkness but decreased to 0.07±0.01 during daylight, except around periods of high tide when it increased to 0.71±0.02. Human visitors confined their activity to shallow water during daylight hours, but no consideration was given to the significant rise in crocodile presence with season and tide. The observed patterns in crocodile and human behaviour, around this shallow-water river crossing, exhibited parallels with historical incidences of crocodile attack.