Park rangers worldwide are tasked with patrolling protected areas to prevent illegal activities, a challenge compounded by the vastness of the land compared to the available manpower. This is exacerbated by the dense vegetation, the rugged terrain, and the adaptability of lawbreakers, compounded by a lack of historical data. To address these constraints, collaborative efforts between park rangers and the computer science community have led to theoretical advances and novel practical implementations. However, the unique features of tropical parks, including vegetation, geography, and the behavior of illegal actors, require adaptation of currently available tools. In this work, we propose two key contributions: employing combinatorial multi-armed bandits for sequential patrol route suggestions, tailored to the trail networks common in Latin American parks, and integrating our algorithm with an acoustic monitoring system. We provide theoretical performance guarantees as well as computational simulations, using information provided by the Jama-Coaque Ecological Reserve (JCR) in Ecuador.