Drone Surveys Revealed Bottom-Up, And Not Top-Down, Effects On The Marsh Deer Local Abundance

Abstract

Context: Spatial variation in large herbivore populations can be highly affected by the availability of resources (bottom-up) but modulated by the presence of predators (top-down). Studying the relative influence of these forces has been a major topic of interest in ecological and conservation research, while it has also been challenging to sample large herbivores. Objective: i) Explore the use of spatiotemporally replicated drone-based counts analysed with N-mixture models to estimate abundance of large herbivores. ii) Evaluate the relative influence of bottom-up (forage and water) and top-down (jaguars) processes on the local abundance of the threatened marsh deer.Methods: We conducted spatiotemporally replicated drone flights in the dry season of Pantanal wetland (Brazil) and imagery was reviewed by either one or two observers. We fitted counts using N-mixture models (for single and double observer protocols) and modelled local abundance in relation to vegetation greenness, distance to water bodies, and jaguar density.Results: We found a positive relationship of marsh deer local abundance with vegetation greenness, a negative relationship with distance to water, but no relation with jaguar density. Individuals were concentrated in the lower and wetter region, even though it is the area expected to be more lethal from jaguar predation.Conclusions: Bottom-up processes are shaping the distribution of marsh deer in the dry season; the benefits of accessing high-quality areas outweigh predation risk from jaguars. Spatiotemporally replicated drone-based counts may serve as an accessible and cost-effective protocol for large herbivores abundance estimation and monitoring while accounting for imperfect detection.

Optimally designing drone‐based surveys for wildlife abundance estimation with N‐mixture models

“…It is important to note that, in "real-case" studies such as the marsh deer example, when modeling the spatial variation in local abundance using covariates (e.g. Brack et al, 2021), the optimal number of visits would be pushed towards lower numbers of visits, to cover more sites.…”

“…Andriolo et al, 2005; Mourão et al, 2000; Ríos‐Uzeda & Mourão, 2012). In 2017, Brack et al (2021) conducted spatiotemporally replicated drone‐based count surveys to estimate the abundance of marsh deer in the Pantanal wetland (Sesc Pantanal Private Natural Reserve; 108,000 ha) and explored the use of this method to monitor that species. Six flight paths (32–42 km) were flown from two to six times each using a fixed‐wing drone equipped with an RGB camera.…”

“…For the purpose of using N‐mixture models for drone‐based counts, we fixed the perception probability p at 0.8, considering what we have seen as a moderately low perception threshold in studies with drone surveys with both human observers (e.g. Brack et al, 2021; Patterson et al, 2016; Preston et al, 2021; Vermeulen et al, 2013) and algorithms (e.g. Dujon et al, 2021; Eikelboom et al, 2019; Kellenberger et al, 2018).…”