Combining acoustic survey and citizen science data yields enhanced species distribution models for tropical rainforest birds.

A key goal in ecology is to develop effective ways to understand species' distributions in order to facilitate both their study and conservation. Many species distribution modeling analyses have been performed using either structured survey data or unstructured citizen science data; these two pools of data have tradeoffs in terms of data density, spatiotemporal coverage, and accuracy. Recent studies have shown that combining structured and unstructured survey data can improve the accuracy of species distribution models for birds, but most of this work has focused on north temperate bird species, using bird atlas data that are less available in the Tropics. Here, we adapted a data pooling approach from the literature on north temperate bird biology to create distribution models for a selection of secretive suboscine bird species that occur in a highly diverse region of the southwestern Amazon. Our approach combined automated acoustic monitoring detections and eBird citizen science data available for the region as well as a high resolution land cover dataset of the region's key ecological gradients. The pooled models outperformed models constructed solely with eBird data for predicting fine grain species responses to habitat gradients in intact forest, but also retained information from the citizen science dataset about species occurrence patterns in non-vegetated areas away from intact forest, including those subject to human disturbance. We present this hybrid approach as a flexible and repeatable means to produce inferences that would not easily be achievable using a single data source, and provide recommendations for other researchers seeking to replicate these methods in Amazonia as well as in other tropical regions.