Data fusion for abundance estimation: community science augments systematically collected removal-in-time distance sampling data

Ecologists use a variety of systematically and opportunistically sampled count data to estimate bird abundance, and integrating or fusing different datasets has emerged as a critical challenge in recent years. While previous work provides data integration methodology for occupancy (presence/absence) estimation, methods for abundance estimation that account for imperfect detection and disparate survey protocols remains an active area of research. Here we show how to integrate systematically collected removal-in-time distance sampling data from the Integrated Monitoring in Bird Conservation Regions (IMBCR) program with North American Breeding Bird Survey (BBS) point counts and eBird community science observations. Using the Grasshopper Sparrow (Ammodramus savannarum) in the Great Plains of the United States as a focal species, we demonstrate that BBS and eBird data improve predictive performance for IMBCR count data, providing more spatially refined and precise estimates of abundance at regional scales. Data fusion increased predictive performance even despite relatively weak spatial correlations among data sets. The methodology developed here provides a principled way to fuse data when estimating abundance with distance sampling, that accounts for imperfect detection and variable effort.