Evaluating citizen vs. professional data for modelling distributions of a rare squirrel. Issue 2 (27th May 2016)

• Record Type: Journal Article
• Authors: Tye, Courtney A.; McCleery, Robert A.; Fletcher, Robert J.; Greene, Daniel U.; Butryn, Ryan S.
• Editors: Wiersma, Yolanda
• Is Part Of: Journal of applied ecology. Volume 54:Issue 2(2017:Apr.)
• Page Start: 628
• Page End: 637
• Publication Date: 2016-05-27
• Subjects: citizen science -- data aggregation -- fox squirrel -- habitat relationships -- opportunistic data -- predictive performance -- road bias -- sample selection bias -- species distribution models -- validation; Agriculture -- Periodicals; Biology, Economic -- Periodicals; Agricultural ecology -- Periodicals; Applied ecology -- Periodicals; 577
• DOI: 10.1111/1365-2664.12682
• Languages: English
• ISSNs: 0021-8901

Abstract

Summary: To realize the potential of citizens to contribute to conservation efforts through the acquisition of data for broad‐scale species distribution models, scientists need to understand and minimize the influences of commonly observed sample selection bias on model performance. Yet evaluating these data with independent, planned surveys is rare, even though such evaluation is necessary for understanding and applying data to conservation decisions. We used the state‐listed fox squirrel Sciurus niger in Florida, USA, to interpret the performance of models created with opportunistic observations from citizens and professionals by validating models with independent, planned surveys. Data from both citizens and professionals showed sample selection bias with more observations within 50 m of a road. While these groups showed similar sample selection bias in reference to roads, there were clear differences in the spatial coverage of the groups, with citizens observing fox squirrels more frequently in developed areas. Based on predictions at planned field surveys sites, models developed from citizens generally performed similarly to those developed with data collected by professionals. Accounting for potential sample selection bias in models, either through the use of covariates or via aggregating data into home range size grids, provided only slight increases in model performance. Synthesis and applications . Despite sample selection biases, over a broad spatial scale opportunistic citizen data provided reliable predictions and estimates of habitat relationships needed to advance conservation efforts. Our results suggest that the use of professionals may not be needed in volunteer programmes used to determine the distribution of species of conservation interest across broad spatial scales.