Habitat change and biased sampling influence estimation of diversity trends

Recent studies have drawn contrasting conclusions about the extent to which local-scale measures of biodiversity are declining and whether such patterns conflict with the global-scale declines that have attracted much attention.¹ A key source of high-quality data for such analyses comes from longitudinal biodiversity studies, which sample a given taxon repeatedly over time at a specific location.² There has been relatively little consideration of how habitat change might lead to biases in the sampling and continuity of biodiversity time series data, and the consequent potential for bias in the biodiversity trends that result. Here, based on analysis of standardized routes from the North American Breeding Bird Survey (3,014 routes sampled over 18 years),³ we demonstrate that major local habitat change is associated with an increase in the rate of survey cessations. We further show that routes that were continued despite major habitat changes show reduced diversity. By simulating potential rates of loss, we show that the underlying real trends in taxonomic, functional, and phylogenetic diversity can even reverse in sign if more than a quarter of diversity is lost from routes that ceased and are thus no longer included in surveys. Our analyses imply that biodiversity loss can be underestimated by biases introduced if continued sampling in longitudinal studies is influenced by local change. We argue that researchers and conservation practitioners should be aware of the potential for bias in such data and seek to use more robust methods to evaluate biodiversity trends and make conservation decisions.