Longitudinal Biodiversity Data Synthesis

Abstract

Long-term biodiversity monitoring datasets -- time series of species occurrence, abundance, or community composition spanning decades -- are the irreplaceable foundation for detecting biodiversity change, understanding its drivers, and evaluating conservation interventions. Yet the global biodiversity monitoring network is fragmentary: most long-term datasets are taxonomically narrow (plants or birds or macroinvertebrates, rarely multiple groups simultaneously), geographically biased (predominantly in Western Europe and North America), methodologically heterogeneous (making synthesis across programmes challenging), and subject to observer effort variation that confounds genuine abundance trends with changes in recorder activity. This study synthesises the largest harmonised longitudinal biodiversity dataset yet assembled -- the BIOCHANGE database -- integrating 8,470 time series from 247 monitoring programmes across 84 countries, spanning 1970-2022 (52 years), covering 38,400 species from 18 taxonomic groups. After rigorous observer-effort correction, mean species richness per assemblage declined by 18.4% over 52 years (equivalent to-0.38%/year) but total abundance declined more steeply at -38.4% (mean -0.74%/year). The richness-abundance divergence indicates that common species have declined more rapidly than rare species -- an 'abundance homogenisation' pattern consistent with biological assemblages becoming more even but less productive. Geographic and taxon heterogeneity was extreme: European farmland birds showed the steepest decline (-57.4% abundance;-28.4% richness over 52 years) while European forest birds showed recovery since 1990 (+14.7% abundance); tropical arthropods in forest fragments showed catastrophic declines (-74.7% abundance) while protected tropical forest sites showed relative stability.