Forecasts for 80% of wild populations and communities predict no change, and why

Abstract

Many retrospective analyses of biodiversity trends reveal a complex picture, including significant declines but also increases. However, prospective analyses (i.e. what near-term trends to expect) remain comparatively rare. Here, we generate and assess forecasts for approximately 43,000 population-level and 10,000 community-level time series from localities across the globe and covering taxa across the Tree of Life. Using unobserved components models and model selection, we found that only 23.5% of wild populations exhibit forecasted time trends (with increasing and decreasing trends roughly equal), while 76.5% show a forecasted constant mean. For communities (temporal - and {beta}-diversity), this figure rises to 84.0%. We identified high variation in the rate of change, var(RoC), as the most important predictor of a forecasted constant mean, and it scaled more strongly with intrinsic ecological dynamics than observation noise. Furthermore, biological structure (taxonomy, life-history traits, and locality) contributed to the identification of trends, highlighting that biological factors underlie, in part, predictions of trends. In summary, despite time trends being identifiable in ecological time-series data, parsimonious statistical models will very often forecast no change. Our results suggest that the absence of directional change may be the norm rather than the exception across the Tree of Life.