Biodiversity increases the forecastability of species abundances in changing environments

Accurate forecasts of ecological dynamics are critical for ecosystem management and conservation, yet the drivers of forecastability are poorly understood. Here we show that experiments are a powerful (but underutilized) tool to explore the limits of ecological forecasting. We conducted a long-term microcosm experiment with aquatic protists and manipulated two potential drivers of forecastability: biodiversity and environmental change (light). We applied data-driven forecasting methods to finely-resolved time series of species abundances and ecosystem properties. The experimental treatments had interactive effects on abundance forecasts. Specifically, biodiversity increased forecastability in changing environments but reduced forecastability in constant environments. However, forecasts of ecosystem properties were only weakly affected by the experimental treatments. Our results indicate that the combination of biodiversity loss and environmental change could make populations particularly unpredictable. The implications for ecological conservation are worrying: diversity loss might impair decision-taking by reducing the forecastability of ecological dynamics in changing environments.