Abstract
Monitoring biodiversity through time is critical for understanding the impacts of global change, yet the reliability of available time-series for estimating population trends remains poorly studied. We simulated population dynamics to study how monitoring duration, inter-annual variability, and temporal autocorrelation influence trend reliability. We then inferred the expected reliability of 16,450 empirical time-series of terrestrial animals based on those simulations. Longer, less variable, and negatively autocorrelated time-series produced more reliable trend estimates. Time-series shorter than 20 years led to large magnitude errors and low power in estimating population trends, despite generally low sign uncertainty. Overall, 80% of the empirical time-series were too short and/or noisy to identify meaningful temporal trends, yet reliability varies among taxa. Our results show important limitations in the available biodiversity data, stressing the need to sustain long-term monitoring to better estimate population trends, a top priority in the current biodiversity crisis.