Ecosystem stability is shaped by resource-trait coordination under multiple interacting global change factors.

Predicting ecosystem productivity and its temporal stability under multiple global change factors is challenging, especially as most experiments so far have manipulated single drivers over short timescales. Here using a 13-year fully factorial field experiment manipulating elevated atmospheric CO2, nitrogen enrichment, warming and reduced rainfall, we show that grassland ecosystem stability is shaped by distinct single-driver effects and strongly non-additive, time-dependent interactions. At ambient levels of all other factors, elevated CO2 and, less markedly, nitrogen enrichment reduced stability by increasing temporal standard deviation more than mean productivity. In contrast, warming and reduced rainfall alone both increased stability through opposite mechanisms, with reduced rainfall suppressing temporal standard deviation and warming enhancing mean productivity disproportionately. Combined driver effects frequently shifted in magnitude and, in some cases, reversed their expected additive direction over time, revealing dynamics that would be missed in short-term studies. Across treatments, stability was governed primarily by species asynchrony, with secondary contributions from soil moisture and functional composition. Together, these findings demonstrate that ecosystem stability results from shifting, non-additive interactions among global change drivers, challenging inferences drawn from single-factor global change studies.