1.As precipitation (PPT) regimes are becoming more extreme worldwide, the inter-annual variation in PPT amount at a given location is becoming greater than ever before. Yet, our understanding of grassland responses to extreme drought and wet years, as well as the pattern of the relationship between PPT and above-ground net primary productivity (ANPP) along a PPT gradient from extreme drought to extreme wetness scenarios, remains unclear, hampering accurate forecasting of the grassland feedbacks to future PPT changes.

2.We addressed this research gap by a 4-year, multi-level PPT gradient experiment that included PPT amounts simulating extremely dry and wet years in combination with a long-term (40 years) observational study in a grassland ecosystem.

3.We found that the ANPP reductions due to extreme droughts were greater than the magnitude of positive responses to extreme wet years. The PPT-ANPP relationship along the gradient of PPT followed a negative asymmetric (concave-down) pattern. We identified the major mechanisms for this pattern were a shift in primary limiting resource for ANPP from water to nitrogen with increasing PPT and different responses of the plant functional groups presented in the community. With the increase of PPT from extremely low to extremely high levels, the relative abundance of forbs increased, while that of grasses decreased and annuals was not significantly affected.

4.This study illustrates how a grassland responded to PPT extremes and highlights the role of shifts in resource limitation in affecting the PPT-ANPP relationship. The negative asymmetric pattern suggests that the reduction of ANPP in extremely drought years cannot be mitigated by its positive response to extremely wet years. Thus, to accurately forecast future ecosystem feedback to climate change, we should take into account the negative asymmetric feature of the ecosystem response to stronger climate variation, as well as recognize that the resource limitations will be fundamentally altered by increasing frequency and intensity of climate extremes in the future.