Aridity-dependent response bias of grassland productivity to precipitation in experiments compared to observations

Precipitation manipulation experiments and long-term observations have been widely employed to examine the response of ecosystem productivity to changing precipitation. However, whether experiments accurately reflect natural responses remains questionable. Here, we conducted a global reality check analysis using paired datasets from 22 grassland sites that included both experimental and observational measurements. The results revealed that the sensitivity of aboveground net primary productivity (ANPP) to changing precipitation was lower in the experiments than in the observations in xeric grasslands. One potential reason is that the experiments fail to simulate the concurrent changes in air humidity that typically accompany natural precipitation events, thereby leading to the underestimation of the changing precipitation effect on ANPP. Conversely, in mesic grasslands, the ANPP sensitivity to experimental changing precipitation was greater than that in the observations likely because low radiation and temperature during the natural raining days have negative effects on ANPP when water is not limited. This study highlights the aridity-dependent bias of changing precipitation effects on ANPP in precipitation manipulation experiments in grasslands, which should be fully considered when conclusions from experiments are incorporated in model simulations.