Experimental drought consistently underestimates productivity responses to natural drought in four Central US grasslands

Climate change is increasing the frequency and severity of droughts globally, and grasslands are particularly vulnerable to such hydrological extremes. Drought effects at the ecosystem scale have been assessed both experimentally and through the study of naturally occurring drought, with emerging evidence that the magnitude of drought effects may vary depending on the approach used. We took advantage of a decadal study of four grasslands to directly contrast responses of aboveground net primary productivity (ANPP) to simulated vs. natural drought. The grasslands spanned a ~ threefold mean annual precipitation gradient (335-857 mm) and were all subjected to a natural 1-year drought (~ 40% reduction in precipitation from the long-term mean) and a 4 year experimental drought (~ 50% precipitation reduction). We expected that the 4 year drought would reduce ANPP more, and that post-drought recovery would be delayed, compared to the 1-year drought. We found instead that the short-term natural drought reduced ANPP more strongly than the simulated drought in all grasslands (~ 10 to ~ 50%) likely due to the co-occurrence of higher temperatures and vapor pressure deficits with reduced precipitation. Post-drought recovery was site specific and each site differed in their recovery from the natural and experimental droughts. These results align with past analyses that experiments that only manipulate soil moisture likely underestimate the magnitude of natural drought events. However, experiments can provide valuable insight into the relative sensitivity of ecosystems to reduced precipitation and soil moisture, a key aspect of drought.