Changes in precipitation variability are known to influence grassland growth. Field measurements of aboveground net primary productivity (ANPP) in temperate grasslands suggest that both positive and negative asymmetric responses to changes in precipitation may occur. Under normally variable precipitation regimes, wet years typically result in ANPP gains being larger 5 than ANPP declines in dry years (positive asymmetry), whereas increases in ANPP are lower in magnitude in extreme wet years compared to reductions during extreme drought (negative asymmetry). Whether ecosystem models that couple carbonwater system in grasslands are capable of simulating these non-symmetrical ANPP responses is an unresolved question. In this study, we evaluated the simulated responses of temperate grassland primary productivity to scenarios of altered precipitation with fourteen ecosystem models at three sites, Shortgrass Steppe (SGS), Konza Prairie (KNZ) and Stubai Valley meadow 10 (STU), spanning a rainfall gradient from dry to moist. We found that: (1) Gross primary productivity (GPP), NPP, ANPP and belowground NPP (BNPP) showed concave-down nonlinear response curves to altered precipitation in all the models, but with different curvatures and mean values. (2) The slopes of spatial relationships (across sites) between modeled primary productivity and precipitation were steeper than the temporal slopes obtained from inter-annual variations, consistent with empirical data. (3) The asymmetry of the responses of modeled primary productivity under normal inter-annual precipitation 15 variability differed among models, and the median of the model-ensemble suggested a negative asymmetry across the three sites, in contrast to empirical studies. (4) The median sensitivity of modeled productivity to rainfall consistently suggested greater negative impacts with reduced precipitation than positive effects with increased precipitation under extreme conditions. This study indicates that most models overestimate the extent of negative drought effects and/or underestimate the impacts of increased precipitation on primary productivity under normal climate conditions, highlighting the need for improving eco-
20hydrological processes in models.
IntroductionPrecipitation is a key climatic determinant of ecosystem productivity, especially in grasslands which limits productivity over the majority of the globe (Lambers et al., 2008;Sala et al., 1988;Hsu et al., 2012;Beer et al., 2010). Climate models project substantial changes in amounts and frequencies of precipitation regimes worldwide, and this is supported by observational 25 data (Karl and Trenberth, 2003; Donat et al., 2016;Fischer and Knutti, 2016). Potential for increasing occurrence and severity of droughts and increased heavy rainfall events related to global warming will likely affect grassland growth Gherardi and Sala, 2015;Lau et al., 2013;Reichstein et al., 2013). As a consequence, better understanding of the responses of grassland productivity to altered precipitation is needed ...