Experimental warming altered rates of carbon processes, allocation, and carbon storage in a tallgrass prairie

Climate warming affects ecosystem functioning by altering the rates of carbon (C) fixation and release. Modeling warming effect on terrestrial C cycling is critical given the feedbacks between climate and C cycling. However, the effect of warming on key model parameters and the resulting long-term C dynamics has not been carefully examined. In this study, measurements from a nine-year warming experimental site in a tallgrass prairie were assimilated into a terrestrial ecosystem C cycle model to assess warming effect on key model parameters and to quantify uncertainties of long-term C projection. Warming decreased allocation of gross primary production (GPP) to shoot, and turnover rate of the live C pools (i.e., shoot and root C), but increased the turnover rates of litter and fast soil C pools. Consequently, warming increased live C pools, but decreased litter and soil C pools, and overall decreased total ecosystem C in a 90-year model projection. Information content gained from assimilated datasets was much greater for plant, litter and fast soil C pools than for slow and passive soil C pools. Sensitivity analysis revealed that fast turnover C pools were most sensitive to their turnover rates and modest to C-input related parameters on both short-term and long-term time scales. However, slow turnover C pools were sensitive to turnover rate and C input in long-term prediction, not in short-term prediction. As a result, total soil and ecosystem C pools were generally insensitive to any parameter in short term, but determined by turnover rates of the fast, slow and passive soil C and transfer coefficients from upstream C to slow and passive C pools. Our findings suggest that data assimilation is an effective tool to explore the effect of warming on C dynamics; the nine-year field data contribute more information for the fast C processes than for the slow C processes; and C cycle model parameters change with warming, and models need to account for that phenomenon not to produce bias in C projections. However, warming-induced changes in parameter values also suggest that some important ecosystem processes may be missing or not adequately represented in the ecosystem C models.