Intra-annual and interannual variability of ecosystem processes in shortgrass steppe

We used a daily time step ecosystem model (DAYCENT) to simulate ecosystem p rocesses at a daily, biweekly, monthly, and annual time step. The model eff ectively represented variability of ecosystem processes at each of these ti mescales. Evolution of CO2 and N2O, NPP, and net N mineralization were more responsive to variation in precipitation than temperature, while a combine d temperature-moisture decomposition factor (DEFAC) was a better predictor than either component alone. Having established the efficacy of CENTURY at representing ecosystem processes at multiple timescales, we used the model to explore interannual variability over the period 1949-1996 using actual d aily climate data. Precipitation was more variable than temperature over th is period, and our most variable responses were in CO2 flux and NEP. Net ec osystem production averaged 6 g C m(-2) yr and varied by 100% over the simu lation period. We found no reliable predictors of NEP when compared directl y, but when we considered NEP to be lagged by 1 year, predictive power impr oved. It is clear from our study that NEP is highly variable and difficult to predict. The emerging availability of system-level C balance data from a network of flux towers will not only be an invaluable source of informatio n for assessments of global carbon balance but also a rigorous test for eco system models.