Carbon balance of the terrestrial biosphere in the twentieth century: Analyses of CO2, climate and land use effects with four process-based ecosystemmodels

The concurrent effects of increasing atmospheric CO2 concentration, climate variability, and cropland establishment and abandonment on terrestrial car bon storage between 1920 and 1992 were assessed using a standard simulation protocol with four process-based terrestrial biosphere models. Over the lo ng-term (1920-1992), the simulations yielded a time history of terrestrial uptake that is consistent (within the uncertainty) with a long-term analysi s based on ice core and atmospheric CO2 data. Up to 1958, three of four ana lyses indicated a net release of carbon from terrestrial ecosystems to the atmosphere caused by cropland establishment. After 1958, all analyses indic ate a net uptake of carbon by terrestrial ecosystems, primarily because of the physiological effects of rapidly rising atmospheric CO2. During the 198 0s the simulations indicate that terrestrial ecosystems stored between 0.3 and 1.5 Pg C yr(-1), which is within the uncertainty of analysis based on C O2 and O-2 budgets. Three of the four models indicated tin accordance with O-2 evidence) that the tropics were approximately neutral while a net sink existed in ecosystems north of the tropics. Although all of the models agre e that the long-term effect of climate on carbon storage has been small rel ative to the effects of increasing atmospheric CO2 and land use, the models disagree as to whether climate variability and change in the twentieth cen tury has promoted carbon storage or release. Simulated interannual variabil ity from 1958 generally reproduced the El Nino/Southern Oscillation (ENSO)- scale variability in the atmospheric CO2 increase, but there were substanti al differences in the magnitude of interannual variability simulated by the models. The analysis of the ability of the models to simulate the changing amplitude of the seasonal cycle of atmospheric CO2 suggested that the obse rved trend may be a consequence of CO2 effects, climate variability, land u se changes, or a combination of these effects. The next steps for improving the process-based simulation of historical terrestrial carbon include (1) the transfer of insight gained from stand-level process studies to improve the sensitivity of simulated carbon storage responses to changes in CO2 and climate, (2) improvements in the data sets used to drive the models so tha t they incorporate the timing, extent, and types of major disturbances, (3) the enhancement of the models so that they consider major crop types and m anagement schemes, (4) development of data sets that identify the spatial e xtent of major crop types and management schemes through time, and (5) the consideration of the effects of anthropogenic nitrogen deposition. The eval uation of the performance of the models in the context of a more complete c onsideration of the factors influencing historical terrestrial carbon dynam ics is important for reducing uncertainties in representing the role of ter restrial ecosystems in future projections of the Earth system.