Carbon balance of the terrestrial biosphere in the Twentieth Century: Analyses of CO₂, climate and land use effects with four process‐based ecosystem models

The concurrent effects of increasing atmospheric CO 2 concentration, climate variability, and cropland establishment and abandonment on terrestrial carbon storage between 1920 and 1992 were assessed using a standard simulation protocol with four process‐based terrestrial biosphere models. Over the long‐term(1920–1992), the simulations yielded a time history of terrestrial uptake that is consistent (within the uncertainty) with a long‐term analysis based on ice core and atmospheric CO 2 data. Up to 1958, three of four analyses indicated a net release of carbon from terrestrial ecosystems to the atmosphere caused by cropland establishment. After 1958, all analyses indicate a net uptake of carbon by terrestrial ecosystems, primarily because of the physiological effects of rapidly rising atmospheric CO 2 . During the 1980s 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 CO 2 and O 2 budgets. Three of the four models indicated (in 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 agree that the long‐term effect of climate on carbon storage has been small relative to the effects of increasing atmospheric CO 2 and land use, the models disagree as to whether climate variability and change in the twentieth century has promoted carbon storage or release. Simulated interannual variability from 1958 generally reproduced the El Niño/Southern Oscillation (ENSO)‐scale variability in the atmospheric CO 2 increase, but there were substantial 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 CO 2 suggested that the observed trend may be a consequence of CO 2 effects, climate variability, land use 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 CO 2 and climate, (2) improvements in the data sets used to drive the models so that they incorporate the timing, extent, and types of major disturbances, (3) the enhancement of the models so that they consider major crop types and management schemes, (4) development of data sets that identify the spatial extent of major crop types and management schemes through time, and (5) the consideration of the effects of anthropogenic nitrogen deposition. The evaluation of the performance of the models in the context of a more complete consideration of the factors influencing historical terrestrial carbon dynamics is important for reducing uncertainties in representing the role of terrestrial ecosystems in future projections of the Earth system.