Combining satellite data and biogeochemical models to estimate global effects of human-induced land cover change on carbon emissions and primary productivity

This study uses a global terrestrial carbon cycle model (the Carnegie-Ames- Stanford Approach (CASA) model), a satellite-derived map of existing vegeta tion, and global maps of natural vegetation to estimate the effects of huma n-induced land cover change on carbon emissions to the atmosphere and net p rimary production. We derived two maps approximating global land cover that would exist for current climate in the absence of human disturbance of the landscape, using a procedure that minimizes disagreements between maps of existing and natural vegetation that represent artifacts in the data. Simil arly, we simulated monthly fields of the Normalized Difference Vegetation I ndex, required as input to CASA, for the undisturbed land cover case. Model results estimate total carbon losses from human-induced land cover changes of 182, and 199 Pg for the two simulations, compared with an estimate of 1 24 Pg for total flux between 1850 and 1990 [Houghton, 1999], suggesting tha t land cover change prior to 1850 accounted for approximately one-third of total carbon emissions from land use change. Estimates of global carbon los s from the two independent methods, the modeling approach used in this pape r and the accounting approach of Houghton [1999], are comparable taking int o account carbon losses from agricultural expansion prior to 1850 estimated at 48-57 Pg. However, estimates of regional carbon losses vary considerabl y, notably in temperate midlatitudes where our estimates indicate higher cu mulative carbon loss. Overall, land cover changes reduced global annual net primary productivity (NPP) by approximately 5%, with large regional variat ions. High-input agriculture in North America and Europe display higher ann ual NPP than the natural vegetation that would exist in the absence of crop land. However, NPP has been depleted in localized areas in South Asia and A frica by up to 90%. These results provide initial crude estimates, limited by the spatial resolution of the data sets used as input to the model and b y the lack of information about transient changes in land cover. The result s suggest that a modeling approach can be used to estimate spatially-explic it effects of land cover change on biosphere-atmosphere interactions.