The objective of this study was to simulate dynamically the response of a complex landscape, containing forests, savannas, and grasslands, to potential climate change. Thus, it was essential to simulate accurately the competition for light and water between trees and grasses. Accurate representation of water competition requires simulating the appropriate vertical root distribution and soil water content. The importance of different rooting depths in structuring savannas has long been debated. In simulating this complex landscape, we examined alternative hypotheses of tree and grass vertical root distribution and the importance of fire as a disturbance, as they influence savanna dynamics under historical and changing climates. MC1, a new dynamic vegetation model, was used to estimate the distribution of vegetation and associated carbon and nutrient fluxes for Wind Cave National Park, South Dakota, USA. MC1 consists of three linked modules simulating biogeography, biogeochemistry, and fire disturbance. This new tool allows us to document how changes in rooting patterns may affect production, fire frequency, and whether or not current vegetation types and life form mixtures can be sustained at the same location or would be replaced by others. Because climate change may intensify resource deficiencies, it will probably affect allocation of resources to roots and their distribution through the soil profile. We manipulated the rooting depth of two life-forms, trees and grasses, that are competing for water. We then assessed the importance of variable rooting depth on ecosystem processes and vegetation distribution by running MC1 for historical climate (1895-1994) and a GCM-simulated future scenario (1995-2094). Deeply rooted trees caused higher tree productivity, lower grass productivity, and longer fire return intervals. When trees were shallowly rooted, grass productivity exceeded that of trees even if total grass biomass was only one-third to one-fourth that of trees. Deeply rooted grasses developed extensive root systems that increased N uptake and the input of litter into soil organic matter pools. Shallowly rooted grasses produced smaller soil carbon pools. Under the climate change scenario, NPP and live biomass increased for grasses and decreased for trees, and total soil organic matter decreased. Changes in the size of biogeochemical pools produced by the climate change scenario were overwhelmed by the range of responses across the four rooting configurations. Deeply rooted grasses grew larger than shallowly rooted ones, and deeply rooted trees out competed grasses for resources. In both historical and future scenarios, fire was required for the coexistence of trees and grasses when deep soil water was available to trees. Consistent changes in fire frequency and intensity were simulated during the climate change scenario: more fires occurred because higher temperatures resulted in decreased fuel moisture. Fire also increased in the deeply rooted grass configurations because grass biomass, which serves as a fine fuel source, was relatively high.