This thesis describes the development and selected applications of a global vegetation model, BIOME4. The model is applied to problems in high-latitude vegetation distribution and climate, trace gas production, and isotope biogeochemistry. It demonstrates how a modeling approach, based on principles of plant physiology and ecology, can be applied to interdisciplinary problems that cannot be adequately addressed by direct observations or experiments. The work is relevant to understanding the potential effects of climate change on the terrestrial biosphere and the feedbacks between the biosphere and climate. BIOME4 simulates the distribution of 15 high-latitude biomes, including five tundra vegetation types, for the present day using observed climate, and the LGM, mid-Holocene, and a “greenhouse” scenario for 2100 using the output of GCMs. In the LGM simulations, the high-latitudes show a marked increase in the area of graminoid and forb tundra, which is also the predominant feature in the paleodata. This vegetation has no widespread modern analog; it was favored by the cold, dry climate, and supported large mammoth populations. Mid-Holocene simulations indicate a modest, asymmetrical northward advance of the Arctic treeline compared to present, with greatest extension in central Siberia (up to 300 km), and little to no change in the Western Hemisphere. This result is in good agreement with pollen and megafossil data from the same period. Differential warming of the continents in response to increased high-latitude solar radiation is hypothesized to account for the asymmetry. Vegetation changes in the 2100 projection, which assumes a continued exponential increase in atmospheric GHG concentrations, are more radical than those simulated for the mid-Holocene. The year-round forcing due to GHGs increases both summertime and annual temperatures in the high latitudes by up to double the mid-Holocene anomaly. However the potential treeline advances and biome shifts in our simulation are unlikely to be realized within 100 years, because of the time required for migration and establishment of new vegetation types. Potential natural wetland area for the present day was simulated by BIOME4 as 11.0x106km2. This value is higher than other estimates but includes small (<50km2) and seasonal wetlands which have not been included in previous surveys. The wetland CH4 source was simulated as 140Tgyr-1. At the LGM, simulated wetland area was increased by 15% but CH4 emissions were 24% less than the present-day. The simulated reduction in the CH4 source is due to substrate limitation induced by low atmospheric CO2 concentrations at the LGM. The 100% increase in atmospheric CH4 concentrations measured in ice cores between the LGM and the preindustrial Holocene may not be due to changes in CH4 source strength alone, as other trace gases influence the atmospheric CH4 sink. The stable carbon isotope composition of the terrestrial biosphere was simulated by BIOME4 and compared to measurements at the leaf, ecosystem and troposphere scales. Model simulations are correlated within one standard deviation to measured means at the PFT and biome scales, and at six Northern Hemisphere CO2 monitoring stations. Global carbon isotope discrimination in the terrestrial biosphere averaged 18.6‰ for potential natural vegetation and 18.1‰ when an agricultural landuse mask was applied. These simulated values are slightly higher than previous estimates, but consistent with measurements. This information is important for the interpretation of contemporary atmospheric observations in terms of carbon sources and sinks on land and in the ocean.