World patterns and amounts of detrital soil carbon

Soil carbon data were collected from published sources for 50 measurement sites spanning the globe's major climate and vegetation types. For each site, climate, vegetation, and land‐use variables were determined and entered into a multiple curvilinear regression program to predict soil carbon. The best model incorporates an estimate of site disturbance, annual actual evapotranspiration, and annual soil moisture deficit, and yields an R = 0‐803. The curvilinear regression equation was coupled with a large climatic database and computer cartography programs to produce first‐generation maps of estimated soil carbon. These maps correctly portray soil carbon as high in boreal and cool temperate zones and low in deserts and tropical zones. Computer planimetry of maps of soil carbon for an ‘undisturbed’ world and for a ‘disturbed’ world resulted in estimates of 1457 × 10 9 mtC and 504 × 10 9 mtC respectively. These estimates compare favourably with recent estimates using other approaches. Clearly, the disturbance factor is critical to future refinements in estimates, suggesting the need for detailed studies of the relationship between land‐use history and the creation and destruction of this important carbon pool.