ABSTRACT
This paper presents sensitivity and scaling analyses using soil database attributes in an ecosystem model of global primary production and soil microbial respiration. The CASA (Carnegie-Ames-Stanford Approach) Biosphere model uses satellite imagery (Advanced Very High Resolution Radiometer and solar radiation), along with climate history (monthly temperature and precipitation) and soil attributes (texture, carbon and nitrogen contents and inundation) from global data sets as CIS input variables. A framework is summarized for spatial modelling and evaluation of potential aggregation errors associated with global gridded data sets. Soil carbon transformations predicted by the model are influenced by moisture effects on microbial activity and soil texture effects on the efficiency of heterotrophic respiration. We tested the assumption that the quality of global soil databases is critical to prediction of ecosystem controls on carbon cycling at large spatial scales. Model sensitivity analysis suggests that predicted soil carbon storage is highly sensitive to texture. A spatially uniform, fine-texture setting resulted in the highest soil carbon pool size; this trend was consistent over the entire global gradient of climate conditions. Using the FAO Soil Map of the World, the model estimates that on a world-wide basis more than 40 per cent of the surface SLOW carbon pool is stored in tropical forest and savanna biomes. Addition of soil inundation effects on microbial activity resulted in a 2 per cent increase in global soil carbon storage, with the most important changes in the needleleaf evergreen forest biome. Improvement of process understanding of organic soil at depth and texture characterization in tropical ecosystems are identified as priorities for global change research.
