Estimating soil carbon fluxes following land-cover change: A test of some critical assumptions for a region in Costa Rica

Changes in soil carbon storage that accompany land-cover change may have significant effects on the global carbon cycle. The objective of this work was to examine how assumptions about preconversion soil C storage and the effects of land-cover change influence estimates of regional soil C storage. We applied three models of land-cover change effects to two maps of preconversion soil C in a 140 000 ha area of northeastern Costa Rica. One preconversion soil C map was generated using values assigned to tropical wet forest from the literature, the second used values obtained from extensive field sampling. The first model of land-cover change effects used values that are typically applied in global assessments, the second and third models used field data but differed in how the data were aggregated (one was based on land-cover transitions and one was based on terrain attributes). Changes in regional soil C storage were estimated for each combination of model and preconversion soil C for three time periods defined by geo-referenced land-cover maps. The estimated regional soil C under forest vegetation (to 0.3 m) was higher in the map based on field data (10.03 Tg C) than in the map based on literature data (8.90 Tg C), although the range of values derived from propagating estimation errors was large (7.67-12.40 Tg C). Regional soil C storage declined through time due to forest clearing for pasture and crops. Estimated CO₂ fluxes depended more on the model of land-cover change effects than on preconversion soil C. Cumulative soil C losses (1950-1996) under the literature model of land-cover effects exceeded estimates based on field data by factors of 3.8-8.0. In order to better constrain regional and global-scale assessments of carbon fluxes from soils in the tropics, future research should focus on methods for extrapolating regional-scale constraints on soil C dynamics to larger spatial and temporal scales.