Regional analysis of greenhouse gas emissions is becoming increasingly important in answering questions related to environmental change, and typically employs a Geographic Information System (GIS) linked with a process-based simulation model. For the Northern Atlantic Zone (NAZ) in Costa Rica (281 649 ha), a regional analysis of soil-atmosphere nitrous oxide fluxes from the dominant land-use types forest, cattle pastures, and banana plantations was performed with both deterministic and stochastic variable representations. The stochastic representation accounted for soil and land management variability across nongeoreferenced fields within 1572 georeferenced land units in 13 relevant classes. Per class, frequency distributions of field-scale fluxes were simulated with a process-based model and Monte Carlo methods. Stochastic incorporation of both soil and land use variability resulted in areal (i.e. land unit-scale) fluxes that were 14-22␕ower than estimates based on averaged inputs. Soil heterogeneity was dominant. In addition, spatial flux patterns for current (1992) land use and two alternative land-use scenarios were evaluated using stochastic inputs. With current management, the regional nitrous oxide-N flux (standard deviation in parentheses) from agricultural land was 0.43 (0.13) Gg y1. Replacing natural grasses with mixtures of grasses and N-fixing species on relevant soil types and introducing different forms of banana plantation management (alternative I) increased the regional flux by 51␝o 0.65 (0.22) Gg y1. When all natural grasses were replaced by fertilized improved species and allowing different forms of banana plantation management (alternative II), the regional flux increased by 126␝o 0.97 (0.68) Gg y1. Using the revised IPCC methodology, the 1992 nitrous oxide emission from agriculture in the NAZ was estimated to be 0.32 Gg y1. Due to formidable data requirements, regional analysis may not easily be used to produce country-level estimates. However, regional analysis does provide a valuable benchmark against which the more straightforward IPCC methodology can be evaluated.