Canada's terrestrial ecostratification framework provides nested spatial units for organizing national data related to soils, landforms and land use. In the agricultural domain, the lack of national, uniform crop yield data on the ecostratification framework severely hinders our ability to evaluate the biophysical data with respect to economic and climatic conditions. We developed a national crop yield database at the regional (ecodistrict) level by aggregating individual records of an existing but very broad-level sample-derived yield database according to the ecostratification hierarchy. Issues related to the different sampling frameworks and the need for confidentiality of individual records were resolved in order to generate an ecostratified crop yield dataset at a reasonably detailed spatial scale. Sixty crops were first statistically arranged into 37 agronomically similar crop groups in order to increase class size, and these crop groups were aggregated into increasingly large spatial units until confidentiality was assured. The methodology maintained data quality and confidentiality while producing crop yield estimates at the ecodistrict level. Comparison to independent crop insurance data confirmed that the resulting crop yield data are valid where estimates were derived from data released at the level of an ecodistrict or an ecoregion, but not at the ecoprovince level. Our crop yield estimates offer a reasonably high level of spatial precision while remaining within standard confidentiality constraints.
