Aims: Identification and characterisation of terroirs depends on knowledge of environmental parameters, functioning of the grapevine and characteristics of the final product. Field studies, resulting in point data, are necessary to investigate the functioning of the grapevine but in order for this information to be of use within zoning studies it must be placed in a spatial context.
Methods and results: A knowledge-driven model used the rules generated in regression tree analyses to directly classify natural terroir units with respect to expected response of Cabernet-Sauvignon and Sauvignon blanc in the Stellenbosch Wine of Origin District. The natural terroir units were then grouped into terroir units that were homogenous with respect to predicted response of selected viticultural and oenological variables for each studied cultivar.
Conclusions: The use of regression tree methodology (CART analyses) enabled the definition of decision trees for spatialisation of this data. Each natural terroir unit could be evaluated with respect to its potential viticultural and oenological response and thus grouped to identify terroir units.
Significance and impact of the study: The identified terroir units can only be considered preliminary but the methodology used has promising implications for different scales of study.
AttachmentsNo supporting information for this article