Spatial Statistics 6 (2013) 57–77 Contents lists available at ScienceDirect Spatial Statistics journal homepage: www.elsevier.com/locate/spasta Continental-scale kriging of gold-bearing commodities Christien Thiart a,∗ , Alfred Stein b a Department of Statistical Sciences, University of Cape Town, Cape Town, South Africa b Department of Earth Observation Science, Faculty of Geo-Information Science and Earth Observation (ITC), Twente University, Enschede, The Netherlands article info Article history: Received 11 December 2012 Accepted 19 July 2013 Available online 26 July 2013 Keywords: Continental kriging Weighted variogram Stratification Gondwana Gold potential abstract This paper focuses on continental-scale kriging on the African con- tinent using the gold-bearing commodities of the Gondwana Geo- science Indexing Database. The mineral layer contains over 20 000 commodities, each containing information on its ordinal interval size value. Boundaries between class intervals across the database are, however, not uniform. We perform spatial interpolation on a continental scale using the commodity gold as the binary variable. First, we select an appropriate distance metric in order to krige on an essentially spherical surface. We use this metric to implement a valid covariance function. Second, the ordinal size classes of the commodities are combined into a unique size classification. In ad- dition, the commodity size classification is used as a proxy for data reliability and is incorporated by using a weighted variogram. The geology is used to stratify Africa into geologically homogeneous strata, leading to stratified kriging. The best model in each stratum is used to produce a map of gold commodities of Africa including the spatial uncertainties. By integrating advanced techniques with high-quality data, a state-of-the-art map of gold commodities was obtained for Africa, including the spatial uncertainties. © 2013 Elsevier B.V. All rights reserved. 1. Introduction A major aim of geostatistics is to predict (or map) values of an attribute of interest at unobserved locations using the observed data at known locations. The use of kriging for that purpose assumes ∗ Corresponding author. Tel.: +27 21 650 3219; fax: +27 216504773. E-mail addresses: christien.thiart@uct.ac.za (C. Thiart), a.stein@utwente.nl (A. Stein). 2211-6753/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.spasta.2013.07.004