Implicit modelling is an approach to spatial modelling in which the distribution of a target variable is described by a unique mathematical function that is derived directly from the underlying data and high-level parametric controls specified by the user. This modelling approach may be applied to discrete variables such as lithology (after converting the discrete codes to numeric values) or to continuous variables such as geochemical grades. This paper discusses the estimation of continuous (grade) variables using implicit modelling.One of the underlying engines of implicit modelling for producing this mathematical function description is the radial basis function (RBF). In essence, the RBF is a weighted sum of functions positioned on each data point. A system of linear equations is solved to derive weights and the coefficients of any underlying drift model coefficients. Once derived, the RBF may be solved for any unsampled point or averaged over any volume to provide an estimate of grade. It is possible, for example, to query the RBF on a regular grid to derive an estimate of block grades. Given the ease of creation of an RBF, and its ability to predict grade, the question arises as to how the grades derived from the solution of an RBF compare with grade estimates derived from conventional geostatistical interpolation methods (eg ordinary kriging (OK))._x000D_
The purpose of this paper is to describe in lay terms: the basic structure of an RBF the role of parametric choice in the solution of RBFs and how this influences the character of the solution the fundamental similarities and differences between RBFs and conventional geostatistical estimators.Using a high-resolution conditional simulation, we show that in many situations RBF and OK estimates are very similar._x000D_
CITATION:Stewart, M, de Lacey, J, Hodkiewicz, P F and Lane, R, 2014._x000D_
Grade estimation from radial basis functions - how does it compare with conventional geostatistical estimation?, in Proceedings Ninth International Mining Geology Conference 2014 , pp 129-140 (The Australasian Institute of Mining and Metallurgy: Melbourne).