A risk-robust development of a heavy mineral resource requires an assessment of the geological uncertainty and spatial variability of the key factors impacting the mining and processing operation. Attributes of interest are the total heavy mineral grade, the slime content and the amount of oversized particles. The modelling of a mineral sand deposit presents two main challenges. Firstly, the exact location of the geological boundaries are obscure due to the complex interaction of the deposit formation processes. Secondly, the properties of interest exhibit a rather complex transitional behaviour across the so-called soft boundaries. Traditionally, the modelling exercise is restricted to a consecutive stochastic simulation of distinct geological facies and orebody properties. This results in different zones separated by hard boundaries, ignoring the typical transitional behaviour.This contribution presents an alternative approach which models the entire sand body as a single domain while accounting for the complex trends present. Error curves are used to infer covariance model parameters without the need of a prior removal of the trend. The principle is based on minimising the differences between empirical and theoretical errors resulting from two methods of prediction calculated for different rings of influence. The added value results from the unambiguous decomposition of the spatial variation into a deterministic trend and a stochastic residual. The trend model neither overfits nor underfits the data. The calculated residuals are thus unbiased and the uncertainty can be correctly assessed.The effort results in a characterisation of the resource uncertainty which provides a more realistic image of the spatial variability of the key properties in and across the geological domains of an existing heavy mineral sand deposit.CITATION:Wambeke, T and Benndorf, J, 2014. Optimal inference of modelling parameters to simulate complex trends across soft boundaries - a case study in heavy mineral sands, in Proceedings Orebody Modelling and Strategic Mine Planning Symposium 2014 , pp 73-84 (The Australasian Institute of Mining and Metallurgy: Melbourne).