| Geostatistical simulation makes strong assumptions of stationarity in the mean and the variance over the domain of interest. Unfortunately, geological nature usually does not reflect this assumption and we are forced to subdivide our model area into stationary regions that have some common geological controls and similar statistical properties. This dissertation addresses the significant complexity introduced by boundaries.; Boundaries are often soft, that is, samples near boundaries influence multiple rock types. We propose two different methodologies for grade estimation in the presence of soft boundaries. In the case of a global soft boundary, we propose to use a conventional linear model of coregionalization to simulate grades using data from adjacent domains. For local non-stationary soft boundaries we propose a non-stationary form of cokriging that accounts for stationary variables within domains and additional non-stationary components near boundaries.; The theoretical framework and notation are developed for both methodologies. Implementation details are discussed and resolved with a number of synthetic examples and a real case study. |
