An approach is developed to simultaneously model the geometry of multiple coal seams in 3-D without directly modeling surface elevations or isochore thicknesses. The approach truncates a locally refined 3-D model of estimated signed distance function values. The signed distance function measures the sign dependent distance to a coal/non-coal contact; negative in coal and positive outside. Global kriging is used to estimate the distance function at unsampled locations so that the resulting model has the correct anisotropy and is free of search artifacts. This approach is aimed at relatively large modeling volumes where the number of coal seams varies and the seams show complexities such as pinch outs and bifurcation. The technique aims to locate all of the coal boundaries with high precision through local grid refinement. The results can be used for resource/reserve evaluation. Crown Copyright (C) 2012 Published by Elsevier B.V. All rights reserved.