Accounting for Extensive Secondary Information to Improve Watertable Mapping

Generally, multiple variables are sampled in addition to the one used to quantify the main phenomenon under study. In some situations, this secondary information is sampled exhaustively for the study area and can be incorporated for helping in estimating the variable of interest. There are several methodologies available for incorporating densely sampled secondary information. Collocated cokriging is one of these methodologies and has the advantage to other methodologies, of accounting for spatial correlation. However, one drawback in its application comes from the difficulty in modeling the coregionalization. Simplified models of coregionalization were developed specifically for the collocated cokriging case. This paper presents, through a case study, the incorporation of dense secondary information using collocated cokriging. The watertable above a coal underground mine is mapped using water-level readings at monitoring piezometers complemented by highly correlated topographic data. The incorporation of dense secondary information through the utilization of a linear model of coregionalization and Markov models is comparatively evaluated in various aspects such as mathematical quality, coherence with the natural phenomena and the ease of utilization and modeling. Simple kriging with local varying means also is evaluated as it accounts for nonstationarity and is easily implemented. All these methods that incorporate topography as secondary information for mapping the watertable provided maps more in accordance with the natural phenomena than ordinary kriging using the principal attribute solely. Among these methods, strictly collocated cokriging using a linear model of coregionalization showed overall superior performance.