Investigation of the impact of sparse data on the use of geostatistical approaches

Abstract

Techniques to estimate semivariogram parameters are investigated with respect to assessing the effects of sparse data and detecting the presence of a trend. These techniques are applied to the determination of semivariogram parameters used in the estimation of hydraulic head values at node locations from measured head data. The presence of no-flow boundaries is included in the estimation of hydraulic heads at node values by applying constraints to the head gradient across the no-flow boundaries. The resulting hydraulic head estimates are used in an inverse groundwater flow model to assess the impact of the no-flow boundary constraints on transmissivities determined from the inverse model. It is found in a case study that when gradients in the prior head distribution do not match assumed no-flow boundaries, the inverse model can produce low transmissivity values along these no-flow boundaries. Prior heads estimated with constraints on the head gradient across no-flow boundaries did not produce the low transmissivity values along no-flow boundaries.