DESIGN OF GROUNDWATER MONITORING NETWORK-A GEOSTATISTICAL APPROACH

Abstract

Geohydrologists rely heavily on site-specific data to describe the geometry and properties of formations of interest. Samples of measurement taken at different locations within the same aquifer display a wide range of variability. As a consequence, one cannot predict with certainty the value of head at a point, even if these quantities have been measured at other nearby wells. Geostatistics provide statistical tools for these problems. One of the commonly used geostatistical method is kriging. The standard kriging method cannot be directly applied to the piezometric head data, since they inevitably display a trend, i.e., non-stationarity. The universal kriging can be applied to a non-stationary data set but could lead to a trend inconsistent with the causative hydrogeological conditions. In the present work, an alternative approach for geostatistical analysis of non-stationary piezometric head data has been adopted. The approach comprises an external identification of the trend purely from the hydrogeological considerations. The residuals around this trend are assumed to be stationary and hence amenable to kriging. This technique has been applied to Yamuna-Hindon Doab for designing a groundwater monitoring network. Analysing the prevailing network of 57 piezometers, the standard error at many subareas was found to be greater than 6. As a consequence of adding 35 additional piezometers, the standard error was found to generally fall below 1 m at many areas.