Application and evaluation of universal kriging for optimal contouring of groundwater levels

This paper deals with the application of universal kriging to interpolate water table elevations from their measurements at random locations. Geographic information system tools were used to generate the continuous surface of water table elevations for the Carlsbad area alluvial aquifer located to the southeast of New Mexico, USA. Water table elevations in the 38 monitoring wells that are common to 1996 and 2003 irrigation years follows normal distribution. A generalized MATLAB (R) code was developed to generate omni-directional and directional semi-variograms (at 22.5 degrees intervals). Low-order polynomials were used to model the trend as the water table profile exhibits a south-east gradient. Different theoretical semi-variogram models were tried to select the base semi-variogram for performing geostatistical interpolation. The contour maps of water table elevations exhibit significant decrease in the water table from 1996 to 2003. Statistical analysis performed on the estimated contours revealed that the decrease in water table is between 0.6 and 4.5 m at 90% confidence. The estimation variance contours show that the error in estimation was more than 8 m(2) in the west and south-west portions of the aquifer due to the absence of monitoring wells.