Rocky Mountain (56th Annual) and Cordilleran (100th Annual) Joint Meeting (May 3–5, 2004)

Paper No. 14
Presentation Time: 8:00 AM-5:00 PM

SPATIAL-TEMPORAL ANALYSIS OF GROUND-WATER QUALITY DATA: THE USE OF MULTI-DIMENSIONAL KRIGING


WELHAN, John A.1, NEELY, Ken2 and HAGAN, Ed2, (1)Idaho Geological Survey, Dept. of Geosciences, Idaho State University, Campus box 8072, Pocatello, ID 83209, (2)Idaho Department of Water Rscs, 1301 North Orchard St, Boise, ID 83705, welhjohn@isu.edu

Water quality monitoring networks are designed to optimize the cost and information content of their coverages, often by compromising trend analysis capability. The detection of temporal trends can be particularly problematic in networks whose data are intentionally or by necessity collected at different locations in every sampling round. We used spatial-temporal kriging to analyze and synthesize data in a temporally-monitored network to delineate areas of statistically-significant change and to provide a framework within which future network sampling design can be optimized.

Kriging is an optimal spatial interpolator that honors all measurements, accounts for redundancy of clustered data, and minimizes the kriging estimation variance, a measure of the uncertainty associated with a kriging estimate. ST kriging is a form of 3-dimensional kriging that involves two spatial coordinates (usually x, y) and the temporal coordinate. In effect, data collected from multiple sampling periods are used to create a 3D space-time "map" of how an analyte varies geographically over the life of the sampling network. ST kriging maximizes information return on monitoring cost by exploiting the temporal persistence of water quality (e.g., nitrate concentration) so as to reduce kriging estimation uncertainty in conventional map representations.

Eleven years of ground water nitrate measurements from the Treasure Valley aquifer in Idaho's Statewide Ground Water Quality Monitoring Network are used to demonstrate the advantages of ST kriging: improved mapping of water quality and regulatory exceedance probability, increased statistical confidence in identifying and mapping temporal trends, and the ability to minimize cost and maximize monitoring efficiency via combined spatial-temporal optimization of the network's sampling design.