QUANTIFYING THE PLACE OF KARST AQUIFERS IN THE GROUNDWATER TO SURFACE WATER CONTINUUM: A TIME SERIES ANALYSIS STUDY OF STORM RESPONSE IN PENNSYLVANIA WATER RESOURCES

Though karst aquifers have commonly been recognized as intermediate between ground and surface water, their location between these end members is generally descriptive rather than quantitative. A more precise measure of a specific aquifer's behavior is revealed by auto-correlation and partial auto-correlation of transient variables in the system. Auto-correlation quantifies how long storm impulses persist in a system, and a system's inertia can identify its position along the continuum.

Auto-correlations of stage data from four springs, water level data from three drought monitoring wells, and discharge data from eight stream gauges, distributed in Blair, Huntingdon, and Lancaster counties in Pennsylvania, illustrate the local range of response to system perturbations. Data at the four springs were collected at 20-minute intervals for a minimum of one year; all well and stream flow data are from long-term hourly data sets collected by the USGS. The largest spring has a drainage basin area similar to the smallest gauged streams, and the wells are sited in areas unlikely to be affected by pumping wells.

The streams generally respond quickly to and recover swiftly from storm perturbations, with signals persisting from 1 to 10 days. Groundwater wells have greater inertia with system responses ranging from 10 to 56 days duration. The four springs as expected occur between ground and surface water response with duration ranging from 5 to 25 days. Surprisingly, the range of response for all three systems overlaps. There was no clear relationship between drainage basin size and inertia in either the springs or the streams, but, while the smallest systems did not necessarily show the least inertia, the largest spring and streams did show the greatest inertia within their type.

Auto-correlation blends different storm responses across a data set, and because differing aquifer conditions can lead to very different responses, the duration of the time series and the measurement frequency play important roles in the appearance of the auto-correlation function. For the purposes of direct comparison among water resources, series for auto-correlation analysis must either be of very long duration and high measurement frequency, or all analysis must be performed on data sets covering the same period with high measurement frequency.