SIGNAL PROCESSING OF TIME-SERIES HYDROCHEMICAL DATA FROM EPIGENIC KARST IN THE CUMBERLAND ESCARPMENT OF KENTUCKY

Epigenic karst aquifers display a range of hydrochemical responses driven by climate patterns at several scales. Responses at springs range in character along multiple gradients including: conduit versus fracture dominated flow, discrete versus diffuse source of input, single versus multiple flow paths, and gravity versus pressure driven flow. Realistically, every spring has a unique fingerprint that is the result of the transformation of an input signal within the aquifer framework.

Signal processing methods including auto- and cross-correlations, fast-Fourier transforms, and wavelet analysis were used to de-convolve time series data from Grayson Gunnar Cave (GGC), an 11-km-long cave system embedded within a karst aquifer spanning an area of 5 km² along Cumberland Escarpment of south-central Kentucky. The data collected in 2015 at 10-minute-intervals using datasondes included discharge (Q – calibrated from a contemporaneous rating curve), temperature (T), specific conductance (SpC), pH, dissolved oxygen (DO), fluorescent dissolved organic matter (fDOM), total phycocyanin, and turbidity. Windows of investigation, partly driven by periods where instruments were offline, comprised the summer (6/1 – 8/1) and late fall (10/1 – 12/17). Similar data collected in 2010 from Sandy Springs in the nearby and larger Redmond Creek karst aquifer provide (RC) a basis of comparison between an open, branching conduit system comprising two basins (GGC) and an aquifer with significant groundwater flow through thick, valley-filling sediment (RC).

Both aquifers record synoptic-scale variations driven by storm events. GGC, with considerably lower system memory (aquifer inertia) than Redmond Creek, contains more internal event structures. For example, these structures include 40- to 80-minute increases in SpC from expulsed storage water at event onset; a 2- to 8-hr lag time between peak Q and T, and minimum SpC and pH; twin pulses of SpC separated by 2- to 8-hrs associated with the two sub-basins of the aquifer; and a 1- to 3-day change in T driven by storm-water migration through the epikarst. These structures are less evident during periods of higher base flow and cooler temperature.