The Art of Tomographic Tracer Tests

Abstract:

In tracer tomography several tracer tests are performed within an aquifer and breakthrough curves are observed at multiple observation points. In the analysis, hydraulic conductivity is estimated as spatially variable, 3-D field subject to some smoothness constraint. Coupled flow and transport models using this conductivity fields are requested to meet observed tracer data. The approach can be combined with hydraulic tomography.
We have performed hydraulic-tomography and tracer-tomography tests using heat and fluorescein as tracers at a field site close to Tübingen, Germany. The aquifer consists of 8-9m alluvials sands and gravels overlain by 1-2m alluvial fines. The hydraulic setup consists of a forced flow field between an injection/extraction well couple, embedded in the forced flow field of another well couple. By turning injection to extraction wells, and vice versa, two different flow fields were considered. Injection wells were separated into several sections by packers, and water was injected into each section proportional to its transmissivity. The water injected into one of the sections contained the tracer. Multi-level observation wells were equiped with thermometers (for heat-tracer tests), on-line fluoremeters (for teh dye tracers), and pressure transducers. Processing of the breakthrough curves included data cleaning, non-parametric deconvolution, and calculation of temperal moments of the estimated transfer functions.
The joint inversion of hydraulic-head measurements and temporal moments of heat-tracer transfer functions was done by the quasi-linear geostatistical approach on a computing cluster. As alternative, we directly invert the time series (without temporal moments) by Ensemble-Kalman filtering.
The high diffusion coefficient of temperature diminishes the penetration of the heat-tracer into the aquifer, which can partially be compensated by reverting the flow field and repeating the tracer tests. In tests with fluorscent tracers the signal disappeared considerably slower, implying that a longer time is needed between individual tracer tests to make sure that the entire tracer has been removed before the start of the next experiment.