The potential of temperature and dissolved gas as smart tracers for process-based heterogeneity characterization

Informative reference data for a realistic assessment of aquifer heterogeneity is a prerequisite for robust transport simulations. Structure-based imaging using salt or a dye as tracer with a known concentration and volume to observe transfer times, is a powerful hydrogeological tool in moderate heterogenous media. Solving then the advection-dispersion equation will explain most of the point to point transport behavior. But, once the aquifer heterogeneity is more complex, e.g. in a double porosity medium like chalk, matrix porosity linked to diffusion processes must be taken into consideration to avoid a biased interpretation of the tracer information. Thus, performing additional local process-based imaging using smart tracers as dissolved gas and hot or cold water, assists to explain the late-time tailing behaviors realistically. Smart tracers were injected in a sub-horizontal fracture connecting two adjacent wells to provide data about the complementary behaviors of each tracer and to focus on matrix diffusion processes. One reference data set is a 70 hours injection of hot water (∆T = + 40 °C) complemented by two 10 minutes uranine pulse injections within an inflatable double packer system isolating the sub-horizontal chalk fracture of interest. The temperature signal arrives at a 7.55 m distance with a delay of 12 hours compared to the first uranine injection and shows a rebound after the injection stopped. Useful reference data for further numerical modelling consists now in (a) local fracture geometry information deduced from interpretation by analytical solutions and, (b) matrix diffusion information. Numerical modelling of those smart tracer experiments may question deterministic models for predictions and motivates for data-driven prediction tools like Monte-Carlo simulation procedures within a direct predictive framework. Distance based global sensitivity analysis (e.g. simultaneous variation of multiple input variables like diffusion coefficient, aperture and matrix storage) will be considered accounting for temperature related changes of viscosity and density. Key information about the most influencing parameters are main model outcomes, as local process understanding is very useful for possible future upscaling in regional models made of structure-based imaging.