A contaminant (or tracer) plume spreads by both molecular and mechanical processes. Variations in groundwater velocity (i.e. mechanical spreading) caused by aquifer heterogeneities can be represented with the parameter longitudinal dispersivity (αL) which is known to increase with distance. The rule of thumb that αL is approximately 10% of the length scale (L) is often used, seemingly based on the work by Gelhar et al. (1992). Zech et al. (2015) revisited the Gelhar et al. (1992) study, demonstrating that not only the spatial distance but also the degree of heterogeneity influences αL.
However, the issue remains that αL determined at one distance from a source will not always be a good predictor of transport behaviour at another. This talk explores the use of spatially variable αL in order to improve the predictive ability of tracer/contaminant behaviour.
Over 79,000 solute breakthrough curves were generated using various synthetic heterogeneous aquifers using the numerical simulator HydroGeoSphere. αL values were determined for breakthrough curves (at various distances from the tracer source) using two approaches: (1) fitting the Ogata and Banks (1961) solution, and (2) fitting a 1D numerical model where αL increases linearly with L. The fitted αL or αL-L relationships were then used to predict the shape of breakthrough curves at both upstream and downstream locations.
The αL-L approach in was generally better able to predict the shape of breakthrough curves at other distances compared to the use of a αLvalue fitted using the Ogata and Banks (1961) solution. However, it should be noted that αL does not always increase linearly, with αLreaching a maximum value after some distance. Nonetheless, the improved predictive ability from the use of spatially variable dispersivities is an approach that warrants further investigation.