Keywords: calibration, inverse modeling, stochastic modeling, nonlinear biodegradation, stochastic-convective, advective-dispersive, travel time, network design, non-Gaussian distribution, multimodal distribution, representers
This thesis offers three new approaches that contribute to the ability of groundwater modelers to better account for heterogeneity in physically-based, fully distributed groundwater models. In both forward and inverse settings, this thesis tackles major issues with respect to handling heterogeneity and uncertainty in various situations, and thus extends the ability to correctly and/or effectively deal with heterogeneity to these particular situations.
The first method presented in the thesis uses the recently developed advective-dispersive streamtube approach in combination with a one-dimensional traveling wave solution for nonlinear bioreactive transport, to study the interplay between physical heterogeneity, local-scale dispersion and nonlinear biodegradation and gain insight in the long-term asymptotic behavior of solute fronts, in order to deduce (the validity of) upscaling equations. Using the method in synthetic small-scale numerical experiments, it is shown that asymptotic front shapes are neither Fickian nor constant, which raises questions about the current practice of upscaling bioreactive transport.
The second method presented in the thesis enhances the management of heterogeneity by extending inverse theory (specifically, the representer-based inverse method) to determinations of groundwater age/travel time. A first-order methodology is proposed to include groundwater age or tracer arrival time determinations in measurement network design. Using the method, it is shown that, in the applied synthetic numerical example, an age estimation network outperforms equally sized head measurement networks and conductivity measurement networks, even if the age estimations are highly uncertain. The study thus suggests a high potential of travel time/groundwater age data to constrain groundwater models.
Finally, the thesis extends the applicability of inverse methods to multimodal parameter distributions. Multimodal distributions arise when multiple statistical populations exist within one parameter field, each having different means and/or variances of the parameter of concern. No inverse methods exist that can calibrate multimodal parameter distributions while preserving the geostatistical properties of the various statistical populations. The thesis proposes a method that resolves the difficulties existing inverse methods have with the multimodal distribution. The method is successfully applied to both synthetic and real-world cases.
|Qualification||Doctor of Philosophy|
|Award date||19 Nov 2008|
|Place of Publication||[s.l.]|
|Publication status||Published - 2008|
- groundwater flow
- biochemical transport
- stochastic models