 |
| Titel |
An evaluation of the predictive capabilities of CTRW and MRMT |
| VerfasserIn |
Aldo Fiori, Antonio Zarlenga, Hrvoje Gotovac, Igor Jankovic, Vladimir Cvetkovic, Gedeon Dagan |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250124656
|
| Publikation (Nr.) |
 EGU/EGU2016-4126.pdf |
|
|
|
|
|
| Zusammenfassung |
| The prediction capability of two approximate models of non-Fickian transport in highly
heterogeneous aquifers is checked by comparison with accurate numerical simulations, for
mean uniform flow of velocity U. The two models considered are the MRMT (Multi
Rate Mass Transfer) and CTRW (Continuous Time Random Walk) models. Both
circumvent the need to solve the flow and transport equations by using proxy models,
which provide the BTC μ(x,t) depending on a vector a of unknown 5 parameters.
Although underlain by different conceptualisations, the two models have a similar
mathematical structure. The proponents of the models suggest using field transport
experiments at a small scale to calibrate a, toward predicting transport at larger scale. The
strategy was tested with the aid of accurate numerical simulations in two and three
dimensions from the literature. First, the 5 parameter values were calibrated by using the
simulated μ at a control plane close to the injection one and subsequently using
these same parameters for predicting μ at further 10 control planes. It is found
that the two methods perform equally well, though the parameters identification is
nonunique, with a large set of parameters providing similar fitting. Also, errors in the
determination of the mean eulerian velocity may lead to significant shifts of the predicted
BTC. It is found that the simulated BTCs satisfy Markovianity: they can be found
as n-fold convolutions of a "kernel", in line with the models’ main assumption. |
|
|
|
|
|
|