|
Titel |
Increased reliability of mean travel time predictions of river-groundwater exchange fluxes using optimal design techniques |
VerfasserIn |
Thomas Wöhling, Moritz Gosses, Karsten Osenbrück |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250089555
|
Publikation (Nr.) |
EGU/EGU2014-3760.pdf |
|
|
|
Zusammenfassung |
In this study, we follow up on previous work at the Steinlach test site (Osenbrück et al, 2013)
near Tübingen, Germany, to investigate hyporheic exchange fluxes in a shallow riparian
aquifer. A steady-state MODFLOW model has been developed for the site and calibrated
using an existing network of 14 observation wells. Due to a relatively steep hydraulic gradient
(0.012 m/m) between the upstream and downstream flow stages of the river bend, water
infiltrates from the river into the shallow aquifer along the upstream section of the river and is
forced to re-enter the river at the downstream end. The passage through the aquifer
potentially allows for mitigation and transformation of river water-bound pollutants. One
important factor to estimate attenuation potentials are travel (and exposure) times through
(parts of) the aquifer.
In our approach we used forward particle tracking (MODPATH) and a flux-weighting
scheme to estimate travel time distributions for the river-groundwater exchange fluxes in the
study domain. Travel times vary significantly within the domain, however, estimates of mean
travel times derived from deconvolution of EC and δ18O–H2O data at selected wells
exhibit a consistent pattern with modelled travel times. The flux-weighted mean
travel time of all river water that passed the riparian aquifer was calculated to 26.1
days.
The uncertainty of the flux-weighted mean travel time was calculated using the prediction
error variance approach by Moore and Doherty (2005) which resulted in a post-calibration
uncertainty of ±93.5 d (1Ïă), i.e. about 350% of the actual prediction. We further analysed the
worth of potential new observations to reduce the large uncertainty of this model
prediction. In our optimization framework, we extend the method by Moore and Doherty
(2005) to simultaneously optimize multiple observations using a modified Genetic
Algorithm (GA) that can also sample from past states for higher efficiency. The
observations considered are hydraulic head, hydraulic conductivity, and river bed
conductance.
Our results show that hydraulic head observations have the largest utility to reduce
predictive uncertainty for up to two new observations while hydraulic conductivity and bed
conductance have the largest utility for monitoring designs with more than two new
observations. The optimized design with new 10 observations significantly reduced
predictive uncertainty to a value ±3.6 d (1Ïă), which is a meaningful value for further
analysis.
References
Moore, C., and Doherty, D. (2005). Role of the calibration process in reducing model
predictive error. Water Resources Research 41(5), W05050.
Osenbrück, K.; Wöhling, Th.; Lemke, D.; Rohrbach, N.; Schwientek, M.; Leven, C.;
Castillo Alvarez, C.; Taubald, H. & Cirpka, O. (2013). Assessing hyporheic exchange and
associated travel times by hydraulic, chemical, and isotopic monitoring at the Steinlach Test
Site, Germany. Env. Earth Sci. , 69, 359-37. |
|
|
|
|
|