 |
| Titel |
What are the respective effects of the protocol, the model and the spatial variability on uncertainty in MSO experiments? |
| VerfasserIn |
Mélanie Weynants, Mathieu Javaux |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250042700
|
|
|
|
|
|
| Zusammenfassung |
| Multistep outflow (MSO) experiments have become a standard to characterize subsurface
hydraulic properties. The main advantages of this transient approach are that the
measurements are relatively quickly conducted, that both the retention and the hydraulic
conductivity curves can be obtained simultaneously, and that several inputs can be used in the
objective function for the inversion. However, the uncertainty on the saturated hydraulic
conductivity, the potential ill-posedness of the inverse problem and the underlying
hypothesis of homogeneity remain noteworthy drawbacks. Moreover users mention
unexpected high flow rates close to saturation leading them to omit the first pressures
step.
The objective of this study is to answer the following questions: (1) What is the most
appropriate model to describe the hydraulic properties from an MSO experiment and what
protocol is the most suitable to be as accurate as possible close to saturation without
observing measurement artefacts? (2) What causes the largest variation in the inverted
hydraulic properties: the spatial variability at the horizon scale; the choice of protocol or the
choice of model? (3) Is the MSO apt to characterize the hydraulic behaviour of structured
soils?
Three horizons close to the soil surface were sampled in the loamy region of Belgium under
different land uses (Crop, Grass, Forest) and MSO experiments were conducted on triplicate
undisturbed 1L soil cores. Fourteen suction and pressure steps were successively
applied on the initially saturated samples ranging from +10 to -960 cm of equivalent
pressure head at the lower boundary. Inversions were performed with AMALGAM-SO
with an objective function combining the cumulative outflow, the pressure head
halfway down the sample and points of the retention curve. The forward model,
Hydrus-1D, was used in combination with the classical Mualem-van Genuchten (MV), the
Durner (DR) and the Gerke (DUAL) models for the description of the hydraulic
properties.
The variation in the cumulative outflow measurements is the greatest in the Forest horizon,
followed by the Grass horizon and finally by the Crop horizon. The fitting performances of
the different models are generally good on the cumulative outflow and the suction data, but
less adequate on retention data. The results show that model DR using unsaturated data
performs the best. It means that it is better to let the sample equilibrate with a water table
levelled with the bottom of the sample before starting to simulate the water redistribution
experiment. Comparing the variation observed in the measurements and the simulations, it
appears that the variation in the Grass horizon is underestimated by the three models,
especially close to saturation. The optimised hydraulic conductivity curves do not
show the same variation at the horizon scale for the different models. The MV
curves tend to be closer to each other than the ones from the other models. The
ability of the MSO to provide relevant information on the hydraulic behaviour of
structured soils is supported by the good fitting performances of model DR. However,
the comparatively less good fits of DUAL tend to moderate this assertion. But the
chosen parameterisation of DUAL might have hindered the flexibility of the model. |
|
|
|
|
|
|