From observation to modelling: An experimental framework for testing the MIPs model on a headwater catchment in Luxembourg.

Abstract:

Background

The quantification of subsurface flow and transport processes at the hillslope scale remains a crucial challenge. The role of macropores, their connectivity, and the mixing processes within the subsurface remain poorly understood. Hydrological modelling has largely focused on input-output relationships, such as rainfall and discharge. The quantification of internal catchment processes has received less attention and is the focus of this work. For this purpose, we designed a series of experimental investigations to characterize the system properties and measure the dynamics of internal system variables. These will then be used to test and potentially improve a hydrological flow and transport model. Model development will be carried out within the Multiple Interacting Pathways model (MIPs), which uses random particle tracking to represent water moving within different pathways (Davies et al., 2011). Our hypothesis is that the fast response of the hydrograph is regulated from fast pathways in the fractured bedrock, which bypass the groundwater.

Objective

This research aims to generate insights into flowpaths and mixing processes through a combined experimental and modelling approach. We follow a two-fold strategy combining a hillslope-scale sprinkling experiment with a hillslope model as a learning tool. The experiment increases our insight in hillslope processes and generates necessary input parameters for the MIPs model that will further be used to evaluate the hillslope system. We will use this combined approach to:

-Quantify the relative importance of subsurface stormflow at the hillslope scale

-Estimate the residence times of water at the selected site

-Generate hypothesis on the behaviour of the system, to be tested with the MIPs model

The main questions we address are therefore:

-What flowpath does water take before reaching the trench face?

-What is the driving force that causes connectivity?

-Is there a clear threshold for this specific hillslope?

Methods

The research is carried out at a hillslope in the Weierbach catchment. The Weierbach is a forested catchment underlain by schist. The altitude ranges from 422 to 512 m a.s.l. Valley slopes are mainly covered with forest, whereas the plateau support cultivated land. The soil type is classified as periglacial deposit with an A horizon (0-5 cm depth). B horizon (5-40 cm depth), and C horizon (40-90 cm depth), which is mixture of weathered schist fragments and loamy deposits. Below 90 cm lies fractured schist bedrock (Juilleret et al., 2011).

The schist formation tends to be fractured towards the surface, forming a system of local reservoirs where water can be stored. Lateral flow can be interpreted as a movement of water across multiple temporarily connected reservoirs (Fenicia et al., 2013). The catchment hydrologic response is characterized by marked threshold behaviour.

In winter, the main discharge peak is reached after a variable lag time (from 20 hours to some days). Runoff is generated from the spilling of water from the reservoirs developed by the irregular bedrock topography, which, when disconnected, empty only due to evapotranspiration. This hypothesis is related to the concept of connectivity of flow pathways, which explains the state-dependent response of the catchment. Connectivity increases during wet conditions, resulting in a larger fraction of the catchment contributing as one or more connected units. In dry conditions, the decrease in connectivity reduces the transmission of flow. Because of the lack of surface runoff outside of the stream channel, most of the incident rainfall infiltrates.

The forested hillslope is located on the north facing slope (left bank) and has dimensions of approximately 6 x 8 m. A combination of electrical resistivity tomography (ERT), soil moisture, soil temperature and trench flow measurements, as well as soil water and trench flow sampling was used during the experiments. The ERT technique was applied to characterize the infiltration processes at the site throughout the sprinkling experiment. During the whole experiment, the hillslope was monitored by a continue ERT measurement. Experimental results were analysed and used to advance hypothesis on the behaviour of the system.

A 3.5 m long trench was installed at the lower part of the experimental site. Subsurface flow was collected at three different depths (25, 50, 130 cm) by 1.5 m long troughs. A gutter was inserted at each depth: one at the boundary between horizon A and B (25 cm depth), one at 50 cm depth, and the third in correspondence with the less fractured bedrock layer (approximately at 130 cm depth). Each gutter was equipped with a tipping bucket, recording at 15 minute intervals. Suction cups, temperature and soil moisture Time Domain Reflectometry (TDR, 30 cm long probes) sensors were placed at 10 and 80 cm depths at three locations. Suction cup lysimeters (PTFE-Quartz) were used to sample soil water during the experiments. Two recording wells (2 m deep) were installed at the base of the trench in order to quantify the losses caused by percolation to deeper layers.

At the stream, 20 m below the hillslope, discharge and conductivity were recorded continuously on 15 minute intervals. Moreover, conductivity and riparian groundwater level were also recorded.

The experiments were carried out between the 31st of March and the 10th of April 2014 during dry conditions (average soil moisture content was 0.21 cm³ cm^-3 at 10 cm depth and 0.35 cm³ cm^-3 at 80 cm depth). Two garden sprinklers Gardena Aquazoom^TM 250/2 were used to apply the water, covering an area of approximately 50 m². During the first week, only one sprinkler was used. Since no trench flow was measured, a second sprinkler was then installed. Intensities ranged from 8 to 20 mm h^-1. On the 2nd, 9th and 10th of April, tracers were added to the sprinkling water. On the 2^nd and the 9^thof April, a tracer solution of 2000 litres containing 10 kg of NaCl was used to sprinkle the area. On the 10th, a solution of 2000 l containing 5 kg KBr was applied. Trenchflow samples were taken at 10-15 min intervals during visible trench flow. All samples were analysed for conductivity, chloride and bromide (using the Ionic chromatography technique).

The MIPs model attempts to represent different possible flow pathways within the system using random particle tracking to represent water moving within different pathways. The movement of these particles is determined by system properties that are expressed in terms of probability distributions. In the MIPs framework, preferential flow pathways are simulated using velocity distributions. Velocity distributions are defined around a mean velocity that is determined by hydraulic conductivity, porosity and hydraulic gradient. The experimental results are the basis for the application of the MIPs model. In order to reproduce the movement of water through the soil, the 2D version of the model will be tested.

Preliminary results

The first analysis of the ERT inversion model showed fast vertical infiltration of irrigated water through the soil. Infiltrating water, likely following preferential flow pathways, quickly reached the fractured bedrock. Drainage efficiency was high; soil moisture returned to its initial state within hours after the end of the sprinkling. There was an absence of a proper wetting front moving along the shallow subsurface. All probes recorded fast peaks of soil moisture during sprinkling followed by a fast recession. For logistic reasons, a higher intensity set up did not allow to sprinkle for more than 3h continuously. The higher soil moisture values were recorded the first week of sprinkling. Unfortunately, with the available setting it was not possible to observe trench flow in concurrence with the highest values of soil moisture. Thus, soil moisture peaks are related to the duration of sprinkling more than with the volume sprinkled.

In the first days of the experiment, no trench flow was observed. The lateral connectivity, i.e. trench flow, was reached only when using two sprinklers, with intensities ranging from 11 to 20 mm h^-1 though the mid-hillslope area received peaks of 35 mm h^-1. Such a large precipitation rate was necessary to generate lateral connectivity and initiate trench flow. Thus, infiltration excess is needed in order to generate lateral connectivity and concurrently activate the trench.

These experiments reveal that lateral connectivity at the hillslope scale was insignificant relative to total rainfall. The total irrigation sprinkled over the experiment was equal to 348.0 mm (a total 68.000 litres). The total effective infiltration, given by the sum of the input that did not reach the trench, calculated only for days with trench flow (i.e. 3 days, the 8th, 9^th and 10^thof April), was 77.2 mm, equal to a 99% of total rainfall. The total amount of flow reaching the trench was only 23 litres. Moreover, trench flow at the 25 cm depth stopped a few minutes after the end of the irrigation. This means that vertical infiltration is the dominant flowpath at this hillslope.

The two wells at the bottom of the hillslope did not show the development of a water table during the experiments. In turn, a response was recorded at the stream. The stream, just 20 m below the hillslope, showed no change in conductivity during the first experiment (02/04). This might be due to the initial storage condition, which needed some days of sprinkling in order to be activated. During the experiments of the 09/04 and 10/04, stream conductivity rapidly increased. On both days, the response was recorded few hours after the onset of irrigation (Table 1).

Table 1. Measured lag times (in minutes) of the change in conductivity for the experiments of the 09/04 and 10/04, respectively with NaCl and KBr mixed input. Start of conductivity raise values is shown.

	09/04/2014	10/04/2014
Start irrigation	0	0
Duration of irrigation	85 min	110 min
Average rate of input	15.8 mm h-1	13.6 mm h-1
Initial response at trench: 25 cm 50 cm 130 cm	20 min 20 min 50 min	35 min 35 min 70 min
Response at stream	135 min	175 min

The relatively fast response of the stream may be related to sprinkling intensity. From day 5 (7th of April) the second sprinkler was installed and consequently the intensities almost doubled. This could be the driving force for the generation of a fast subsurface movement that was measured in a double peak conductivity rise at the stream.

We argue that the fast response is regulated from fast pathways in the fractured bedrock bypassing the groundwater. Groundwater conductivity in the riparian zone started rising on the 17/04, a week after the arrival of the conductive water to the stream, showing that the water reaching the riparian zone is slowly released from the bedrock, whilst a first rapid increase in the stream conductivity is due to preferential pathways.

The fast response in stream conductivity seems to be consistent with the double peak observed in streamflow discharge. The second, larger peak likely reflects slower movement through the subsurface.

This information reveals the need to consider the bedrock system carefully. Fast-responding flowpaths within the bedrock discontinuities likely explain these rapid changes in stream conductivity, whilst slower flowpaths within the fractured bedrock form a complex reservoir that comprise the slow, second peak. This work also demonstrates that it is crucial to consider the whole hillslope-riparian-stream system to determine connectivity across different scales.

References

Davies, J., Beven, K.J., Nyberg, L., & Rodhe, A. (2011). A discrete particle representation of hillslope hydrology: hypothesis testing in reproducing a tracer experiment at Gardsjon, Sweden. Hydrological Processes, 25, 3602-3612.

Fenicia, F., Kavetski, D., Savenije, H.H.G., Clark, M.P., Schoups, G., Pfister, L., & Freer, J. (2013). Catchment properties, function, and conceptual model representation: is there a correspondence? Hydrological Processes

J. Juilleret, J.F. Iffly, L. Pfister, C. Hissler (2011) Remarkable Pleistocene periglacial slope deposits in Luxembourg (Oesling): pedological implication and geosite potential. Bulletin de la Societe des naturalistes luxembourgeois 112: 125-130.