The study deals with the identification and characterization of
rapid subsurface flow structures through pedo- and geo-physical measurements
and irrigation experiments at the point, plot and hillslope scale. Our
investigation of flow-relevant structures and hydrological responses refers
to the general interplay of form and function, respectively. To obtain a
holistic picture of the subsurface, a large set of different laboratory,
exploratory and experimental methods was used at the different scales. For
exploration these methods included drilled soil core profiles, in situ
measurements of infiltration capacity and saturated hydraulic conductivity,
and laboratory analyses of soil water retention and saturated hydraulic
conductivity. The irrigation experiments at the plot scale were monitored
through a combination of dye tracer, salt tracer, soil moisture dynamics, and
3-D time-lapse ground penetrating radar (GPR) methods. At the hillslope scale
the subsurface was explored by a 3-D GPR survey. A natural storm event and an
irrigation experiment were monitored by a dense network of soil moisture
observations and a cascade of 2-D time-lapse GPR “trenches”. We show that
the shift between activated and non-activated state of the flow paths is
needed to distinguish structures from overall heterogeneity. Pedo-physical
analyses of point-scale samples are the basis for sub-scale structure
inference. At the plot and hillslope scale 3-D and 2-D time-lapse GPR
applications are successfully employed as non-invasive means to image
subsurface response patterns and to identify flow-relevant paths. Tracer
recovery and soil water responses from irrigation experiments deliver a
consistent estimate of response velocities. The combined observation of form
and function under active conditions provides the means to localize and
characterize the structures (this study) and the hydrological processes
From a general perspective the interplay of processes and spatial structures
It is a long-standing vision in eco-hydrology to observe and characterize the form and function of all possible different flow paths in the subsurface. However, this is hindered by a lack of observation techniques which are capable of measuring and visualizing flow paths across the relevant range of scales in a continuous manner. In this study, we address the challenge of in situ observation, identification and characterization of flow-relevant structures through a series of complementary methods at the point, plot and hillslope scale.
While heterogeneity is seen as a purely random variation of soil properties,
organized heterogeneity implies a spatial covariance of these properties and
connected flow paths. As such we define structure based on their functional
implication in line with
Despite observation of fast responses through such macroporous networks,
e.g., as tracer breakthroughs
Due to limited direct observability of subsurface flow, most evidence is
either inferred from integral responses or derived from model applications:
in the field, a large spectrum of methods is applied to investigate
subsurface connectivity
Furthermore, breakthrough curves of precipitation or irrigation events at
trenches or springs are commonly used
So far, relatively few studies managed to actually in situ image spatially
distributed subsurface flow paths at larger scales. On the one hand,
applicability is also often technically limited to very small scales:
Hydrological “standard approaches” attempt to explore parameters like soil
layer depth, porosity and hydraulic conductivity based on distributed
point-scale measurements. Also, state and flux monitoring most often consists
of a set of point observations, e.g., of hydro-meteorological conditions and
soil moisture. An appropriate sampling design is substantial for the
statistical inference
The rationale of this study is to analyze insights into flow-relevant subsurface structures based on qualitative and quantitative measurements at the point, plot and hillslope scale. Specifically, we hypothesize that a combination of quantitative field methods and in situ imaging of subsurface response patterns with dye staining and time-lapse GPR provides the missing link between form of the flow structures and how their interactions determine rapid subsurface flow and thus function.
We test this hypothesis by addressing three main research questions. What kind of information on sub-scale flow-relevant structures,
their characteristics and their distribution can be inferred from a large set
of direct point-scale measurements of soil hydraulic properties? How do salt tracer data, dye tracer stains, soil moisture response
patterns, and 3-D time-lapse GPR compare with respect to inference on
vertical flow channels and apparent flow velocities at the plot scale? How do methods and identified structures convey to the hillslope scale?
The study approaches the identification and characterization of flow-relevant
subsurface structures as the aspect of
The study at hand approached the topic on three complementary scales with a range of different methods: as a standard reference, results from auger exploration and in situ measurements of hydraulic conductivity and infiltration capacity were collected. They were extended with pedo-physical laboratory examination of 250 mL undisturbed ring samples for bulk density, porosity, texture, soil water retention characteristics, and saturated hydraulic conductivity. We then broadened the perspective to the plot scale with irrigation experiments accompanied by TDR (time domain reflectometry) measurements of soil moisture dynamics in a 1-D profile, 3-D time-lapse GPR imaging, and tracer recovery of dye, salt and stable isotopes. At the hillslope scale, 3-D GPR was used to identify flow-relevant structures in a static survey. For dynamic investigation, an irrigation experiment specifically designed to identify lateral flow structures was observed by a dense network of TDR soil moisture profiles and a series of trench-like 2-D time-lapse GPR transects.
The study is situated in the headwaters of the Colpach River, a tributary of
the Attert which has been investigated by several studies before
Map of the study sites in the upper Attert basin, Luxembourg.
The soil physical exploration addressed our research question Q1 using an
intentionally large set of hydrological and geophysical methods to survey the
subsurface. The sampling is guided by a network of hydro-meteorological
monitoring stations measuring all relevant fluxes and states in the
atmospheric boundary layer and the subsurface (research project “Catchments
As Organized Systems”,
Aligned with the monitoring stations, infiltration capacity and saturated
hydraulic conductivity were measured. In order to address plot-scale (few
meters) and hillslope-scale (a few hundred meters) heterogeneity, the design
consisted of clustered sets of point measurements along two catenas plus one
set at the site of the hillslope irrigation experiment presented in
Sect.
The distance between the clustered sets was 80–200 m. In each, three
nested sets with a lag distance of 10–20 m along and perpendicular to
the contour line were defined. In such a nested set at least one measurement
of infiltration capacity and two profiles (laterally spaced 1 m) of
saturated hydraulic conductivity in different depth levels were conducted. To
complete the scale triplet
In addition to the point measurements, a series of percussion drilled profiles (drill head diameter of 4 cm) as 1-D profiles were drawn and 250 mL ring samples were taken within the top 0.6 m for laboratory analyses.
Infiltration capacity was measured at 40 points with a Hood Tension-Infiltrometer (IL-2700, UGT GmbH). It employs a tension chamber (12.4 cm radius) as infiltration water supply. Inside the chamber, a defined low negative pressure head is established, which allows a precise measurement of infiltration capacity at different tensions. Three to five tension levels between the 0 and 5.5 cm water column were applied at each spot.
In addition to infiltration capacity at the surface, we used a Compact
Constant Head Permeameter (CHP, Ksat Inc.) for determination of saturated
hydraulic conductivity in 32 borehole profiles with 3–7 depth levels of
about 20 cm increments, with the lowest level at a depth where further
hand-drilling was inhibited by stones. The permeameter establishes a constant
water level (10.5 cm in our cases) above the bottom of a borehole (here
5 cm in diameter). The outflow is measured to calculate saturated
hydraulic conductivity
The 63 undisturbed soil ring samples were analyzed for bulk density, porosity (assumed to be equal to saturated soil water content), soil water retention properties (Hyprop, UMS GmbH and WP4C Decagon Devices Inc.), saturated hydraulic conductivity (Ksat, UMS GmbH), and soil texture (ISO 11277, wet sieving and pipette method sedimentation).
In order to explore the network of flow-relevant structures and patterns of
rapid subsurface flow, we conducted three plot-scale irrigation experiments.
This relates to our second research question Q2. The general setup is very
similar to the one described by
Three plots of 1 m
Plan view layout of the plot-scale irrigation experiments. Three
irrigation plots (1 m
The irrigation was accomplished by spray irrigation (full-cone nozzle
Spraying Systems Co.) using a wind-protection tent. Brilliant Blue dye tracer
(4 g L
In addition, temporal dynamics of soil moisture along a selected profile was
monitored throughout the experiments through continuous TDR measurements in
an access tube (Pico IPH, IMKO GmbH) down to 1.5 m depth and with a
diameter of 4.2 cm. This technique is chosen to minimize the impact of
sensor installation (percussion drilling and installation of the tubes from
the surface) and to avoid interference with the GPR (sensor probe was removed
during GPR measurements). The sensor measured an integral of about 1 L
(depth increment of 18 cm, mean signal penetration of 5.5 cm). It was
manually lowered in the tube to the respective depth for each reading. Each
measurement took about 10 s. Hence the whole procedure added up to
4–10 min per profile record. The procedure was continuously repeated until
1.5 h after irrigation onset in line with the findings of
Two hours after the end of each irrigation, a percussion drilled soil core was taken
(drill head diameter of 8 cm) and sampled in 5 cm depth increments down
to 1 m. The plot was excavated 24 h after irrigation for vertical and
horizontal recovery of Brilliant Blue stains. This was done by successive
digging of three vertical faces into the plot (aligned with the slope line,
0.1 m distance starting from the lateral edge) and five to seven
horizontal cuts in different depth levels down to the first deposit layer
(
All samples were analyzed for bromide (
A recovery coefficient (RC) is calculated as proportion of recovered mass of
Prior to the bromide analysis, the percussion drilled soil core samples were
also analyzed for their stable isotopic composition (
The quantitative measurements allow one to infer apparent vertical flow
velocity along the profiles. For bromide we employ a cumulative curve method
The individual TDR soil moisture measurements (
GPR is known as geophysical imaging technique with high spatial resolution
Using this setup, we acquired one 3-D GPR data cube before irrigation, one
directly after the end of irrigation, and a last one about 20 h after
irrigation for each plot. One survey took about 45 min.
In our study we calculate the structural similarity attribute
As an additional estimate of vertical response velocities, the same approach
as for the soil moisture responses (Sect.
In order to examine the characteristics of flow-relevant structures and the
periglacial deposit layers at the hillslope scale, we conducted an experiment
on 21 June 2013 at a close-by hillslope. The experiment was specifically
designed to explore the response in lateral preferential flow paths and to
replicate the plot-scale experiments without tracer application. The site had
to be chosen for facilitation reasons (permissions, accessibility,
collaboration within the CAOS research project). With reference to its
hydrological responses
As an additional reference to the soil core profiles, a 3-D GPR survey of the
hillslope was conducted prior to the natural event and the irrigation. The
GPR data processing relies on a standard processing scheme including bandpass
filtering, zero time correction, envelope-based automatic scaling, gridding
to a regular 0.03 m by 0.1 m grid, inline fk-filtering and a 3-D
topographic migration approach as presented by
For structural analysis, the processed data are imported into the OpenDtect
software (dGB Earth Sciences). Under heterogenous soil conditions the derived
data cube is dominated by complex reflection patterns which prohibit a
classical structural analysis based on picking reflectors
The experimental site is located at the lower part of a north facing
hillslope. Vegetation is dominated by mixed beech forest. However, the
experimental site is placed in an area with no major trees. Except for few
young trees at the downhill monitoring area, all shrubs were carefully
removed from the experimental site to accomplish GPR measurements and allow
for undisturbed and homogeneous irrigation. The topographic gradient is about
14
Layout of the hillslope-scale irrigation experiment as vertical
view
The experimental layout is given in Fig.
Moreover, a surface runoff collector was installed across 2 m of the lower boundary of the core area. It was built from a plastic sheet installed approximately 1 cm below the interface between litter layer and Ah horizon of the soil profile. At the downhill end of the sheet, the water was captured by a buried and covered gutter. An in-ground tube was attached to the deepest point of the gutter to conduct the water to a tipping bucket downhill of the investigated area. The tube had been filled with water prior to the experiment to ensure an immediate reaction to the occurrence of surface runoff.
We monitored soil moisture dynamics in a setup of 16 access tubes with 3 manual TDR probes like in the plot-scale experiments (Imko GmbH, two with 12 cm integration depth and one with 18 cm). Measurements required manual positioning of the sensor probes for each reading. We continuously recorded the states in all tubes in 10 cm depth increments, realizing revisiting intervals of 5–20 min. The tubes were installed to reach to a depth of about 1.7 m. The layout consisted of three diverging transects with four TDR profiles in the lower half of the core area, the highest density of profiles just downhill from the rain shield, and the furthest profile about 9 m downhill.
Four 2-D time-lapse GPR transects were treated as
In order to synchronize the almost 5000 individual TDR soil moisture records to a regular grid in time and depth, interpolation and resampling were required. To do so, we generated an intermediate grid of high data density onto which linearly interpolated versions of the time series of each profile were projected. We then resampled from this intermediate grid to derive a synchronized version of the records at 0.1 m depth and 15 min time increments. With this the spatial aggregation remains below the integration length of the TDR probes.
The temporal resampling and the therefore necessary linear interpolation is
close to the acquisition timing of one profile (4–10 min each). Since the
correlation length of distributed soil moisture observations is rather short
and because we explicitly aim to analyze the responses of preferential flow
structures, the issue of interpolation needs special attention and will be
discussed in Sect.
All soil moisture measurements are converted to changes in soil moisture referenced to the state previous to irrigation onset to identify activated flow paths. Lateral interpolation between different TDR profiles over distances of about 1 m and above is unfeasible. Soil moisture as extensive state variable is discontinuous at interfaces. The found subsurface setting does not exhibit any isotropic continuum required for such interpolations.
As in the plot irrigation experiments, vertical response velocities are
calculated for the TDR profiles in the core area. The calculation of lateral
response velocities is given in the companion study
The 2-D time-lapse GPR data are derived from nine repeated recordings along
the four vertical GPR transects. Each record is processed after a standard
processing scheme of bandpass filtering, zero time correction, exponential
amplitude preserving scaling, inline fk-filtering, topographic migration with
constant velocity (0.07 m ns
Most time-lapse GPR data analyses are based on calculating trace-to-trace
differences
Like for the plot-scale experiments, we use the time-lapse structural
similarity attribute
Due to the presence of remaining event water from the preceding storm event
The experiment was preceded by two strong storm events of 43 mm in total
on 20 June 2013. In reference to the gauge reaction the experiment was
conducted shortly before the second peak of the runoff reaction to the
preceding storm events
The in situ point measurements of infiltration capacity and saturated
hydraulic conductivity showed high variability without clear relationships
with simple morphological descriptors like depth, hillslope position or
topographic flow gradient (details given in Fig.
On average the area is dominated by silty soils
Results of laboratory analyses of 63 undisturbed 250 mL ring
samples.
The soil core profiles (Fig.
Soil core profiles from the upper Colpach River basin. See
Figs.
Based on these standard techniques the overall setting of a heterogeneous silty soil deviating from expected low hydraulic conductivity was revealed. So far gained insight is limited to the general existence of periglacial deposit layers (high gravel content in soil profiles), rapid flow paths (hydraulic conductivity several orders of magnitude above literature references), and some integral retention properties. However, details about its spatial organization and the detection of specific and potentially continuous structures remained obscured by high heterogeneity.
In the plot-scale tracer experiments the Brilliant Blue dye stains identified
patchy infiltration patterns partially bypassing large sections of the soil
without clear traces of the actual flow path (Fig.
The connectedness and large transport capacity of this network of
inter-aggregate pores is corroborated by the distributions of bromide tracer
recovery (Fig.
At plot XII we found a stronger interaction with the soil matrix, which led to more dye staining and a higher bromide recovery. Overall, tracer recovery was incomplete (0.45, 0.38, and 0.83 for plots X to XII, respectively) and even declined when including the core samples (0.24, 0.3, 0.63) once more, pointing to strongly irregular soil water redistribution.
Recovered dye patterns in plot irrigation experiments.
The observed soil moisture changes (Fig.
Results from plot-scale irrigation experiments with 50, 30, and
50 mm spray irrigation for 1 h.
The structural similarity attribute of the 3-D time-lapse GPR measurements
provided qualitative information of changes in soil moisture in a spatial
context. At all plots the response patterns of low structural similarity
pointed out quick vertical flow to a depth of 80 ns or about 1.4 m
within 1.5 h after irrigation start (Fig.
The contrasting attribute distributions over time and comparing plots X and
XII not only revealed diverse patterns. It also highlighted the qualitative
nature of the analytical method of the GPR data. Although visual
interpretation of the radargrams (top rows in Figs.
Time-lapse 3-D GPR of irrigation experiment at plot X. Center line radargrams at the marked transect (gray dashed line in lower panels) for the three acquisition times (before 0:00 h, directly after irrigation 1:00 h, 20:00 h after irrigation) are given in the top row. Two way travel time (TWT) is given as original depth reference. The structural similarity attribute of the 3-D data cube is given in three different depth layers (top 20–40 ns, mid 40–60 ns, low 60–80 ns) in the lower panels. The irrigation plot is marked by a black dashed box/line. Slope line distance is increasing downslope.
For the identification of structures, the results did not exhibit specific macropores like the dye stains, but areas of response to the irrigation. Nevertheless, the patchy characteristic of the found response patterns was very similar to that of Brilliant Blue.
Based on all applied techniques, hydraulic conductivity and apparent vertical
flow velocities were calculated (kernel density estimates plotted in
Fig.
All results ranged several orders of magnitude above the literature reference
of
Saturated hydraulic conductivity and apparent vertical flow velocity
kernel density estimates.
The 3-D GPR survey at the site of the hillslope experiment identified
fragmented structures at about 1.5 m depth (Fig.
Potential subsurface structures from 3-D GPR survey and setup of
hillslope experiment. Structure identification guided by the dip corrected
semblance attribute. Depth estimated based on mean measured effective radar
velocity in soil of 0.07 m ns
The results of the hillslope-scale irrigation experiment can be distinguished
into the core area observations with TDR profiles only and observations at
the downhill monitoring area, including TDR profiles as well as 2-D GPR
transects. The change in soil moisture at the core area (TDR 2 and 8 in
Fig.
Development of soil moisture in TDR profiles during and after the
hillslope irrigation experiment. Exemplary transect with changes referenced
to pre-irrigation conditions and attributed to irrigation water. Time is
given in hours after irrigation start. Individual measurements and probe
reference marked with triangles. More data and explanation in
The four successive GPR transects across the downhill monitoring area
provided spatially distributed images of hillslope-scale flow patterns and
boundary fluxes. The structural similarity attribute of storm event water
(green) and irrigation water (blue) revealed distinct, heterogeneously
distributed patterns
(Fig.
The patches which reacted to the storm event are mostly different ones than
the structures used to drain the irrigation water. Apparently the irrigation
experiment initiated flow in more shallow structures (compare transect 1
irrigation reaction with transect 3 storm water in
Fig.
Structural similarity attribute in time-lapse 2-D GPR transects.
Blue: irrigation event water; green: storm event water. Columns: time series
in one transect; rows: different transects at the same time.
The patchy structures at the transects highlighted the irregularly
distributed nature of lateral preferential flow paths which was similarly
observed in the plot experiments. Although some areas exert a higher density
of reacting flow paths than others, no continuous patterns could be specified
throughout the hillslope. We also saw a decay of the signal strength and
areal share with distance from the core area. As the patterns from transect 1
did not simply propagate further downslope, the flow paths must be tortuous
and leaky. Hence inferring the configuration of the connection between the
four transects in the downhill direction is not feasible. A comparison of the
suggested structures of the 3-D GPR survey to the overall response to
irrigation recorded at the GPR transects did not correlate well (compare
identified potential structures with the reaction summary at GPR transects in
Fig.
Our results have shown that the silty soils coincide with high porosities and high hydraulic conductivity at the point scale. Such a coincidence is not what is expected for cohesive, fine textured soils and can be explained by a setting of aggregated fine material in conjunction with a network of inter-aggregate pores. With respect to our research question Q1, the pedo-physical analyses initiated the recognition of these sub-scale structures. However, neither their position nor their general setup can be identified based on point observations because of its scale below the support of the measurements. Vice versa, methods at the next scale do not provide information about porosity and bulk density.
Irrigation experiments at the plot scale visualized that a network of these
inter-aggregate voids connects the surface to the periglacial deposit layer
and is responsible for highly diverse soil water redistribution. These
structures are different from what we usually expect (cracks, worm burrows,
roots channels) at this scale. This could be depicted from dye tracer stains
(Fig.
At the hillslope scale (Q3), applications of 3-D time-lapse GPR are technically impossible due to the long acquisition times. Consequently we altered the setup to four trench-like 2-D time-lapse GPR profiles to facilitate the required high temporal resolution. The responses suggest structures similar to but less diverse than the found inter-aggregate voids at the plot scale. They are spatially persistent and leaky and apparently feed from diverse sources. As such the irrigation experiment caused a similar response in different structures than the previous storm event. Moreover, the relatively high input rates have proven adequately chosen to identify lateral subsurface flow paths. At this scale the capability of point-based methods for structure identification is even more limited as the dense network of soil moisture profile observations did not allow the derivation of a conclusive picture.
Interestingly, static methods failed to unravel structures from overall
heterogeneity. This corroborates our idea that responses to an event are
required for the identification of flow-relevant structures. Furthermore, it
confirms that a combined assessment of form and function is needed to
mutually reduce ambiguity. This is also shown in the companion study with a
focus on function and processes at the hillslope scale
In our case TDR measurements through access tubes were employed as low-impact means to monitor soil water dynamics in order to detect areas of quick and strong response. Structures in general and the inter-aggregate voids in our case cover only a very small fraction of the measured volume. We may underestimate detected flow paths when they do not alter the total volumetric soil water content much (bypassing). This can explain the observed patterns of low response in the topsoil and changes in regions where the fast flow is decelerated at some kind of bottleneck. Referring to the theoretical integration volume of 1 L, it would require a macropore of about 1 cm diameter within the support of the sensor to be filled to just reach a threshold of 2 % vol. Adding this 20 mL of water diffusively would result in the same measurement. This shows that soil moisture measurements exhibit a conceptual bias towards the diffusive fraction of the soil water.
The quantification of advective water from the recorded changes in soil
moisture has been proven as not feasible. Given the insight of the discretely
structured flow domain and the high lateral response velocities identified in
the companion study
The potential horizons identified by the static 3-D GPR survey do not
coincide with the observed responses (Fig.
The comparison of radargrams in time needs further attention: In other
time-lapse GPR applications for soil water dynamics in structured domains
On the one hand, we minimized methodological problems concerning the noise
arising from the imperfect positioning of repeated GPR measurements by using
a measuring platform at the plot scale, transect guides at the hillslope
scale, and an automatic-tracking total station
Although the 3-D time-lapse attribute data of the plot irrigation experiments are of low spatial resolution (blur due to similarity attribute method and long duration of one acquisition) and limited temporal resolution (few acquisition times), they are suitable to identify regions of flow-relevant structures and their characteristics. In the multi-2-D transects resolution was enhanced (short duration of one acquisition and many repeated measurements) which depicted the structures much better. Hence, time-lapse GPR can especially be improved by enhancing the acquisition time and frequency.
The observation of changes during activation of flow-relevant structures generated the required contrast to overall heterogeneity. For large structures, this ledto precise identification and localization. Smaller flow paths cannot be fully resolved. Nevertheless, the continuous 2-D and 3-D images of the subsurface response patterns provide means to non-invasively study the form–function relationship in situ and to overcome some of the restrictions of retrospective and destructive tracer methods. However, quantitative interpretation of time-lapse GPR data remains challenging.
In contrary to our first expectation, the value of pedo-physical analyses of
soil core samples has been relatively high even for characteristics of flow
facilitated by the revealed paths at larger scales. Structure identification
is not only obscured in heterogeneity as one would expect, but properties
deviating from the standard situation (fine texture, low bulk density and
high porosity) gave rise to the identification of the inter-aggregate flow
paths. However, the spatial organization of structures below and above the
support of the samples cannot be revealed. This is also the reason for the
relatively low information which could be drawn from the in situ infiltration
measurements: The observed flow rates are largely affected by the capacity of
the connected flow paths draining the measurement point. This adds to the
critical assumption of homogeneity
Besides the high information gain through the state shift of flow-relevant
structures in irrigation experiments, the employed methods at the plot scale
have very specific advantages and disadvantages: Especially the laborious and
costly analysis of salt tracers and stable isotopes is contrasted by
relatively little additional information. Moreover, the lack of a temporal
information about when the solute or water molecule was retained in a certain
depth is seen problematic. Soil moisture profile dynamics and time-lapse GPR
do not suffer this drawback. Both can be employed with very low or even no
impact on the subsurface from the surface. While GPR requires to be operated
in higher temporal resolution (see Sect.
Under strongly structured conditions as at the hillslope under study, point
observations remain a needle in a haystack. Unlike for vertical structures at
the plot scale, the dense network of soil moisture profiles could not depict
the lateral flow paths well. Here, the
With regard to our a priori model application, the combination of vertical and lateral flow paths (identified in the irrigation experiments) with layers of low permeability just below the structures (observed in the soil core profiles) could refine the domain towards more lateral soil water transport. The mean retention properties (derived from pedo-physical analyses) are adequate. Hence, the combination of data from all scales can contribute to a refinement of the model.
Based on numerous point-scale measurements, the overall layering and mean
property of a heterogeneous soil with periglacial deposit layers were
described in Sect.
At the hillslope scale, the attribute supported picking of potential
structures in the 3-D GPR data cube also had high discrepancies from the
actual relevant structures (see the differences between potential subsurface
structures and recorded reaction in the TDR and GPR profiles in
Fig.
It has proven particularly difficult to distinguish heterogeneity and
structure. This has conceptual implications: as introduced, we regard
statistical heterogeneity as random small-scale changes in hydrological soil
properties
In more general terms, heterogeneity can be seen as deviation of the found reality from the concept of quasi-homogeneous elementary volumes. If this deviation concerns only the apparent parameters of the same physical process, more samples are adequate to determine their distribution. In cases (like here) where this deviation also means a shift in the physical processes, heterogeneity may introduce bias as it becomes a scale problem: any measurement will consist of an unknown subset of connected or non-connected flow paths. This makes it impossible to unravel the properties of the different flow domains without knowing the composition of the explored ensemble of each measurement. Hence the point-based techniques cannot determine the super-scale organization outside the support of the measurement. Without detection of organization and thus flow-relevant structures, they can only recover heterogeneity independently of the number of samples.
In the context of preferential flow studies in watersheds around the globe
and in many different models, our results open new ways to visualize
subsurface flow and to facilitate more field studies to understand stormflow
generation
In the form and function framework one implication of the study is that a
disjunct analysis of the two is a source of unnecessary ambiguity and
susceptibility to bias. Although the conjugated nature of form and function
is very much in line with the general findings and perception of early
studies
While models require specific parameters about the site under study which are
coherent with their conceptual assumptions or modeler's perception
In the hillslopes under study, silty, cohesive soils coincide with high porosity and high flow velocities at the Darcy scale. This motivated in depth investigation of flow-relevant structures explaining this. We have shown that subsurface heterogeneity and the mismatch of observation and process scales obscured the identification of flow-relevant structures under static conditions without a shift between active and non-active states. The pedo-physical analyses initiated the recognition of these sub-scale inter-aggregate structures. The point-scale exploratory methods could quantify the general characteristics of the subsurface only within a wide spectrum of the respective target properties. However, they failed to identify flow-relevant structures in terms of position, distribution and capacity at larger scales. Measurements of infiltration capacity and hydraulic conductivity require special attention, because they integrate over an unknown set of advective and diffusive flow paths. The discrepancy between results from the soil core profiles and a 3-D GPR survey on the one hand and the time-lapse approaches on the other hand indicates that structures identified from inhomogeneities are not necessarily flow-relevant pathways.
Joint application of tracers and time-lapse GPR during irrigation experiments
revealed details about the structures and their activation by flow. At the
plot scale a network of inter-aggregate pores enables fast soil water
redistribution in a less directed manner and at much finer scales than
usually expected in macropores like cracks, worm burrows or root channels.
This facilitates high apparent vertical flow velocities ranging around
10
Our findings show that form and function in hydrological systems operate in
conjugated pairs. This implies that it is very difficult to observe them
separately and that their projections are inherently non-unique and
scale-dependent. Besides the fine scale of the inter-aggregate voids, form
needs to be addressed in its context to reveal information about its
structure and characteristics, but addressing function also needs details
about the spatial circumstances to be conclusive. Overly strong assumptions
about structures or processes can be avoided by the presented non-invasive
time-lapse GPR method, which can visualize and localize response patterns at
the plot and hillslope scale. They compare well with soil moisture dynamics
and tracer recovery. As such the localization of responses provides the
missing link to relate form to function
All data used in this study is foreseen to be openly published in Earth System Science Data (ESSD) as concise outcome from the research project. Until then they are available from the authors on request.
In addition to the dye tracer stain records, quantitative analysis of salt tracer recovery distribution in the excavated profiles underneath the irrigation plots was done. One challenge to address was the required time to collect an adequate array of such soil samples with known volumetric reference. We developed a re-loadable core sampler with a calibrated sample volume of 66 mL.
The sampler is applied like a ring sample with an attached hammering adaptor. In order to minimize time and impact on the profile we enabled a pull-withdrawal of the sample. For this, the sampler is about 15 mm longer than the desired sample. The irregular open edge is scraped off by a calibration twist drill. The prepared and accurate to volume sample is finally pushed out by a piston from the sampler into a sealable brown glass bottle for further treatment in the laboratory.
In situ measurements of infiltration capacity and saturated hydraulic
conductivity had a highly heterogeneous distribution. To detail the
respective records and found profiles, Fig.
Hydrological exploration results.
Based on the findings of the pedo-physical exploration, we setup the 2-D
process model CATFLOW
CATFLOW model reference of the hillslope experiment.
Comparing the results from the model with the experiment shows strong deviation in terms of the activation of a conductive layer. However, this could be improved by adding a layer of low permeability below, since the modeled reaction on the bedrock interface is quite similar but slower than the observed dynamics.
In addition to bromide as a conservative salt tracer, the same percussion
drilled core samples were analyzed for their stable isotopic composition
(
Plot-scale irrigation experiments. Proportion of event water derived
from deviations in concentrations of deuterium and bromide in the soil water
of sampled cores to the reference
We calculate the volumetric event water portion (–) in the soil water as
Figure
Time-lapse 3-D GPR of the irrigation experiment at plot XI. Center line radargrams at the marked transect (gray dashed line in the lower panels) for the three acquisition times (before 0:00 h, directly after irrigation 1:00 h, 20:00 h after irrigation) are given in the top row. The structural similarity attribute of the 3-D data cube is given in three different depth layers (top 20–40 ns, mid 40–60 ns, low 60–80 ns) in the lower panels. The irrigation plot is marked by a black dashed box/line. Slope line distance is increasing downslope.
Time-lapse 3-D GPR of the irrigation experiment at plot XII. Center line radargrams at the marked transect (gray dashed line in the lower panels) for the three acquisition times (before 00:00 h, directly after irrigation 01:30 h, 20:00 h after irrigation) are given in the top row. The structural similarity attribute of the 3-D data cube is given in three different depth layers (top 20–40 ns, mid 40–60 ns, low 60–80 ns) in the lower panels. The irrigation plot is marked by a black dashed box/line. Slope line distance is increasing downslope.
In line with the findings of
In addition to the results in Sect.
Standard deviation of the structural similarity attribute at the different GPR transects in the hillslope experiment over time (solid lines) and standard deviation of the differences of two successive attribute distributions (dotted lines). The used threshold for the detection of flow-relevant structures is marked as the dashed purple line.
The demands on the precision of the repeated acquisition with spatial
determination and antenna contact to the ground are very high and are assumed
to be nearly perfect within our experiments. Under field conditions precision
is limited due to numerous effects like micro-topography, topsoil conditions,
signal attenuation and even weather. The lack of distinguished reflectors
also inhibited any estimation of quantitative values. Further, the referenced
depths in Fig.
The highlighted assumptions clearly frame the limits of the technique. The
overall sensitivity of the approach can be judged from the structural
similarity attribute of the last pairs of records in the hillslope experiment
when we assume the soil water to be in equilibrium again.
Figure
Another limit is the interpretability of changes in the radargrams, as water can have different effects under different situations. A wetted well-defined surface may quickly become a reflector which is easy to detect. However, tortuous flow paths may not be as ideal. Small structures might be well below the limits of detectability in the complex reflection pattern. As such the structural similarity attribute can only detect zones of significant changes which can be induced by many lumped small structures, one big flow path, or even a favorably oriented stone which gets wetted.
The authors declare that they have no conflict of interest.
Thanks to Elly Karle and the Engler-Bunte-Institute, KIT, for the IC measurements of bromide. We are grateful to Selina Baldauf, Marcel Delock, Razije Fiden, Barbara Herbstritt, Lisei Köhn, Jonas Lanz, Francois Nyobeu, Marvin Reich and Begona Lorente Sistiaga for their support in the lab and during fieldwork, as well as Markus Morgner and Jean Francois Iffly for technical support and Britta Kattenstroth for hydrometeorological data acquisition. Laurent Pfister and Jean-Francois Iffly from the Luxembourg Institute of Science and Technology (LIST) are acknowledged for organizing the permissions for the experiments. Moreover, we thank Markus Weiler (University of Freiburg) for his strong support during the planning of the hillslope experiment and the preparation of the manuscript. This study is part of the DFG-funded CAOS project “From Catchments as Organised Systems to Models based on Dynamic Functional Units” (FOR 1598). The manuscript was substantially improved based on the critical and constructive comments of the anonymous reviewers, Christian Stamm and Alexander Zimmermann, and the editor Ross Woods during the open review process, which is highly appreciated. The article processing charges for this open-access publication were covered by a Research Centre of the Helmholtz Association. Edited by: Ross Woods Reviewed by: Christian Stamm and three anonymous referees