Introduction
Managing water resources in a changing world (Cassardo and Jones
2011), water security (IHP
2012), change in hydrology and society (Montanari et al.
2013) as well as the need for integrated water resources management (EU WFD
2000) are the challenges for hydrological research in the next decades. The water cycle at local, regional and global scales is increasingly under pressure. Climate and land use changes and other global drivers, e.g., population growth and rapid urbanization, will put pressure on water resources with a tremendous impact on the natural environment (IHP
2012). To understand the functioning of the hydrological cycle under the ongoing changing conditions, hydrology has made enormous efforts with regard to small relatively homogeneous systems over a relative short timescale. In the next decade, hydrological research has to focus on the understanding of complex systems over much longer timescales (Wagener et al.
2010; Milly et al.
2008) considering nonlinearities, heterogeneities and highly dynamic processes.
Hydrology in complex settings
Hydrology builds on a wealth of process studies that revealed numerous causal relationships and often rather complex, nonlinear interactions between different processes or influencing factors. Consequently, water resources management has to face the challenge to differentiate between different effects that occur in parallel. Different effects might add to each other in some cases and might compensate each other in others. Often considerable lag times have to be accounted for as well. On the other hand, climate change as well as massive anthropogenic effects, both intended and unintended, and partly being on the way for centuries or even millennia in most parts of the world, needs to be accounted for.
Hydrologists have developed a whole zoo of powerful models. These models can handle many of the challenges even for complex systems. In practice, however, there is a substantial misbalance between the potential of spatially distributed models and the available data to condition and to constrain them. Consequently, models often are massively over-parameterized due to a lack of data and of knowledge of the respective local conditions, leading to substantial uncertainty (Beven
2001). In addition, models combine findings from preceding studies about single causal relationships. However, feedback between different processes is often hard to quantify in empirical studies in complex settings. Consequently, feedbacks in models are often necessarily implemented on experts’ judgement rather than on sound data.
Landscape hydrology approaches
The term “landscape hydrology” has been introduced for a scientific approach to meet the challenges given above. In contrast to, e.g., “catchment hydrology”, the term “landscape hydrology” usually does not only imply a larger spatial scale, but very heterogeneous settings as well (Hyndman et al.
2007), including very different soil types (Terribile et al.
2011), the nexus between groundwater and surface water systems (Hyndman et al.
2007; McLaughlin et al.
2014; Yuan et al.
2015), or large-scale feedbacks like evapotranspiration as a source for rainfall in other regions (Woodward et al.
2014). Thus, landscapes are primarily considered as highly interconnected systems comprising numerous abiotic, biotic and anthropogenic elements. Due to the large number of feedback links between landscape elements, landscapes are far from being random ensembles of elements. Instead, they are highly constrained, that is, the number of effective degrees of freedom is much smaller than one could expect with regard to the number of landscape elements (Lischeid et al.
2015).
Landscape hydrology aims at systematically exploring these constraints and at making efficient use of them. Following that approach has in many cases revealed that although the structure of hydrological systems might be complex, hydrological functioning often is surprisingly simple or low dimensional. The term “functioning” is used for time series of hydrological variables like discharge, groundwater head or soil water content. On the one hand, the effects of small-scale heterogeneities often level off at larger scales (Hohenbrink and Lischeid
2015). On the other hand, among a variety of processes that affect hydrological functioning at the landscape scale usually only a few prevail and need to be considered (Seyfried and Wilcox
1995; Blöschl
2001; Sivakumar
2004).
Taking that seriously implies that hydrological functioning should reflect the effects of the prevailing processes and could in turn be used to infer major features that are relevant for hydrological functioning. For example, time series of groundwater head, lake water level and stream runoff do not only depend on geological structures, but could be used to reveal information about major geological features. This information can then be used, e.g., to reduce the uncertainty of a hydrological model.
This is the basis of the forensic hydrology approach. The term “forensic” came up in the environmental sciences in the 1970s in the context of identifying causes of environmental pollution (Hurst
2007). In the case of hydrology, it usually requires an identification of flowpaths and flow velocities, or residence time, respectively (Hurst
2007). In addition, the term “forensic hydrology” has been used for analysis of major floods and droughts (Loáiciga
2001; Borga et al.
2014; Ramirez and Herrera
2016). It is now used in a wider sense for systematic analysis of cause–effect relationships in complex settings in hydrology, combining a variety of different methods of direct and indirect inference (e.g. Kappel
2014) in a systematic way. Thus, hydrologists act as “geodetectives” (Hurst
2007), similar like detectives in a crime story, although not restricted to court cases. In this study, both well-known and rather simple as well as innovative sophisticated data analysis approaches were used and the results merged in a systematic way for developing a consistent conceptual model of the Quillow study in a very complex geological setting. This is a necessary prerequisite for assessing pathways of subsurface transport of contaminants and nutrients from arable fields and the respective residence time. Key questions are: Are the small lakes that are abundant in that region hydraulically connected to a regional aquifer? Are there major aquitards in the catchment, and how far do they extend? Which area does a given stream drain?
Data
Digital elevation model
For the Quillow catchment, except for the outermost western part, digital elevation data at high spatial resolution (1 m) were provided by Landesvermessung und Geobasisinformation Brandenburg (State Survey and Geospatial Basic Information Brandenburg; LGB) acting on behalf of the Landesamt für Umwelt, Gesundheit und Verbraucherschutz Brandenburg (State Office of Environment, Health and Consumer Protection Brandenburg; LUGV). The data are based on an airborne laser survey in spring 2011. They were used with kind permission of LGB Brandenburg, ©Geobasis-DE/LGB 2012.
For open water bodies, the digital elevation data approximate the water level during the time of the survey. Due to aberration of the laser beam, the error of water level determination might be in the order of 0.1 m or even more, especially for large water bodies. Water level in the streams was determined at 100-m intervals along the streams, resulting in 1273 elevation points.
In addition, water level in 1176 small lakes (kettle holes) was determined using the same data set. They were identified first manually on the basis of different digital maps including aerial views, biotope types and topographical maps, scale 1:10,000. Then, the elevation of the water level was determined at these points based on the digital elevation model. We did not make full use of the high spatial resolution of elevation data for streams in order to roughly balance the number of stream and lake water level data points and to achieve a more homogeneous coverage of the area.
Soil hydrological data
Soil water content had been measured automatically at 1-h intervals using TDR probes or FDR probes (site Kraatz) in 20–300 cm depth at seven different locations within the catchment (Table
1; Schindler et al.
2008). In addition, the same measurements were taken in a series of lysimeters located at the Dedelow research station (Fig.
1). Lysimeters were filled with homogenized substrates excavated at different sites in the catchment. Surface area is 1 m
2, depth 2 m. Lysimeters were vegetated with different crops. Table
1 gives an overview over site properties and measurement depths.
Table 1
Soil hydrological measurement sites
KF | Kraatz, foothill | Sandy loam to loam | Luvic Stagnosol (Colluvic, drainic) | Arable field | 60, 100, 200, 300 |
KM | Kraatz, midslope | Sandy loam | Stagnic Luvisol (eutric) | Arable field | 60, 100, 200, 300 |
KK | Kraatz, hilltop | Sandy loam | Calcic Luvisol (eutric) | Arable field | 60, 100, 200, 300 |
FeHo | Ferdinandshorst | Sand to sandy loam | Haplic Cambisol (eutric) | Arable field | 60, 100, 200 |
CH-K | Christianenhof | Sand to loamy sand | Haplic Cambisol (Dystric) | Pine forest, ca. 80 years old | 30, 50, 100, 200, 300 |
Sk-B | Schlepkow, beech | Sandy loam | Haplic Cambisol (Dystric) | Beech forest, ca. 40 years old | 30, 50, 100, 200, 300 |
Sk-M | Schlepkow, mixed forest | Sandy loam | Haplic Cambisol (Dystric) | Mixed forest, ca. 80 years old | 20, 60, 200, 300 |
Lys08 | Lysimeter, Dedelow research station | Loam | Haplic Luvisol | Arable field | 130, 160, 185 |
Lys14 | Lysimeter, Dedelow research station | Loam | Haplic Luvisol | Arable field | 130, 160, 185, 195 |
Lys25 | Lysimeter, Dedelow research station | Sandy loam | Haplic Stagnosol, drainic | Arable field | 130, 160, 185 |
Lys26 | Lysimeter, Dedelow research station | Sandy loam | Haplic Stagnosol, drainic | Arable field | 130, 160, 185 |
Lys30 | Lysimeter, Dedelow research station | Coarse sand | Haplic Cambisol, eutric, arenic | Arable field | 130, 160 |
Lys31 | Lysimeter, Dedelow research station | Coarse sand | Haplic Cambisol, eutric, arenic | Arable field | 130, 160, 185 |
Groundwater data
For this study, groundwater head data from two different data sets were used. Details are given in Table
2. The authors’ group has been operating four groundwater wells (198, 201, 203, 204) that are located in the central part of the catchment (Fig.
1). They have been equipped with automatic ventilated pressure transducers, recording groundwater head at daily intervals. For more details, the reader is referred to Merz and Steidl (
2015).
Table 2
Groundwater observation wells
198 | 75.27 | 24.00 | 22.00 | ZALF |
201 | 78.29 | 12.50 | 14.50 | ZALF |
203 | 79.46 | 16.00 | 18.00 | ZALF |
204 | 65.00 | 16.00 | 18.00 | ZALF |
25470023 | 100.33 | 37.80 | 35.80 | LUGV |
25470024 | 100.38 | 54.7 | 52.70 | LUGV |
26471092 | 96.45 | 20.10 | 17.10 | LUGV |
26471094 | 101.33 | 38.79 | 36.79 | LUGV |
26471095 | 97.55 | 35.58 | 33.58 | LUGV |
26471096 | 97.52 | 49.58 | 47.58 | LUGV |
26471097 | 86.00 | 23.61 | 21.61 | LUGV |
26471098 | 86.00 | 33.63 | 31.625 | LUGV |
26471099 | 86.00 | 42.45 | 40.45 | LUGV |
26480022 | 62.50 | 16.80 | 14.80 | LUGV |
26481052 | 56.64 | 42.28 | 40.28 | LUGV |
26486011 | 59.86 | 44.64 | 40.54 | LUGV |
26490030 | 17.40 | 9.30 | 7.30 | LUGV |
26490031 | 17.37 | 40.20 | 38.20 | LUGV |
26491061 | 40.87 | 61.62 | 59.62 | LUGV |
26491062 | 40.86 | 98.73 | 96.73 | LUGV |
26491066 | 50.20 | 53.80 | 49.80 | LUGV |
28481093 | 54.75 | 20.00 | 18.00 | LUGV |
In addition, groundwater head data were provided by the State Office of Environment, Health and Consumer Protection Brandenburg (LUGV). These wells were located within or close to the Quillow catchment (Fig.
1). Groundwater head had been determined by pressure transducers (daily intervals) or manually (once or four times per month).
Stream discharge
Water level of the Quillow stream has been measured at hourly intervals close to the Dedelow research station (Fig.
1) by an automatic ventilated pressure transducer. That site is located underneath a road bridge where the stream exhibits a clearly defined cross section. Thus, water level data could be converted to discharge based on a rating curve that has been regularly calibrated by current meter measurements.
In addition, differential stream gauging was performed at different measurement points in the main stream of the catchment during baseflow conditions. A current meter was used to determine flow velocity at different positions in cross sections of the stream. The measured values of flow velocity were integrated over the respective cross-sectional area. As this approach was meant to yield a quick first insight into spatial patterns along the stream, no sound error analysis was performed. Based on experience with that approach in preceding studies, an uncertainty range of about 10 to 20% is assumed.
Methods
The methods used in this study are only briefly described here. Please refer to the cited papers for methodological details. All analyses and graphs presented in this study were performed using the R software (R Core Team
2014).
Principal component analysis of hydrological time series
Hydrological time series usually exhibit substantial spatial heterogeneity. That might be due to the spatial heterogeneity of rainfall, interception, evapotranspiration, soil and aquifer properties as well as anthropogenic impacts. Principal component analysis (PCA) of hydrological time series aims at exploiting the observed heterogeneity as a source of information for quantitative assessments of the respective effects. In other disciplines, terms like Empirical Orthogonal Functions or Karhunen–Loève transformation are used instead of PCA.
In mathematical terms, a principal component analysis performs an eigenvalue decomposition of a covariance matrix of a set of variables into a set of independent principal components. When time series are subjected to a PCA, the resulting principal components constitute time series as well. Any time series of the input data set can then be represented as a linear combination of the principal components without any loss of information. The respective weighting factors reflect the bivariate correlation (called “loadings”) between observed time series and principal components as well as the eigenvalues of the principal components. The sum of the eigenvalues of selected principal components over the sum of the eigenvalues of all components is equal to the fraction of variance of the total data set explained by the respective components. Principal components usually are sorted by decreasing eigenvalues, that is, the most important components are listed first.
Principal component analysis has been used in hydrology for many decades but has only rarely been applied to time series, in contrast to other disciplines like climatology. However, it has proven to be a powerful tool to identify and to quantify the impact of river water stage on riparian groundwater head (Lehr et al.
2015), of groundwater production wells on adjacent lakes (Böttcher et al.
2014), of different crops and tillage practices on soil water content (Hohenbrink et al.
2016), or of climatic gradients on stream discharge (Thomas et al.
2012). Application of PCA on a data set comprising soil matrix potential, groundwater head and stream discharge nicely illustrated the hydrological continuum within a catchment (Lischeid et al.
2017).
When PCA is applied to time series, the first component usually is similar to a time series consisting of the spatial averages of normalized observed values per time step. Subsequent components reflect then different effects that cause deviation from that mean behaviour each (Hohenbrink et al.
2016). One of these processes that is prominent in groundwater head and soil hydrological data sets is increasing transformation of the input signal (e.g. rainfall or snowmelt) with depth in the vadose zone, that is, increasing attenuation of the amplitudes and deceleration of the signal (Hohenbrink and Lischeid
2015; Lischeid et al.
2010). This phenomenon will be called “damping” in the following.
Principal component analysis of time series does not require equidistant time series. However, the data need to be synchronous, that is, measurement dates need to be identical for all observed variables. To ensure equal weighing of all observables, each time series was normalized to unit variance and zero mean prior the analysis. For more detailed information about that approach, the reader is referred to Lischeid et al. (
2010), Hohenbrink and Lischeid (
2015) and Hohenbrink et al. (
2016).
Differential stream gauging
Net inflow of groundwater into a stream, or net loss of stream water due to seepage or discharge to the aquifer within a given stream reach can be assessed by the net difference of discharge at both ends of the respective stream reach. In addition, tracer tests can be used to assess gross inflow (Bencala et al.
2011; Bergstrom et al.
2016). It is recommended to perform these measurements during baseflow (Cey et al.
1998; Ruehl et al.
2006; Kalbus et al.
2006; Kikuchi et al.
2012) to minimize the effect of surface runoff or interflow feeding the stream. In addition, short-term temporal variability should be negligible as discharge measurements are usually taken consecutively.
Applicability of that approach often is limited by the limited precision of discharge measurements. On the other hand, it allows rapid assessment of groundwater—stream interaction at medium to large scale with little effort. Thus, it has been widely used in hydrology. However, there is no uniform nomenclature. The approach has been termed “longitudinal variation in discharge” (Zellweger et al.
1998), “measurements of stream discharge along the reach” (Cey et al.
1998), “seepage runs” (Ruehl et al.
2006), “incremental stream flow” (Kalbus et al.
2006), “differential flow gauging” (McCallum et al.
2012) or “differential discharge measurements” (Kikuchi et al.
2012). Here the term “differential stream gauging” will be used.
Comparison of groundwater head and surface water level
According to Darcy’s law, hydraulic gradients in the subsurface are determined by groundwater flow density and hydraulic conductivity of the respective substrate. Small-scale heterogeneities of the latter are likely to level out at larger scale, resulting in a fairly smooth groundwater head surface (e.g. Fleckenstein et al.
2006). Consequently, major deviations of groundwater head or lake water level at selected sites from the mean groundwater surface provide strong evidence for missing hydraulic links between the respective hydrological systems.
In a first step, comparing the shape and smoothness of the respective surfaces could provide substantial evidence for or against hydraulic connection between surface water systems and groundwater. As a quantitative measure of the smoothness variograms were used. Variograms describe the variance of spatial data as a function of distance between respective data points. Spatial variance is low for data sets with a smooth surface and for small distances, tending to increase with distance. In contrast, for abrasive surfaces spatial variance is close to the maximum already for very short distances.
Variogram values
\(\upgamma\) as a function of distance
\(h\) are calculated as mean squared distances between two points
x
i
,
x
j
$$\gamma \left( h \right) = \frac{1}{2N\left( h \right)} \cdot \mathop \sum \limits_{i,j = 1}^{N(h)} (x_{i} - x_{j} )^{2}$$
usually using a rather small number of distance classes with
\({N\left(h\right)}\) data points each for pairwise comparison.
The data of the variogram of the observed values were fitted to an exponential model given by
$$\gamma \left( h \right) = c \cdot \left( {1 - e^{{\frac{ - h}{a}}} } \right)$$
where
\(h\) denotes distance, c denotes the sill, and a denotes the range. For large data sets and large distances, the sill approaches the variance of the data. For the exponential model, the range is a measure of how rapidly spatial variance increases with distance.
As low-frequency patterns and trends would mask the patterns of spatial variance as a function of distance, spatial data were detrended prior the analysis by linear regression with the northing and easting coordinates.
The total number of data points of small lake water level was 1176, that is, some orders of magnitude smaller compared to the number of data points of the digital elevation model. Thus, 1000 data points of the latter were selected randomly in ten different realizations to be analysed via variogram and to be compared with that of the lake water level data.
Data of deep groundwater wells were too sparse to be interpolated. Here a different approach was used. For each of the wells, mean groundwater level was compared to linear interpolation between lake and stream water level points in up to 1 km distance. The difference between these respective two approaches was compared to the square root of the error variance of the linear interpolation, performed without the groundwater head data.
Discussion: Resulting conceptual model and implications for modelling
According to the results of the principal component analysis, the time series of soil water content exhibited substantial spatial heterogeneity. Although in general the damping of the input signal tended to increase with depth, soil water content at one site in 185 cm depth behaved like that at 60 cm depth at another site (Fig.
3). Much of that heterogeneity can be ascribed to the enormous heterogeneity of unconsolidated sediments (Merz et al.
2009) and soils in this Pleistocenic landscape, ranging from coarse sandy soils to clayey loam soils. In addition, erosion processes resulted in substantial spatial heterogeneity at the range of a few 10 m as well (Sommer et al.
2008). Consequently, soil hydrological and chemical processes exhibit enormous spatial heterogeneity (Rieckh et al.
2012; Gerke et al.
2016) in this region.
However, except for some of the lysimeters, there was no evidence for systematic differences between different sites, even between arable field and forest sites (Fig.
2). This could have been masked by substantial within-site heterogeneity. Only two of the lysimeters that were filled with coarse sand clearly stood out, indicating that only extremes of soil texture needed to be taken into account. Applying the same approach to another soil hydrological data set, Hohenbrink et al. (
2016) found clear differences between two crop rotation schemes. However, that difference accounted only for 3.6% of the total variance. It is very likely that a difference of that magnitude would not been detectable in the Quillow data set given the enormous spatial variability. However, these differences of soil hydrological dynamics must not be mismatched with differences of the total sum of deep seepage or groundwater recharge. In fact, Schindler et al. (
2008) found clear differences of deep seepage rates between forest sites and arable fields as well as between different texture classes in a data set that comprised among others the Quillow data used for this study.
Thomas et al. (
2012) used the same approach and found clear evidence for a climatic gradient reflected by respective components when analysing hydrographs from all over the Federal State of Brandenburg. Thus, although the annual sum of precipitation is known to decrease from west to east in the catchment (Sommer et al.
2008), the temporal patterns obviously are the same.
In addition, time series of groundwater head and of discharge plot right in between those of soil water content time series in Fig.
3, illustrating a hydrological continuity between vadose zone, aquifer and stream. According to Fig.
3, the hydrograph is substantially less damped compared to the latters. Damping of the hydrograph corresponds more to that of soil water content at approximately 1.5 m depth. It can be concluded that runoff generation to a large degree occurs at rather shallow soil depth. Here tile drains might play an important role.
Comparing water levels in kettle holes and streams, there was strong evidence of a common hydrological system (Fig.
6). In addition, the smoothness of the surface spanned by surface water level compared to that of topography supported that inference (Fig.
8). In contrast, groundwater head in wells screened at greater depth were substantially lower except for three wells that were located close to the stream in the eastern lowland part of this and in an adjacent catchment (Figs.
1,
7). On the other hand, deep groundwater wells close to the stream in the more upstream part of the catchment obviously are neither directly hydraulically linked to the stream nor to the kettle holes (Figs.
1,
7).
The line of demarcation of these two parts of the catchment is approximately identical with the upper end of the gaining reaches of the stream between stream measurement points 13 and 12 (Fig.
5). Combining these findings with those of the discrepancy between groundwater heads and surface water level in the upstream parts of the catchment suggests that upstream that demarcation line one or more aquitards separate the uppermost aquifer from a deeper one. That aquitard(s) obviously thin(s) out upstream the demarcation line, and about 2 km downstream an area of high kettle hole density that stretches from North to South, perpendicular to the main stream (Fig.
1). This is an area that is drained by some of the major tributaries of the Quillow stream (Fig.
1). That feature could indicate that the uppermost aquifer thins out here, e.g., due to intersection of topography with an approximately horizontal lower confining bed.
The line of intersection seems to be located close to well 26486011, the only artesian well in the catchment. About 2.5 km downstream of this well, close to stream measurement point 8 (Fig.
5), a fen exists close to the stream from which water discharges to the stream. Substantial efforts to drain that fen by trenching to separate it from the presumed interflow or shallow groundwater from the adjacent hillslope have not been successful. Fairly high electric conductivity of the discharging water from the fen points actually to deep groundwater that discharges here but presumably had recharged in more upstream parts of the catchment.
According to Fig.
4, the degree of damping found at the artesian well 26486011 points to recharge in an area rather far from this site where the thickness of the vadose zone is about 6–7 m. This points to a substantial lateral extent of the upper confining layer. A corresponding mismatch between the degree of damping of groundwater head time series and depth of pressure head below surface has been found at wells 198, 201, 203 and 204 as well. They are all located close to the stream, up to 5 km upstream the artesian well (Fig.
1). The degree of damping found at these wells points to recharge substantially further uphill where the thickness of the vadose zone is up to 13 m approximately (Fig.
4). High electric conductivity and absence of oxygen and nitrate in these wells (Merz and Steidl
2015) are indicative for long groundwater residence time, supporting this inference.
These findings can be summarized as follows. Hydrological processes in the topsoil exhibit substantial spatial heterogeneity, urging for a large number of replicates of soil hydrological data for model calibration. Besides, there is no evidence for clear spatial patterns within the catchment that need to be considered. In general, damping of the input signal increases with depth in the vadose zone and seems to exert first-order control on the dynamics of groundwater head and stream discharge.
Kettle holes and streams are part of a common uppermost shallow hydrological system that is separated from an underlying major aquifer in the western part of the catchment and close to the catchment boundary in the eastern part as well. However, the underlying confining layer must be leaky. Otherwise neither periodical drying-up of stream reaches in this part of the catchment nor recharge of the underlying deeper aquifer would be possible. In contrast, in the central western part of the catchment, there was strong evidence for an extended tight confining layer.
Conclusions
Landscape hydrology considers landscapes as systems being subject to numerous feedback links, that is, as highly constrained systems, and applies modern methods of system analysis to make efficient use of the available data. That approach opens a pathway to handle systems with very complex structure like that of the thick unconsolidated Pleistocenic sediments in North Central Europe. An example was given in this study.
Although we observed substantial small-scale spatial heterogeneity of soil moisture dynamics in the vadose zone, there was no evidence of systematic differences between different plots within the catchment, even not for different land use classes. In addition, measurements on field sites did not differ systematically from those in lysimeters filled with homogenized material, except for lysimeters filled with coarse sand. Discharge dynamics corresponded to that of soil moisture at 1.5 m depth, emphasizing the role of shallow flowpaths including tile drains for runoff generation. Small lakes (kettle holes) that are abundant in the catchment are generally hydraulically connected to the groundwater system. The main aquifer is unconfined in the eastern half of the catchment, but is confined and separated from an overlying shallow aquifer in the western part that is linked to lakes and streams. This information is essential for assessing nutrient and agricultural contaminant transport and turnover in the catchment, including the fate of pesticides. Using this information as a starting point for modelling studies, model uncertainty can be substantially reduced right from the beginning.
Landscape hydrology is not restricted to a single tool, like, e.g., a certain type of model, that could be applied in a schematic way for solving any problem. Rather it aims at systematically determining the constraints of the respective hydrological system currently being under study in order to achieve an internally consistent conceptual model. This is realized by means of a well-balanced set of complementary methods and tools of hydrological systems analysis like demonstrated in the Quillow study. It might not be spectacular at the first sight but is a very promising pathway to follow for the sake of sound understanding of given hydrological systems and for solving real-world problems for the benefit of society and nature.
However, likewise there is urgent need for developing advanced tools for analysis of hydrological behaviour and for determining constraints of the given system in a more sophisticated way. This approach is called “forensic hydrology”. Like in criminalistics, hydrologists need to be well equipped with a variety of powerful diagnostic tools to be able to meet the demands of modern society.
Acknowledgements
The authors are deeply indebted to Roswitha Schulz, Dorith Henning, Ralph Tauschke, Joachim Bartelt, Bernd Schwien and numerous students who did most of the field work, partly under very harsh conditions. We highly appreciate the groundwater head and discharge data provided by the Landesamt für Umwelt, Gesundheit und Verbraucherschutz Brandenburg (State Office of Environment, Health and Consumer Protection Brandenburg; LUGV) and the high-resolution digital elevation data by Landesvermessung und Geobasisinformation Brandenburg (State Survey and Geospatial Basic Information Brandenburg; LGB) acting on behalf of the State Office of Environment, Health and Consumer Protection Brandenburg. Part of the data used for this study have been collected within or associated with the LandScales project, financed by the Leibniz Association, which is highly appreciated.