http://duerreych.bau-verm.uni-karlsruhe.de
e-mail: markus.casper@bau-verm.uni-karlsruhe.de
Fax: 0049 721 661329
Tel.: 0049 721 608-4222

Abstract
Introduction
Organic Substances in Surface Waters
Site description
Methods
Results and Discussion
Conclusions and Outlook
References
 
 

In order to distinguish runoff generation processes, investigations of the spatial and temporal dynamics of dissolved organic carbon (DOC) export and d ¹⁸O (VSMOW) in precipitation, soil water and runoff were carried out in the catchment Duerreychbach (northern Black Forest, Germany). For a headwater DOC measurements during selected precipitation-runoff events were used to distinguish the dominating runoff processes and their chronology. The DOC-rich water originated from the upper soil horizons and acted as a rapid runoff component. Runoff was found to be dominated by lateral percolation through the upper soil horizons The combination of DOC and d ¹⁸O data in a mixing diagram allowed to quantify the fraction of event water. In opposition to d ¹⁸O, DOC concentrations always showed a distinct contrast between soil water and precipitation water. Determination of a representative mean pre-event value for both DOC and d ¹⁸O was difficult, but the possible range of soil DOC concentrations was rather limited and fell between 68 to 80 mg 1^-1 in our study. Using a DOC mass balance approach, a hydrograph separation for the Duerreychbach catchment succeeded for low antecedent soil moisture or low precipitation sums: The headwater covering only 22% of the catchment delivered approx. 98% of the total DOC load and more than 50% of the water. Increasing precipitation sums or wetter antecedent soil moisture increased the connectivity of other areas to the main channel, which then contributed about 52% of the exported DOC.

Knowledge of runoff generating processes is fundamental for testing and improving predictive rainfall-runoff models. Moreover, expected changes in climate or land use require models, which are capable of predicting the resulting changes in runoff dynamics and water quality. Therefore, the knowledge of the dominating runoff generation processes and their interactions is important in order to prove a model's applicability for these purposes (Bonell, 1998; Buttle, 1994; Mehlhorn et al.,1999).

To identify the dominating runoff processes, flow paths need to be tracked. Here the chemistry of water solutes can provide valuable information (Bishop 1991; McDowell & Likens, 1988; Easthouse et al., 1992). In addition, the analysis of stable isotopes is a common method to provide data on sources of runoff (Bishop, 1991; Buttle, 1994; Ambroise et al., 1995; Brown et al., 1999). It allows to separate event water (rain or throughfall) and pre-event water which was stored as soil or ground water prior to the event. However, a significant contrast in isotopic composition of event-water and pre-event water is essential.

Hydrometric, isotopic and hydrochemical data complement to each other. Hence, a data set covering a wide range of event characteristics and antecedent soil moisture conditions of the catchment helps to elucidate the sources and flow paths of water and allows to understand the relative contributions and the dynamics of the different runoff generation processes (Bonell, 1998).

In this study selected rainfall-runoff events were analysed to describe the sequence of dominating hydrological processes, to quantify surface runoff, and to estimate the contribution of three nested subcatchments to runoff. Investigations with the natural tracers d¹⁸O (VSMOW) and dissolved organic carbon (DOC) were performed to estimate proportions of event and pre-event water during runoff events. The analysis of DOC dynamics was used to describe the sequence of dominating hydrological processes and to quantify surface runoff.
 
 

Natural organic substances play a major role in the description of ecosystems, since they characterise the linkage of organisms with their environment. Soil processes such as accumulation, turnover and export of organic carbon play an important role in the ecosystem, driven by water as the transport medium (McDowell and Wood, 1984). Thus knowledge of the fluxes of DOC can contribute significantly to the understanding of runoff generation (Hooper et al., 1990). For dissolved organic matter there exist various terms with overlapping or partially matching definitions. Perdue & Gjessing (1989) described DOC as complex mixture of natural dissolved organic acids. In this study DOC refers to the fraction of total organic carbon (TOC), that passes a 0.45 µm pore-diam-filter (EN 1484 ,1997; Eaton et al., 1995). This definition also comprises the fraction of humic substances, which are refered to as coloured organic substances of natural origin, derived from biogenetic transformation (Allard et al., 1991), or sum of the fractions humic and fulvic acids, obtained by a fractionation procedure of organic substances with XAD-8 resin (Thurmann, 1981; Malcom, 1989). DOC occurs in soil water, natural aquifers and drinking water in concentrations between 0.1 to 80 mg l-1 (Wetzel and Corners, 1977; McDowell and Likens, 1988; Scott et al., 1998). In temporate regions the portion of DOC in streams usually exceeds the particulate organic matter by one order of magnitude. In aquatic systems the significance of DOC is abundant. In the food chain for example it plays a central role. Low order streams like the Duerreychbach are basically heterotrophic ecosystems that obtain almost all of their DOC from terrestrial primary production (Hynes, 1963), which continuously delivers organic matter. Partly transformed by biological degradation soluble fractions reach the stream, carried by precipitation water through the soil (Mulholland et al., 1990). In their study Fiebig and Lock (1991) described immobilisation dynamics of DOC in the stream bed. They observed in laboratory experiments, that substantial amounts of DOC in groundwater can be adsorbed to the stream sediment and consumed by microorganisms. Nevertheless due to rapid degradation in the soil the fraction of bioavailable organic carbon in the stream water is usually below 10% of the total organic carbon, based on steady state concentrations (Frimmel, 1990). Especially in waters with a high load of humic substances the recalcitrant components dominate. In particular during stormflow the Duerreychbach carries water with a dark brown colour caused by the high concentration of humic substances. On the other hand adsorption and desorption dynamics of organic solutes at sediment surfaces have to be considered. Own extractions of DOC with NaOH have shown, that sediment surfaces of both, streambed and spring sediments of the Duerreychbach were highly loaded with humic substances, revealing rather slow desorption processes. The amount of humic substances indicated, that the sediment surfaces were almost saturated. Since travel time of water from the headwaters to the main gauge seldom exceeded a few hours, we considered DOC to be a conservative water solute. The biological stability and the variability of mobilisation processes depending on soil characteristics are preconditions for the use of DOC as a tracer of water from different source areas and different flow paths through the soil into the stream.

Figure 1: Site map of the hydrological test site Duerreychbachtal: subcatchment delineation and instrumentation

The Duerreychbach catchment is a 700 ha V-shaped valley covered with coniferous forest in the northern Black Forest, SW-Germany, with elevations between 600 and 950 m a.m.s.l.

Mean precipitation is 1530 mm a^-1 with a mean temperature of 6-7°C. The bedrock is Triassic sandstone covered by deep blocky periglacial regolith. The very steep slopes are carved by corries and topped with humid plateaux. Typical soils are podzols, stagnic gleysols or degraded peats. For this study the catchment was subdivided into two nested headwater catchments LE (27 ha) and SH (156 ha) and the subcatchment DUE (544 ha) (Figure 1).

The headwater LE is characterised by a flat humid plateau and relatively flat hillslopes. It is closely linked to the stream by a dense artificial drainage system. Soils with stagnic or peaty properties cover 50% of the area (Figure 2). Discharge was only activated during periods of high soil moisture content or after a heavy rainfall. The catchment SH includes LE and shows similar characteristics. 32% of the soils have peaty and stagnic properties. In addition, it comprises a plain which was deforested by a wind break in 1991. Though, during wet periods a small periodic spring contributes to the catchment discharge, the stream channel falls dry periodically.

Figure 2: Distribution of peaty or stagnic soils in the Duerreychbach catchment

The subcatchment DUE has steep slopes. Except for the alluvial plain of the narrow valley floor, soils with stagnic or peaty properties are only found in corries and on the mountain tops and cover only 10% of the area. They have a low connectivity to the main stream due to a sparse drainage system and great distances. Podzols are dominant on the forested slopes. Discharge is observed at the catchment outlet all year round. In dry periods only three perennial springs are contributing to runoff (Figure 2).
 
 

Field measurements

As shown in Figure 1, the study area was equipped with several gauging sites. Two weather stations provided data of precipitation, air temperature, soil temperature and soil moisture content. Soil moisture content was measured by TDR probes at 20 and 35 cm below the surface. V-notch weirs with a pressure gauge connected to ISCO autosamplers were used for measurements of water levels at LE and SH. A rectangular weir was installed at DUE. Runoff was calculated with a stage-discharge curve deduced from the weir geometry and calibrated with bucket measurements and the salt dilution method according to Rodel (1993). Using the salt dilution method flow velocity and discharge were measured temporary for a number of points along the main channel. Water samples were taken at intervals of 48 hours, and during runoff events at intervals of 3 hours. Soil water was collected weekly with PE/Nylon suction cups at depths between 5 and 70 cm of a podzol soil which is the dominant soil type of the study area.

Laboratory methods

Determination of DOC concentrations in runoff and soil water was carried out by IR detection of CO₂ after combustion at 950°C with synthetic air with a HighTOC analyser (Elementar, Hanau/Germany) with an accuracy of ±  2%. Atomic mass spectrometry was used to determine d ¹⁸O‰ (VSMOW) with a gas ion mass spectrometer and Aquaprep (Micromass Optima). The sample were prepared with the "glove bag technique". The accuracy is ±  0.15‰.

Delineation of subcatchments

Using a GIS (Arc/Info 7.2) and a hydrologically corrected, high resolution DEM (10 meters grid size) subwatersheds were delineated by defining the corresponding source points. The results were confirmed by field survey.

Hydrograph separation techniques

Different hydrograph separation techniques were applied: (1) two-component isotopic separation (Bishop, 1991; Brown et al., 1999), (2) two-component separation using DOC as a tracer, (3) two-tracer, two-component end-member mixing analysis (Christophersen and Hooper, 1992)

1. Two-component isotopic separation for the headwater LE.The technique (Bishop, 1991; Brown et al., 1999) is a mass-balance approach that separates stormflow into pre-event water (in our case: water stored in the soil prior to the event) and event water on the basis of the stable isotope ratio of each component:

(1)

where Q is discharge, C is the d ¹⁸O composition. The subscripts s, p, e represent the streamflow, pre-event and event components, respectively. The relative contributions of pre-event and event water were calculated for each sample. Since the isotopic composition of the pre-event (soil water) component was unknown (no discharge during dry periods) it was determined using a DOC-d¹⁸O mixing diagram. The isotopic composition of event water was calculated as the weighted mean of the precipitation samples

2. Two-component separation using DOC as tracer. Since runoff from the plateau area (headwaters LE + SH) showed a stable (only +/-10% variation during event) and very high concentrations of DOC, we used DOC similarly to d ¹⁸O: for the catchment outlet we calculated the instantaneous fraction X(t) of runoff which originates from those areas:

(2)

where:

C_DOC(t): DOC concentration at catchment outlet, C_DOC(t = 0): DOC background concentration at catchment outlet and C_DOC(Plateau): constant DOC concentration from export area.

3. Two-tracer, two-component end-member mixing analysis (Christophersen and Hooper, 1992). To estimate the isotopic composition of pre-event water, both isotope and DOC data were analysed. Since DOC and d¹⁸O show a strong correlation we determined the mixing line for pre-event and event water by plotting all measurements in a bivariate DOC/d¹⁸O diagram (Figures 6b and 7b). Since the exact position of the end member "soil water" was unknown, we incorporated additional information. First we used d¹⁸O data from a soil profile nearby. Second we made the assumption that the last pre-event measurement of low flow DOC concentration is the lower limit for the actual DOC concentration of the soil water. This allowed us to find a better estimation for the actual d¹⁸O content of soil water by plotting the DOC value on the mixing line and reading the corresponding d¹⁸O value (detailed example in Figure 6b).

Spatial and temporal distribution of DOC concentrations

DOC concentrations measured in the stream varied between 0.5 and 81 mg l^-1. The highest concentrations were observed during stormflow in the headwater catchments. The DOC concentrations of soil water decreased with increasing soil depth. In the soil water of a podzolic soil profile highest concentrations and dynamics of DOC were obtained in the upper soil horizons. DOC concentrations of a perennial spring in the Duerreych valley never exceeded 1 mg C l^-1. This is coherent with earlier studies about the adsorption kinetics and sources of DOC in soils (McDowell and Wood, 1984; McDowell and Likens, 1988; Kaplan et al., 1980; Dalva and Moore, 1991): The deeper the water percolates through the mineral soil, the more DOC is adsorbed to the soil body and degraded Thus in the Duerreychbach stream water with a high DOC concentration must originat from the upper soil horizons and traced rapid runoff components (Brown et al., 1999).

Headwater LE

Figure 3: Soil moisture response on a precipitation input

The two nested headwaters LE and SH showed a very similar hydrological and hydrochemical behaviour. For the following description of runoff processes, we focus on the headwater LE.

Soil moisture reaction to precipitation input is shown in Figure 3. Initially, percolating rain caused only a slight increase in water content. Soil moisture rose quickly to total saturation, starting in greater soil depths, gradually saturating the soil body completely. The time lag between main precipitation input and total saturation indicated strong lateral flow components. Due to low slope and the high infiltration capacity, the headwater represents a potential area of saturation excess flow (Peschke, 1999). Hortonian overland flow was excluded.

Three typical rainfall-runoff events (events A,C and K in Table I) and the corresponding soil moisture, discharge and DOC concentration of the LE catchment are shown in Figure 4.
 
 

Table I. Event characteristics.
Event number	Date	Precip. sum [mm]	Max. intensity [mm h-1]	Pre-event condition *
A	5.9.1998	20	7	Dry
B	5.10.1998	28	5	(very) Wet
C	24.10.1998	43	4	Intermediate
D	11.5.1999	60	6	Wet
E	7.7.1999	27	13	Intermediate
F	14.7.1999	39	21	Intermediate
G	4.8.1999	42	30	Dry
H	22.9.1999	8	2	Dry
J	3.10.1999	17	4	Intermediate
K	16.9.2000	100	12	Dry
* Wet: wetness front above 20cm depth; Intermediate: wetness front between 20 and 35 cm depth; Dry: wetness front below 35cm depth (TDR probes at weather station in headwater SH)

Figure 4: Three precipitation-runoff events as examples for the reaction of the catchment on precipitation (events A,C and K, Table I)

The precipitation input was 20 mm, 43 mm and 100 mm respectively, causing rapid runoff reactions. Despite of the different precipitation intensities all three events showed the same decrease of DOC concentration during precipitation on water saturated soil. Using event C from Figure 4, the sequence of the runoff processes are described in Figure 5.

At the beginning of runoff the DOC concentration was low. Runoff consisted mostly of surface runoff on forest tracks (Figure 5 (ix)), channel precipitation, drainage of deeper horizons and a few small saturated areas near the stream (vi, vii). The second rainfall peak caused complete saturation of the soil. A maximum concentration of DOC in the stream was reached because lateral runoff components through the organic layer (v) became dominant, discharging old soil water and slowly replacing it. Further infiltration was inhibited by water saturation, causing surface runoff in the form of saturation excess flow (iii) , which diluted the DOC-rich soil water with precipitation water. A new maximum of the DOC concentration was reached when surface runoff ceased. Decreasing levels of saturation due to lateral drainage (v, vi, vii) and groundwater recharge (viii) lowered the level of saturation and caused a decrease of DOC concentrations because of an increasing dominance of deeper runoff components. This is different to the study of Peters et al. (1995), where dilution effects of laterally drained soil water due to flushing was reported. In the headwater LE this process was considered to be of minor importance. In other catchments the riparian zone in proximity of the stream is regarded to be the major source of DOC (Bishop et al., 1994; Fiebig et al., 1990). However, in the headwater LE the DOC contributing area is too large to be exploited only by lateral flow. Nevertheless, in addition to a vertical pattern of runoff contributors, also a spatial pattern must be expected. The principal processes of runoff generation can be detected on a plot scale. But the heterogeneity of the headwater LE and the increasing responding time with increasing distance from the stream and the gauging point might also contribute to the observed increase in DOC in the first part of the chemograph and the decreasing DOC concentration at the end. However, the data of DOC concentrations in soil and stream water are not sufficient to quantify different lateral subsurface runoff components. Here it would be necessary to find appropriate methods to characterise DOC from different origin of mobilisation that can be found in both, soil and stream water (Gron et al., 1996; Bishop et al., 1994).
 

Table II. Portion of event water calculated from d 18O and DOC data (headwater LE).
Event number	Date	Pre-event condition	Percentage runoff (27 ha)	Percentage event water	Max. DOC conc.
H	22.9.1999	Dry	0	0	0
J	3.10.1999	Intermediate	8	15-20	76.2
E	7.7.1999	Intermediate	10	22-30	58.7
B	5.10.1998	Very wet	33	25-32	60.5
D	11.5.1999	Wet	26	25-35	51.7
K	16.9.2000	Dry	34	30 (DOC only)	52.5
G	4.8.1999	Dry	6	50-60	42.7

For a number of 6 events we measured both, DOC and d¹⁸O concentration in order to combine the potential of both tracers (Table II). Two events are discussed in detail. One thunderstorm (event F, Table I) on dry soil conditions (Figure 6a) showed an exceptionally high portion of event water (50 to 60%) and a small percentage runoff of approx. 6%. Most precipitation on the soil caused an increase in soil moisture without contributing to runoff, while runoff on tracks was proportional to the amount of rain. This means that only saturated areas close to the drainage system contributed to runoff. Figure 6b shows the adjustment of the isotopic composition of soil water. Since the measured d¹⁸O concentration from a soil profile nearby didn’t fit exactly on the mixing line, we adjusted the d ¹⁸O value using a measured pre-event streamwater DOC concentration of 75 mg l^-1. The diagram also shows that uncertainties in the determination of pre-event DOC (70 to 80 mg l^-1) resulted in an error of only ±0.1‰ for d ¹⁸O which –in this case- is below the analytical error of d¹⁸O.
 
 

Figure 6: Hydrograph separation at LE for wet pre event conditions (event G, Table I), (a) DOC and runoff reaction to precipitation input, (b) adjustment of d¹⁸O using pre-event DOC

An advective rain (event B, Table I) on wet soil (Figure 7a) produced high discharge volumes containing a mean of 25 to 32% event water. Uncertainties in the determination of pre-event DOC concentration (69 to 78 mg l^-1) resulted in an error of less than ±0.3‰ for d¹⁸O (Figure 7b). The percentage runoff reached 33%, indicating that the area of saturation excess flow expanded to locations not directly connected to the main drainage system. More pre-event water ( = soil water) contributed to stormflow in comparison to the previous event (Figure 6).

Table II shows the fractions of event water in the headwater LE for a range of events. On dry antecedent soil moisture conditions discharge started only after filling of the soil storage (event H with no runoff reaction). A thunderstorm on dry soil (event F) activated only the drainage near areas causing a high fraction of event water. Wetter pre-event conditions resulted in higher discharge volumes but lower portions of event water, indicating that more soil water could be exchanged during the event (events B, D, K). Especially for smaller events on intermediate conditions, the fraction of event water was reduced because less surface runoff was produced (event J). In general, higher portion of pre-event water from the upper soil layers were also verified by higher DOC concentrations (last column in Table II).
 
 

Hydrometric, DOC and d¹⁸O data showed that soils with stagnic and peaty properties controlled the runoff processes. During a rainfall event these soils showed saturation caused by low hydraulic conductivity in deeper soil layers. The dominating runoff generation process is saturation excess flow (surface and shallow subsurface runoff) with a high degree of soil water exchange during events, indicated by the high DOC load and a low pH. The fast response of the headwater was a result of the dense artificial drainage system. It enhanced the runoff reaction by a better connectivity to the main stream. Although the mixing process and mixing locations of event water and soil water remain unclear, we suggest a mixing at the surface-soil-interface (e.g. due to return flow), and a mixing in the stream, where surface runoff and lateral subsoil inputs meet.

Catchment scale

On the catchment scale we used a mass balance approach of DOC for a hydrograph separation. Figure 8 shows a hydrograph separated into 3 components using DOC. The runoff was caused by a heavy thunderstorm (event F, Table I) on intermediate soil moisture conditions. For this type of event, the groundwater contribution ("baseflow") was assumed to be constant. DOC originated exclusively from the plateau region of our catchment because all other DOC-rich soils had no connection to the drainage system. This conclusion was verified by the comparison of the net DOC export at the catchment outlet (approx. 415 kg for the entire event) and at the gauge draining the headwater SH (408 kg). An area of only 22% of the catchment (headwater SH) delivered approx. 98% of the total DOC load and more than 50% of the water.

The average DOC concentration of the water from SH was 52 mg l^-1. The background concentration at the catchment outlet was 1.2 mg l^-1. Using equation (2) the instantaneous fraction of water originating from the headwater SH was calculated for each time step of 1h at the catchment outlet. The shape of the separated hydrograph shows both, translation and diffusion during the flow process. The time lag between the peak discharge at SH and the DOC peak at the main gauge was approx. 3h which fits with flow velocity measurements in the main channel. The percentage runoff for the headwater SH was 12%. The water added in subcatchment DUE was low in DOC concentration. It originated mainly from tracks, impervious areas and the alluvial plain. The percentage runoff for this subcatchment was only 3.1% and it contributed mainly to the fast runoff components. It dominated the ascending limb of the hydrograph.

For events with wet antecedent soil moisture or high precipitation sums, this simple method can not be applied. The runoff contribution from the subcatchment DUE increased due to a better connectivity, especially of the forest soil on the hillslopes. Figure 9 shows a sample event from September 2000 (event K, Table I). The antecedent soil moisture condition was dry. Approx. 25 mm of rainfall was necessary to fill the soil water storage. Then runoff started and reached its maximum after another 65 mm of rainfall within 12 hours. The net DOC export at the catchment outlet added up to 3300 kg, whereas the headwater SH contributed only 1600 kg (48.5%). The remaining 1700 kg (51.5%) came from the hillslope catchment DUE, where forest soils as well as saturated areas on the mountain tops (see also Figure 2) became connected to the drainage system. The shape of the DOC concentration curve at the catchment outlet looked similar to the event J (Figure 8). It also reached a similar maximum of approx. 35 mg l^-1. This may be explained by the same process delivering the DOC (saturation excess flow with a high degree of soil water exchange causing a quite constant DOC concentration) but now active on a larger contributing area. The average DOC concentration of the water from SH was again 52 mg l^-1. The average DOC concentration of the water added in subcatchment DUE increased to 25 mg l^-1. In comparison to the first event, the percentage runoff of the headwater SH increased only by a factor of 1.6 to 19%, whereas the percentage runoff for the subcatchment DUE increased by a factor of 4 to 12.5%. The headwater SH and the subcatchment DUE became more similar from a hydrological (percentage runoff) and hydrochemical (DOC load) point of view.

In the observed headwaters, measurements of DOC concentrations in the stream can be used to detect subsoil and surface runoff. DOC concentrations and d ¹⁸O showed a strong correlation, though they represent different processes - DOC measurements track flow paths, d ¹⁸O detects flow sources. Nevertheless the correlation can be explained by a broad contributing area with a large reservoir of DOC-rich pre-event water and a good lateral mixing of pre-event and event water during the runoff process. This is consistent with the study of Brown et al. (1999). Hence combining the information derived from DOC and d ¹⁸O measurements supplemented each other and allowed a fitting of d¹⁸O values. However, using DOC fluxes for studies of runoff processes bears advantages besides low cost for the analysis. In opposition to d¹⁸O, DOC concentrations always show a distinct contrast between soil water and precipitation water. Determination of a representative mean pre-event value for both DOC and d¹⁸O is difficult, but the possible range of soil DOC concentrations is rather limited and fell between 68 to 80 mg 1^-1 in our study. Models for a better determination for pre-event soil DOC are needed, since it is not completely understood, whether soil water sampling represents the mobilisation processes of DOC under natural conditions (Bishop et al., 1994).

For events with low precipitation sums and events with high intensities on dry pre-conditions DOC originated mainly from the headwaters. This fact enabled us to track this water along the main channel. In addition, the headwaters were the areas with the highest percentage runoff. Therefore the knowledge of the runoff generation process in headwaters is sufficient to understand a catchment behaviour for this type of events. Increasing precipitation sums and/or wetter antecedent soil moisture conditions result in increasing connectivity of other contributing areas to the main channel (Sidle et al., 2000). The hydrologic and hydrochemic similarity of the headwaters and the subcatchment DUE increased due to a complex superposition of runoff generation processes (Peschke, 1998). Especially knowledge of the runoff generation on steep hillslopes (e.g. discussed in McDonnell, 1990) and the ground water reaction becomes more important to understand the catchments response on precipitation. Only the use of additional tracers for hydrograph separation can help to elucidate this complex interaction of processes. In September 2000 we measured a data set which may illustrate the direction of further investigations.

First hypothetical interpretations are given. Figure 10 shows the same event (G in Table I) as displayed in Figure 9. Using 4 hydrochemical tracers (calcium, aluminium, silica and DOC) we tried to separate the hydrograph into 5 components (method according to Hoeg et al., 2000). Sodium concentration was only used to check the results. Five reservoirs of water were distinguished using hydrochemical data listed in Table III.
 

Table III. Hydrochemical data of the different reservoirs [mg l-1], in ( ): values used for the 5-component separation, bold numbers indicate the main characteristics of a reservoir.
Location / reservoir (No)	Sodium (Na)	Silica (Si)	Calcium (Ca)	Aluminium (Al)	DOC	pH	Number of samples
Main gauge	0.5-1.5	2-2.5	3.0-6-0	0.0-1.0	0.5-40.0	5.5-6.9	21
Shallow groundwater (1)	0.75-1.4	2.1-2.8 (2.3)	1.5-4.5 (3.5)	0.0-0.25 (0.0)	0.0-1.0 (0.0)	4.5-5.5	10
Deep groundwater (2)	0.7-1.5	4.8-5.8 (5.25)	3-8.8 (6.0)	0.0 (0.0)	0.0 (0.0)	5.8-6.0	6
Soil water from hillslope (3)	0.3-1.3	0.1-0.6 (0.5)	0.5-3 (3.0)*	1.4-1.9 (1.9)	10.0-35.0 (30.0)	3.7-4.2	11
Soil water from saturation area (4)	0.3-0.9	1.3-1.9 (1.7)	3.0-12.0 (8.5)*	0.8-1.2 (1.2)	40.0-80.0 (70.0)	3.9-5.0	44
Rain water (5)	0-0.8	0.0 (0.0)	0-1.0 (1.0)	0.0 (0.0)	0.0 (0.0)	5-7	8
* these high calcium concentrations are caused by continuous liming of the forest soils and therefore trace mainly shallow subsurface flow

Figure 10 gives us a first guess concerning the main processes contributing to stormflow: (1) saturation excess flow, (2) shallow subsurface runoff and/or perched aquifers in the organic layer of the forest soils forming the steep hillslopes, (3) direct runoff from impervious areas and direct precipitation on the alluvial plain, (4) groundwater ridging and/or pressure propagation in the aquifer. The concentrations of DOC and the selected cations aluminium, silica and calcium are closely correlated with the different runoff contributing reservoirs. Especially aluminium seems to be a good tracer for water from acid forest soils (but of course is very sensitive to variations in pH, DOC and transport time). In addition silica content is closely correlated with contact time. Nevertheless all solutes showed a large variation within time. Therefore continuous sampling is necessary for a reliable application of the presented method. Taking additional tracers into consideration (such as d ¹⁸O or chloride) might help to prove or falsify our assumption of the existence of five contributing reservoirs. In particular, the processes delivering the deep groundwater component could not be sufficiently clarified.

This work was carried out as part of the Graduate Program of Ecological Water Resources Management, financed by the Deutsche Forschungsgemeinschaft (DFG) and the state of Baden-Württemberg.