HYDROLOGICAL PROCESSES, VOL. 8, 335-349 (1994)

A CASE STUDY IN CATCHMENT HYDROCHEMISTRY: CONFLICTING INTERPRETATIONS FROM HYDROLOGICAL

AND CHEMICAL OBSERVATIONS

ALAN JENKINS

Institute of Hydrology, Wallingford, UK

ROBERT C. FERRIER

Macaulay Land Use Research Institute, Aberdeen. UK

RON HARRIMAN

Freshwater Fisheries Laboratory, Pitlochry, UK

AND

YINKA 0. OGUNKOYA

Obafemi Awolowo University, Ile Ife, Nigeria

ABSTRACT

Soil water, stream water, groundwater and rain water were sampled through a storm event in a moorland catchment. Samples were analysed for major ions and deuterium. Chloride and deuterium are used as tracers to enable separation of the stream runoff hydrograph into three components: rain, soil and groundwater. The results indicate that rain water arrives in the stream quickly during the event and contributes a significant volume to the runoff peak. The chemical signal in the rain water is, however, significantly damped, apparently due to mixing with soil water held in the catch- ment before the event. This is further modified before reaching the stream, apparently through mixing with a deeper groundwater component. Interpretation of tracer, chemistry and hydrological data to present an integrated picture of catchment hydrochemical response is difficult due to problems in the chemical and conceptual definition of the flow components.

KEY WORDS Tracers Hydrograph components Groundwater Soil water Rain water

INTRODUCTION

Hydrograph separation using mass balance equations for water and chemical tracers to determine the contributions of pre-event (‘old’) and event (‘new’) water in storm runoff is now an accepted and widely used technique in hillslope hydrology. Examples may be drawn from studies covering a range of hydro- logical and geomorphological settings to illustrate the relative importance of pre-event water (e.g. Hooper and Shoemaker, 1986; Obradovic and Sklash, 1986; Pearce ef al., 1986) and event water in storm runoff (e.g. Bottomley et al., 1984; Bonnell er al., 1990; Maule and Stein, 1990). The quantification of these flow components has necessitated a reconsideration of streamflow generation mechanisms and many stu- dies have followed the logical step towards using the results to infer flow pathways and hillslope hydrolo- gical processes (Sklash and Farvolden, 1979; Pearce ef al., 1986; Rodhe, 1987), although Kennedy et al. (1986) reported significant variability in the isotope ratio of catchment soil water and advised caution in the interpretation of isotope data. The most important conclusion from this breadth of hydrological study

CCC 0885-6087/94/040335- 15 0 1994 by John Wiley & Sons, Ltd.

Received 13 November 1992 Accepted 10 April 1993

336 A. JENKINS ET AL.

is that runoff generation mechanisms are difficult to generalize from one catchment to another and even from storm to storm within the same catchment (Ogunkoya and Jenkins, 1991).

Refinements to the isotope hydrograph separation approach have incorporated at least three flow com- ponents (DeWalle et al., 1988; Swistock et al., 1989; Maule and Stein, 1990) and so, necessarily, have used more than one tracer to split the old water component into distinctive groundwater and soil water com- ponents. The most important assumptions in these techniques are: (i) the chemistry (solute and isotopic) of the waters from the various sources are distinguishable, that is, each source has its own identity; and (ii) the chemical identity of each component is maintained in transit from source ‘reservoir’ to stream channel and is only changed by mixing with other identified components either in the channel or during travel to the channel, that is, there is spatial and temporal uniformity in the identity of each end-member component. One problem with such an analysis lies in identifying all of the contributing end-members and characterizing them through some sampling scheme (De Walle et al., 1990; Genereux and Hemond, 1990).

A further problem with these essentially hydrological studies lies in the interpretation of the results with respect to observations of stream chemical behaviour. Many hydrochemical studies have observed baseflow water to be well buffered with high pH and alkalinity, and storm flows to exhibit low pH and low alkalinity (Neal et al., 1990; Jenkins et al., 1990; Christopherson ef al., 1990a). This has led to the identification of two or more flow components based on chemistry (Hooper et al., 1990) such that, for example, deep ground- water provides the well buffered water and soil water provides the acidic storm flow (Robson and Neal, 1990; Christophersen et al., 1990b).

Clearly, hydrological and chemical interpretations are incomplete in isolation and more information is required to identify water source areas and flowpaths rather than simply using highly correlated rain- fall-runoff relationships or rainfall and stream water chemistry with little regard to soil and groundwater measurement. In this paper detailed quantitative and qualitative observations of rainfall, stream water, soil water and groundwater through a storm event at a mountain catchment in Scotland are brought together to assess the possibility of developing an integrated and consistent interpretation of catchment hydrology and hydrochemistry. The study site has been the subject of a five year hydrochemical study and a substantial background knowledge of the chemical characteristics and hydrological response of the catchment exists (Jenkins, 1989; Jenkins et al., 1990; Wheater et al., 1990; Ferrier ef al., 1990). The existing database presents a unique opportunity to interpret tracer observations and hydrograph separation results within the context of the detailed hydrochemical observations.

STUDY AREA

The Allt a Mharcaidh catchment (9.98 km’) lies on the western edge of the Cairngorm Mountains of north- east Scotland. It consists of three distinct topographic units: a gently rising upland plateau which lies above 700 m: steep valley sides lying between 550 and 700 m; and a gently sloping valley floor, below 550 m. The catchment is underlain by biotite granite with thick deposits of boulder clay derived from local rock cover- ing the valley floor. Alpine podsols predominate on the plateau, peaty podsols on the valley sides and blan- ket peat on the valley floor. The peaty podsols have a weakly indurated horizon which promotes the occurrence of a perched water-table. Downslope drainage from the valley sides promotes saturated condi- tions in the peat soil on the valley floor. The peat is characterized by the presence of pipes and incised eroded channels to the level of the underlying mineral horizon. Vegetation consists of northern blanket bog communities on the valley floor, lichen-rich boreal heather moor on the valley sides and alpine azalea-lichen heath on the plateau. The average annual precipitation is between 1100 and 1500mm. The main catchment consists of two smaller headwater catchments G2 and G3 (2.96 km2).

Streamflow was monitored at 20 minute intervals at the outlet of the catchment (Gl) and the larger of the subcatchments (G3). Runoff chemistry was sampled hourly during the storm event at both sites. Rainfall was monitored using a network of five tipping bucket gauges (RI-R5 in Figure 1) and bulk samples for chemical analysis were collected at each site. In addition, discrete samples for each millimetre of rainfall were collected at R1 and hourly bulk samples were taken at R5. Six boreholes, arbitrarily numbered

CATCHMENT HYDROCHEMISTRY 337

Figure 1. Allt a Mharcaidh study site showing the location of the monitoring equipment. GI, 2 and 3 are stream monitoring stations. PH locates the peat soil lysimeters and PP the peaty podsol soil lysimeters; rain gauges are marked RI-R5. BH shows the location of

the piezometer network and the individual borehole numbers are expanded in the inset

BHI, BH2, BH22, BH23, BH24 and BH26, were sampled hourly during the early part of the storm and then less often towards the end. The boreholes form two profiles perpendicular to the stream (Figure I ) . Soil throughflow was sampled at 50cm (PH2) and 1 m depth (PH4) in the valley bottom peat and a depth corresponding to the base of the organic horizon (PPO) in the peaty podsols on the valley side using gutter lysimeters.

Samples were analysed for deuterium (D20), relative to standard mean ocean water (SMOW), by mass spectrometry. Concentrations of Na, K, Ca, Mg, A1 (total monomeric), NH4, C1, SO4, NO3 and total organic carbon (TOC) were determined on stream, borehole rain and soil lysimeter water samples using the methods described by Harriman et at. (1990).

338 A. JENKINS ET AL.

RESULTS

All the data presented here relate to a rainfall-generated high flow event in the Allt a Mharcaidh during the period 13-15 June 1989.

Hydrological observations For several days before the onset of rainfall (13 June) the stream was at baseflow levels: 0.1 and

0.04m3 s-l at G1 and G3, respectively. Despite these low flow conditions water continued to flow from the peat soil lysimeters at a very low but constant rate, indicating slow drainage from the peat (PH2)

20

10

- -

0 .

Figure 2. (a) Storm rainfall and runoff at GI (solid line) and G3 (broken line) and flows in lysimeters at (b) PH4 (deep peat), (c) PH2 (shallow peat) and (d) PPO peaty podsol

CATCHMENT HY DROCHEMISTRY 339

and podsolic soils on the catchment hillslopes (PH4). Rainfall began at 0600 hours on 13 June 1989 and the storm comprised three bursts of rainfall over the following 15 hours (Figure 2). Catchment average rainfall is used for comparison with the flow at G1 and the rainfall at a high altitude site (R3) was used for G3 as most of this subcatchment is at a higher altitude than the main catchment. Total rainfall input to the whole catchment was 26 mm with a peak intensity during the early part of the storm of 6mm h-' . At R3 the total input was 31.5 mm with a corresponding peak intensity of 9.5 mm hr-'. At G I , flow increased, after a lag of 1-2 h from the onset of rainfall, from a baseflow level of 0.1 m3 s-' to a peak of 0*62m3 s-', which receded until further rainfall caused a second flow peak of 0.63 m3 s-I. The total storm runoff was 3.4mm, giving a runoff coefficient of only 13% which reflects the dry antecedent conditions, and so much of the rainfall was stored within the catchment over the storm period (Jenkins et al., 1990). At G3, the peak flow reached 0.2 and 0.21 m3 s-l following the two main peaks in rainfall and the lag was shorter than at G1, being about 1 h. This may reflect the higher peak intensity at the high altitude site (R3), but G3 is also a smaller catch- ment than G1, dominated by steep slopes and this will tend to cause shorter lags between rainfall and run- off. Certainly the variations in flow at G3 more closely match the variations in rainfall input than at GI. The total storm runoff at G3 was 3.5mm, giving a runoff coefficient of only 10.9%.

The throughflow response of the soil layers at the pit locations is shown in Figure 2. The gaps in this record do not necessarily indicate zero flow from the lysimeter (except at PPO which responds very rapidly to rainfall), as this is mainly attributable to the failure of the tipping bucket mechanism at low flows, and indeed flow has been observed in the PH4 lysimeter throughout long summer dry spells. Flow in the upper layer of the peaty podsol responds immediately to rainfall and stops almost instantaneously with the rain- fall. Both peat layers, on the other hand, show a lag of up to 4 h between rainfall and throughflow and are characterized by long recessions after the rainfall has stopped. All of the lysimeters reach a peak flow at the same time as flow in the stream. These observations indicate that the soils approach saturation through the storm, leading to lateral movement of water downslope, or the lysimeters collect water transmitted rapidly along preferential flowpaths. The response is entirely consistent with other observations at the site (Jenkins et al., 1990; Wheater et al., 1990) although the origin of the throughflow water can only be indicated by study of the solute load.

Depths to the water-table through the early part of the storm, encompassing the first flow peak, are

+

-1

h

E v

-2 8 5 + + u)

E 2 5 P -3 c

-4 0200 2200 1800 1400

Time (hours GMT)

Figure 3. Depths to the water-table in the piezometers through the storm event

340 A. JENKINS ET AL

shown in Figure 3. All boreholes show a decrease in depth to the water-table (that is, a rise in the water- table surface) in response to the rainfall input. The changes are non-uniform between each borehole and the rise in level varies between 36cm at B26 (nearest the stream) and 8cm at B22 (furthest upslope from the stream). The borehole levels indicate that within the valley bottom near to the stream channel there was a general movement of the water-table towards the surface. The timing of water-table rise relative to the increase in flow also differs widely and in some instances the water-table continues to rise until the end of sampling. Changes in the local topography are probably responsible for this varied response. At all bore- holes a marked decrease in the water-table depth is observed 48 h after the storm.

Tracer observations Rainfall D20 concentrations are initially enriched and become progressively more depleted as the storm

progresses (Figure 4), but are always more enriched than stream water. Stream D20 concentrations at GI and G3 show a consistent response to the rainfall, becoming more enriched as the flow increases and moving back to pre-storm levels as the flow recedes. This response implies that event water is reaching the stream and contributing to the flow peak.

Both the peaty podsol and the upper peat layer show an initial depletion in D20 concentration and then a gradual enrichment. This behaviour ties in closely with the rainfall D20 signal. At the deep peat lysimeter, on the other hand, D20 concentrations are initially more enriched than the stream water, but quickly assume a similar concentration. No discernible trend is observed in the concentration of D20 at any borehole and they all remain relatively constant at about -60%0. Given that the water level in all of the boreholes is increasing, this implies that the rise in the water-table is due to an influx of similar water and so tends to support the general hypotheses of the development of a groundwater ridge close to the stream generated by a pressure mechanism. The D20 data taken as a whole indicate that very rapid runoff of rain water produces the initial change in stream D20 and further contributions of more enriched soil water maintain the stream D20 response through the peak flow. The response only tails off after the peak flow has passed when soil and rain water contributions recede and groundwater again dominates the stream flow.

Chloride concentrations during the event are shown in Figure 5 . Initially, the rainfall concentration is low and increases through the storm, although levels are always lower than the stream concentration a t GI. Initial peaks of C1 in the stream at GI and in all three soil lysimeters probably represent wash-off of C1 accumulated over the dry period preceding the event. At G3, C1 is generally lower than G1 and rain- fall concentrations exceed those observed in the stream only in the latter part of the storm. There is no observed dilution effect in the stream to support the implication from the D 2 0 data that rain water con- tributes significantly to the stream flow. On the other hand, in the PH2 and PH4 lysimeters C1 concentra- tions are consistently higher than the stream, implicating a source of C1 on or near the surface. Only at the peaty podsol lysimeter (PPO) do C1 concentrations match the observed variation in rainfall concentration. This raises the possibility that C1 is not conservative on the short time-scale and that short-term retention occurs in the soils (Harriman et al., 1990).

Borehole concentrations of C1 are considerably higher than in the stream. Each borehole shows a differ- ent response with respect to each other. In the early part of the storm BH24, BH23 and BH26 decrease in C1 concentration whereas BH2 shows no pattern. The day after the storm, however, all of the boreholes return to a similar C1 concentration which is close to that of stream baseflow. To explain this pattern, a source of water low in C1 must contribute to the groundwater component during the storm. The only water observed with low C1 was the rainfall and it is difficult to hypothesize a mechanism for moving rain water quickly to this groundwater zone. In any case, the D20 signal in the borehole does not imply such a process. Possibly further stores of water exist within the catchment in the form of discrete pools of groundwater with a finite capacity which, when exceeded, mix together.

Hydrochemical observations Before the start of rainfall, flow at GI is assumed to be dominated by the groundwater contribution as

there was no significant rainfall at the site in the preceding 10 days and the soil lysimeters were dry, except

CATCHMENT HYDROCHEMISTRY 34 1

for a very small flow from the base of the peat (PH4) indicating some slow drainage of the peat soils. This 'background' soil water at PH4 is very acid (pH 3.8-4.2), rich in total monomeric A1 (426 pequiv. l-'), but relatively low in Ca (25 pequiv. 1-I) and SO4 (37 pequiv. 1-I) (Figure 6). Stream baseflow chemistry is characterized by pH values greater than 6.0, A1 near zero and Ca and SO, around the long-term mean concentration for the stream at 51 and 50pequiv. 1-', respectively (Figure 6.).

Very soon after the rainfall begins, flow occurs in the peaty podsol lysimeter, indicating that the upper

0 Rain

-20 1 0

-40 2 - 3 0 ] .- 0 0 0

-50

G3 -60

00 PPO

-36 0 . PH4 '.

I 0200 1 ioo 2100 0800

Figure 4. Deuterium concentrations in (a) rain and stream water, (b) soil lysimeters and (c) piezometers. Note the longer time-scale for the piezometers

342 A. JENKINS ET A L

200 -

160 - r '- 5 s 120 - D a, 1 - a,

.-

./

and intermediate valley slopes are transmitting rain water rapidly downslope. This water has a chemistry very similar to rain water (Figure 6). Flow in the peat lysimeters increases equally rapidly as water draining the podsol soils flows along the interface of the peat and underlying mineral horizon (PH4). At PH2 a build-up of dry deposits at the surface and recently mineralized S in the upper peat layers produces the initially high SO4 concentration. This rapid runoff of S04-enriched water causes the initial peak SO4

~

Rain 20

- b ' 4 0 -

* I: PPO

0: 1200 2200 0800 0200

BH23 rn

BH24 ' BH2 o

BH26 0

I _ , , r- ' I . . r- 1200 1200 1200

Time (hours, GMT)

Figure 5. Chloride concentrations in (a) rain and stream water, (b) soil Iysimeters and (c) piezometers. Note the longer time-scale for the piezometers

CATCHMENT HYDROCHEMISTRY

-- L- loo., 80

0

2o t m 0

343

Figure 6. (a) Sulphate and (b) Ca concentrations in the stream at G1 (m), PH2 lysimeter (O), PH4 lysimeter (0). rain water (0) and the piezometers (range of concentrations shown as bars)

concentration in the stream at G1 as the rainfall SO4 concentration is low. Thereafter, the SO4 concentra- tion at PH2 remains relatively constant and at PH4 remains markedly lower than stream concentrations.

The stream water chemistry through the event responds in close association with the change in flow. This presumably reflects both changes in the chemistry of water following different flowpaths and the changing relative contributions from different flow components, that is, soil, rain and groundwater. Both SO4 (an atmospheric-derived ion) and Ca (dominantly a catchment-derived ion) concentrations increase with flow. Soil water SO4 concentrations, however, are well below those in the stream and show only a very damped response to changes in rain water input. As rainfall concentrations of Ca and SO4 are low, at least initially in the case of SO4, this increase in concentration points to the influx of a source of water not sampled. Although borehole water chemistry spans a wide range of Ca and SO4 concentrations it is not of sufficiently high concentration to account for the increase in the stream if mixing with only soil water and rain water is assumed. In any case, it is unlikely that the groundwater SO4 concentration can be greater than the long-term mean for the stream and so this indicates that the boreholes are affected by influxes of stream, soil or rain water during the event.

The observed hydrochemical response can be interpreted to represent a picture consistent with long-term observations at the site (Jenkins et al., 1990). The initial increase in stream SO4 is driven by wash-off of dry

344 A. JENKINS E r AL.

160 -

140 - ,.. 'L

.f 120 -

J 100 - CT m -

0)

0 80 - 0 .c

60 -

40 -

and mineralized deposits and is reflected in the initial peak in the PH2 lysimeter. This high SO4 water, originating as rain water, moves rapidly into the stream and at the same time Ca is mobilized from soil exchange sites. During the second flow peak, the rainfall contribution to the flow is sufficient to enable the higher rainfall SO4 concentrations to produce the second observed SO4 peak in the stream. Soil water in the catchment before the event, and now augmented by rain water, tends towards an quilibrium with the soils through ion-exchange processes and slow drainage of this soil water after the storm peaks allows Ca and SO4, and consequently alkalinity and hydrogen, to recover only slowly during the recession limb. Throughout the storm period the groundwater contribution would then need to be assumed to remain essentially constant.

HYDROGRAPH ANALYSIS

End-member mixing The assumption that stream water represents a mix of stored water which resided in the catchment before

the storm and incoming rain water is intuitively correct. The pre-storm stored water may be further assumed to consist of soil and groundwater components. As the storm progresses, if the end-members

1

m PH4 Borehole 150

130 - - - - > 110 - .- 3 0- 2 90- v

0) z ' O : i U 50 -

t PH2

-70 -50 -30

PH4 .v

-70 -60 -50 -40

Deuterium ("i")

0

Figure 7. End-member mixing diagrams for stream water samples at (a) G1 and (b) G3 through the storm event

CATCHMENT HYDROCHEMISTRY 345

have been correctly identified and the tracers are truly conservative, all of the stream samples should be bounded within the limits of the three end-members (Figure 7). A number of observations complicate the choice of end-members: (i) each of the soil profiles sampled has different concentrations of D 2 0 and C1; (ii) borehole D 2 0 concentrations are different to those in the stream before the event; (iii) borehole C 1 concentrations differ markedly; (iv) borehole C 1 concentrations change through the storm period; and (v) rainfall concentrations of both C1 and D20 change through the storm. To simplify this situation volume-weighted mean rainfall D20 and C1 concentrations were calculated, pre-storm (12 June) soil water from lysimeter PH2 is used for the soil end-member, and the pre-storm (12 June) stream water chemistry is taken to represent the groundwater end-member. Figure 7 shows that most of the stream samples fall within the end-member boundaries defined under this scheme. Figure 6 also indicates the degree to which this analysis is changed by using other definitions for soil and groundwater end-members, e.g. using soil lysi- meter PH4 to describe the soil end-member. The borehole chemistry cannot be used as the groundwater end-member as C1 concentrations are too high and a further source of water, low in C1, would be needed to produce stream C1 concentrations. Furthermore, the D20 concentration is constant between boreholes at a value which is 5 %O lower than that observed in the stream at pre-storm baseflow level. This indicates that the groundwater sampled does not contribute to streamflow, even at baseflow, that the borehole sam- ples do not represent true groundwater chemistry, or that other groundwater contributions of different chemistry contribute to the flow.

Hydrograph separation The overall pattern of tracer concentrations in the rainfall, stream, soils and boreholes (Figures 4 and 5)

indicate that event water is contributing to the hydrograph. To assess the relative contributions of water from these different storages within the catchment a hydrograph separation technique has been used. In an earlier two-component hydrograph separation analysis, event water was calculated to contribute 46% of the total storm runoff with an instantaneous event water contribution of about 60% during the second flow peak, but that soil water, or at least water of a different isotopic identity to rain water or groundwater, contributed significantly to the storm flow (Ogunkoya and Jenkins, 199 1). A three-component separation, therefore, seems more appropriate in this instance. The assumptions underlying the use of the three- component model are: (i) storm runoff has three end-members, in this instance we assume these to be incident precipitation, soil water and ground water; (ii) each end-member has a distinctive chemical and isotopic identity in terms of the tracers used and this is certainly the case for the three components assumed here (Figure 6); (iii) temporal variations that occur in the concentration of the end-members through the storm are known and these are mainly caused by Rayleigh distillation or rain-out effects in the case of precipitation and by the mixing of one end-member into the reservoir of another, such as rain water mixing with soil water; and (iv) streamflow at the inception of storm runoff, i.e. baseflow, is a mixture of ground- water and other compartments not identified or sampled.

The three-component hydrograph separation model uses mass balance equations for two conservative tracers. Assuming that a, b and c are the proportions of incident precipitation, soil water and ground- water, respectively, x and y are the concentrations of conservative tracers in stream water then

x = aAl + bA2 = cA3 y = aB1+ bB2 + cB3

where A l , A2, A3, B1, B2 and B3 are the corresponding tracer concentrations in the incident precipitation, soil water and groundwater, respectively. Rearranging these equations, the proportions of the three components flow are

-(x - A3)(B2 - B3) + ( y - B3)(A2 - A3) (A2 - A3)(B1 - B3) - (B2 - B3)(A1 - A 3 )

a =

(X - A3)(Bl - B3) - ( y - B3)(A1 - A3) (A2 - A3)(B1 - B3) - (B2 - B3)(A1 - A3)

b =

346 A. JENKINS ET A L .

Any variation in tracer concentrations through the event are assumed to be ascribed to mixing processes influenced by flowpaths in or on the hillslope. On the other hand, a volume weighted mean concentration for rain water implies a temporally random variation of tracer concentration around the weighted mean. The temporal variability may not be random, however, given the Rayleigh distillation and rain-out effects (Ingrham and Taylor, 1986; McDonnell et al., 1990). Furthermore, a weighted mean allows tracer concen- trations of the last raindrops in the storm event to influence hydrograph separation at the beginning of the storm runoff response. This leads to an underestimation of the groundwater and soil water components if the rainfall concentrations are in the direction of groundwater and soil water concentrations. For this rea- son, volume-weighted rainfall inputs were incrementally adjusted for each time period (McDonnell et d., 1990). Hence, although the time synchronization of mixing on the catchment and entry into the channel remains unknown, rain falling later in the storm cannot influence the analysis for an antecedent period.

The groundwater is necessarily represented by stream water chemistry before the onset of the storm and not the borehole chemistry. As discussed earlier the water from the boreholes was isotopically enriched compared with the pre-storm stream water, whereas C l concentrations in the stream are considerably lower than in the boreholes. Using a mean, or individual, borehole composition will clearly not produce sensible separation as in this instance most of the stream samples during the storm plot outside the bound- aries of the mixing diagram (Figure 7). Using baseflow stream water to represent the groundwater end- member assumes that soil water (as defined at the lysimeter) does not contribute to baseflow. Both PH2 and PH4 are used to represent the pre-storm soil water chemistry (Table I).

The results of the hydrograph separation at G1 and G3, are shown in Figure 8 and the calculated contributions of each end-member to total streamflow are given in Table I. At both sites the contributions of groundwater are similar using both PH4 and PH2 to represent the soil water. Calculated groundwater contributions are higher for G3 than G1. The choice of soil water, however, produces a trade-off between rain water and soil water contributions. If PH4 is used, the soil water contribution (23%) is less than the rain water contribution (36%) at G1. If PH2 is used, the soil water contribution (37%) dominates over the rain water contribution (18%). At G3, the rain water contribution is always greater than the soil water contribution. Instantaneous contributions of rain water and soil water (Table 11) are highest around the first flow peak.

These results are in close agreement with the two-component hydrograph separation detailed by Ogunkoya and Jenkins (1991) and it is clear from the mixing diagram (Figure 7) that only two components are identifiable from the stream water chemistry as C1 is rather constant. In this respect, it could be argued that only a two-component separation is justified. On the other hand, the distinction between the three components assumed here, in terms of C1 and D20 concentration, is clear. Nevertheless, the soil water component is difficult to define in the field as concentrations vary between lysimeters and so a two- component split is perhaps the optimum interpretation these data can support.

It is interesting, however, to speculate on the dynamics of the soil water component through the event. With respect to the relative contributions to the streamflow peak from catchment stored and new water, the

Table I. End-member C1 (pequiv. I- ') and deuterium (%) concentrations

Deuterium Chloride

Stream (pre-event baseflow) G1 G3

BH26 BH24 BH23 BH2

PH2 PH4

Boreholes (pre-event)

Soil lysimeters (pre-event)

-64 - 59

- 60 -58 -58 - 60

-44 - 50

91 76

103 190 143 129

1 I0 158

CATCHMENT HYDROCHEMISTRY

220 -

180 - h

'm 140- m E

ii

v

g 100-

60 -

20 -

347

600

500

,"' 400 E 6 300 ii

200

100

- 7

v

I I " " I " " I

0200 1200 2200 0800 Time (hours GMT)

Figure 8. Three-component storm hydrograph separation at (a) GI and (b) G3. Rain water contribution is represented by the dotted line, rain plus soil water contribution by the broken line and total streamflow (rain plus soil plus groundwater) by the solid line

question hinges on which category the soil water is assigned to. The D20 signal in the soil water before the rainfall event implies that both PH2 and PH4 were distinct sources of water from both each other and from the stream (therefore, implicitly, the groundwater). As the storm progresses, however, D20 in the PH2 lysimeter decreases rapidly before increasing again, apparently in response to the changing rain water concentrations. In this respect, the soil water should be regarded as new or event water and the peak flow is generated by quickflow mechanisms transmitting rain water to the channel. PH4, on the other hand has a dissimilar trend and D20 concentrations later in the storm are very similar to stream D 2 0 concentrations.

The results indicate that pre-event groundwater makes a significant contribution to total runoff by dominating the recession limb. Peak flows are dominated by rain and soil water contributions which are transmitted rapidly along preferential pathways, notably along the interface between the peat and mineral soil layers and through the upper organic layers of the peaty podsol. These quickflow processes were previously identified by Jenkins et al. (1990) along with the identification of a well developed rill and pipe system draining variable source areas within the soil profile (Wheater et al., 1990). In this instance, however, soil water should be regarded as pre-event as the end-member used in the analysis was deter- mined by sampling before the event.

348 A. JENKINS ET A L .

Table 11. Calculated total storm runoff and peak instantaneous contributions

Site Soil water Total runoff contribution (%) Peak instantaneous contribution (%) end-member

Rain Soil water Groundwater Rain Soil water ~~ ~ ~ -~~

G1 PH2 17 37 46 GI PH4 36 23 41 G3 PH2 24 20 56 G3 PH4 29 13 58

24 48 40 54

51 32 37 24

CONCLUSIONS

The data from Allt a Mharcaidh shows a complexity of pattern that goes against the interpretation of hydrograph response which calls for large contributions of groundwater to storm flows. It must be recog- nized, however, that this analysis is based on detailed hydrochemical data from a single storm event and very different hydrograph responses have been observed from storm to storm in this catchment as a func- tion of antecedent and storm characteristics (Jenkins, 1989; Jenkins et al., 1990; Ogunkoya and Jenkins, 199 1). The interpretation of a relatively simple hydrological pattern is complicated by chemical data which imply sources of water that have been neither sampled nor hypothesized. The end-members vary through time (chemically) through reaction with the soil solid phase and without more detailed information it is unclear if soil and groundwater from particular areas of the catchment contribute to streamflow genera- tion. Only at a very broad scale do the chemical and tracer results fit with a simple hydrological picture. To extrapolate this concept to a detailed quantitative scale is dubious. In this respect the distinction between ‘old’ and ‘new’, quick and slow, groundwater and soil water, must be clear; these terms are not interchangeable as they each represent different conceptual components and will be characterized by different chemistry and isotope ratios. Consequently, the present result underpins the complex nature of catchment hydrology and runoff generation and the need for a more thorough investigation of end- members and pathways. Without such a development it remains questionable if reliable process based short time-scale predictive models of hydrochemistry for upland catchments are achievable.

REFERENCES

Bonnell, M., Pearce, A. J . , and Stewart, M. K. 1990. ‘The identification of runoff production mechanisms using environmental isotopes in a tussock grassland catchment, eastern Otago, New Zealand’ Hydrol Process., 4, 15-34.

Bottomley, D. J., Craig, D., and Johnston, L. M. 1984. ‘Neutralization of acid runoff by groundwater discharge to streams in Canadian Precambrian shield watersheds’, J . Hydrol., 75, 1-26.

Christophersen, N., Hauhs, M., Mulder, J., Seip, H-M., and Vogt, R. D. 1990a. ‘Hydrogeochemical processes in the Birkenes catch- ment’ in Mason, B. J. (Ed.) The Surfuce Wa/ers Acidifica/ion Programme. Cambridge University Press, Cambridge, 97- 107.

Christophersen, N., Neal, C., and Hooper, R. 1990b. ‘Modelling streamwater chemistry as a mixture of soil water endmembers, a step towards second generation acidification models’, J . Hydrol. 116, 307-321,

DeWalle, D. R., Swistock, B. R., and Sharpe, W. E. 1988. ‘Three component tracer model for stormflow on a small Appalachian forested catchment’, J . Hydrol., 104, 301-310.

DeWalle, D. R., Swistock, B. R., and Sharpe, W. E. 1990. ‘Tracer model for stormflow on a small Appalachian forested catchment - reply’, J . Hvdro/., 117, 38 1-384.

Ferrier, R. C., Miller, J. D., Walker, T. A. B., and Anderson, H. A. 1990. ‘Hydrochemical changes associated with vegetation and soil’ in Mason, B. J. (Ed.) The Surface Waters Acidificafion Programme. Cambridge University Press, Cambridge, 57-69.

Genereux, D. P. and Hemond, H. F. 1990. ‘Three component tracer model for stormflow on a small Appalachian forested catchment - comment’, J . Hydrol., 117, 377--381.

Harriman, R. Gillespie, E., King, D., Watt, A. W., Christie, A. E. G., Cowan, A. A., and Edwards, T. 1990. ‘Short term ionic re- sponses as indicators of hydrochemical processes in the Allt a Mharcaidh catchment, western Cairngorms, Scotland’, J . Hydro/. ,

Hooper, R. P. and Shoemaker, C. A. 1986. ‘A comparison of chemical and isotopic hydrograph separation’, Wu/. Resow. Reg., 22,

Hooper, R. P., Christophersen, N., and Peters, J. 1990. ‘Endmember mixing analysis (EMMA): an analytical framework for the

Ingrham, N. L. and Taylor, B. E. 1986. ‘Hydrogen isotope study of large scale meteoric water transport in northern California and

Jenkins, A. 1989. ‘Storm period hydrochemical response in an unforested Scottish catchment’, Hydro/. Sci. J . , 34, 393-404.

116,267-285.

1444- 1454.

interpretation of streamwater chemistry’, J . Hydro/. , 116, 321-345.

Nevada’, J . Hydrol., 85, 183-197.

CATCHMENT HYDROCHEMISTRY 349

Jenkins, A,, Harriman, R., and Tuck, S. J. 1990. ‘Integrated hydrochemical responses on the catchment scale’ in Mason, B. J. (Ed.) The

Kennedy, V. C., Kendall, C., Zellweger, G. W., Wyerman, T. A., and Avanzino, R. J. 1986. ‘Determination of the components of

Made, C. P. and Stein, J. 1990. ‘Hydrologic flow path definition and partitioning of spring meltwater’, Wat. Resour. Res. 26, 2959-

McDonnell, J. J., Bonnell, M., Stewart, M. K., and Pearce, A. J. 1990. ‘Deuterium variations in storm rainfall: implications for stream

Neal, C., Robson, A. J., and Smith, C. J. 1990. ‘Acid neutralization capacity variations for the Hafren forest stream, mid-Wales:

Obradovic. M. M. and Sklash, M. G. 1986. ‘An isotopic and geochemical study of snowmelt runoff in a small arctic watershed’,

Ogunkoya, 0. 0. and Jenkins, A. 1991, ‘Analysis of runoff pathways and flow contributions using deuterium and stream chemistry’,

Pearce, A. J., Stewart, M. K., and Sklash, M. G. 1986. ‘Storm runoff generation in humid headwater catchments 1. Where does the

Robson, A. J. and Neal, C. 1990. ‘Hydrograph separation using chemical techniques: an application to catchments in mid-Wales’, J.

Rodhe, A. 1987. ‘The origin of streamwater traced by oxygen-l8’, Uppsala Univ. Depf. Phys. Geogr., Div. Hydro/., Report Series A 41,

Sklash, M. G. and Farvolden, R. N. 1979. ‘The role of groundwater in storm runoff, J. Hydro/., 43, 45-65. Swistock, B. R., DeWalle, D. R., and Sharpe, W. E. 1989. ‘Sources of acidic stormflow in an Appalachian headwater stream’, Waf.

Resour. Res. 25, 2139-2147. Wheater, H. S., Langan, S. J., Miller, J. D., Ferrier, R. F., Jenkins, A., Tuck, S. J., and Beck, M. B. 1990. ‘Hydrological processes on

the plot and hillslope scale’ in Mason, B. J. (Ed.) The Surface Waters Acidification Programme. Cambridge University Press, Cambridge, 121-137.

Surface Waters Acidification Programme. Cambridge University Press, Cambridge, 47-55.

stormflow using water chemistry and environmental isotopes, Mattole River Basin, California’, J. Hydro/. 84, 107-140.

2971.

hydrograph separation’, Waf. Resour. Res. 26, 455-458.

inferences for hydrological processes’, J. Hydro/. 121, 85-101.

Hydro/. Process. 1, 15-30.

Hydrol. Process., 5, 271-283.

water come from?’, Wat. Resour. Res. 22, 1273-1282.

Hydrol. 116, 345-365.

260 pp.