Introduction

There has been enhanced interest in streamflow-generation processes in permafrost regions in the last decade, spurred largely from observations of changing streamflow hydrochemistry, increased low flows and changes to the recession limb of the hydrograph (Keller et al. 2010; Lyon et al. 2009; Lyon and Destouni 2010; Frey and McClelland 2009; Smith et al. 2007; Schuster et al. 2011; St. Jacques and Sauchyn 2009; Striegl et al. 2005; Walvoord and Striegl 2007). Virtually all of these observations are ascribed to thawing permafrost, which has been documented across the pan-arctic (Hinzman et al. 2005; Jorgenson et al. 2006; Quinton et al. 2011) and is projected to continue in response to warming surface temperatures.

Streamflow generation processes in permafrost environments are distinguished from more temperate regions in that: (1) snowmelt is typically the major hydrological event, although late-summer and fall storms are becoming increasingly important (Kane et al. 2003; Spence et al. 2011), (2) deep drainage is restricted where permafrost occurs, supporting near-surface water tables, and (3) surface organic soils and preferential flow pathways allow a rapid means of this supra-permafrost water to reach the stream. When permafrost thaws, it is hypothesized that: (1) deeper groundwater contributions are able to connect to the drainage network, as sub- and intra-permafrost pathways become increasingly important compared with supra-permafrost groundwater, and (2) increasing active layer thickness will result in deeper supra-permafrost pathways and slower subsurface drainage to the stream network (see Woo 2012 for review). In both cases, changes in streamflow geochemistry and runoff timing and magnitude are expected to occur. The hydrological response will be a transition from flashier responses to a slower, more gradual response dominated by baseflow and the occurrence of over-winter flows where previously absent (MacLean et al. 1999; Petrone et al. 2006; Slaughter and Kane 1979; St. Jacques and Sauchyn 2009). Hydrochemically, as catchments transition to less permafrost, dissolved organic carbon (DOC) is expected to decline and solute concentrations increase in response to deeper flow pathways and enhanced contact time with mineral exchange sites (Frey and McClelland 2009; MacLean et al. 1999; Petrone et al. 2006; Smith et al. 2007; Striegl et al. 2005).

There is good support of this conceptual model of change from process-based research in headwater catchments (e.g. Carey and Woo 2001; Hinzman et al. 1993; McNamara et al. 1997; O’Donnell et al. 2010; Slaughter and Kane 1979). However, the majority of reporting on permafrost thaw and its impact on groundwater contributions to streamflow have been observed on large river systems such as the Yukon in Canada (Dornblaser and Striegl 2007, 2009; Schuster et al. 2011; Striegl et al. 2005; Walvoord and Striegl 2007), and large Russian rivers such as the Ob’, Yenisey and Lena rivers (Frey et al. 2007; Smith et al. 2007), using synoptic sampling and data obtained from national hydrometric monitoring networks. Results have identified quantifiable changes in inferred groundwater contributions that range from ∼1 %/year for the Yukon river for the last 30 years (Walvoord and Striegl 2007), to winter baseflow increases of 0.5–272 % in large rivers across the Northwest Territories, Canada (St. Jacques and Sauchyn 2009). There has, however, been less work understanding flow processes in headwater catchments, where much of our conceptual understanding is obtained, and how these are responding to change. One exception is Jones and Rinehart (2010) who have synthesized data from Caribou-Poker Creeks Research Watershed in Alaska over the past 30 years, and identified complex responses among watersheds with different percentages of permafrost cover. While watersheds with less permafrost had greater baseflow, inter-annual variability in precipitation, pre-freezeback wetness, snowpack total and other factors all complicated understanding the hydrological response to thawing permafrost. In addition, alternate hypotheses have largely been absent from the debate of thawing permafrost on streamflow, although Spence et al. (2011) have presented results from an investigation in the Canadian Shield since the mid 1990s which suggests that changes in streamflow regimes can be explained by changing the phase and timing of autumn precipitation.

Evidence for changing climate in permafrost regions is unequivocal, particularly in northwestern North America where increases in temperature of several degrees Celsius are observed in instrumental records (ACIA 2004). However, inter-annual variability is large in terms of both temperature and precipitation, and the influence of this variability on streamflow response is considerable (Jones and Rinehart 2010). Furthermore, short-term climate variability undoubtedly influences streamflow hydrochemical response inter-annually. Information on change from large rivers in response to thawing permafrost is compelling, yet headwater catchments continue to provide essential information on process interaction, seasonal memory effects, and the nature of short-term variability in runoff and hydrochemical response. The latter is important considering the use of chemical indicators to assess change.

Considering the lack of multi-year headwater catchment investigations in permafrost regions, the objective of this paper is to use 4 years of hydrometric, stable isotope and hydrochemical data to identify the relative influence and variability of near-surface supra-permafrost groundwater versus deeper intra and sub-permafrost flow pathways during freshet. Stable isotope hydrograph-separation techniques (e.g. Obradovic and Sklash 1986) will be linked with dissolved ion hydrochemistry, DOC and hydrometric conditions to assess inter-annual variability in flow pathways and the sensitivity of these to climate during melt and antecedent conditions. By using 4 years of data, the relative influence of intra-annual variability on response can be identified in a period presumably short enough to limit considerable permafrost thaw.

Study site

Granger Basin (60°31′N, 135°18′W) is a 7.6 km2 headwater catchment located within the Wolf Creek Research Basin (WCRB), Yukon, Canada. WCRB and Granger Basin are long-term watershed research facilities (Fig. 1). Granger Basin ranges in elevation from 1,310 to 2,250 m above sea level, and has a continental subarctic climate with mean annual precipitation of ∼400 mm at higher elevation, of which ∼40 % falls as snow. Mean annual temperature is −3 °C, with July and January temperatures of +14.1 and −17.7 °C respectively.

Fig. 1
figure 1

Location of (a) Yukon Territory and Wolf Creek Research Basin in Canada, and (b) Granger Basin

Full size image

The geology of the basin is primarily sedimentary, comprising limestone, sandstone, siltstone and conglomerate and is overlain by a mantle of glacial till. Approximately 70 % of the basin is underlain with permafrost (Lewkowicz and Ednie 2004), with low-elevation southerly exposures permafrost-free. Granger Basin is located above treeline in the alpine taiga ecozone with low-lying grasses, herbs and shrubs vegetation such as dwarf birch and willow. At higher elevation, alpine tundra predominates. For a more detailed description of the basin, see McCartney et al. (2006).

Methods

Field data

Data were collected and analyzed for four snowmelt seasons: 2002, 2003, 2006, and 2008. Discharge was calculated using a stage-discharge relationship during open water and salt dilution when the channel had significant ice and backwater effects. Streamflow samples were collected typically on a daily basis during freshet, and less frequently post-freshet. Samples were filtered, and sealed in 60-ml vials with airtight caps and kept cool and dark until analysed. Samples for DOC were filtered through pre-combusted Whatman GF/F glass fibre filters and acidified with H2SO4 to 0.035 M prior to storage in sterilized 50-ml vials. Snowmelt lysimeters (0.25 m2) were installed at the base of the snowpack to collect meltwater, which was also sealed in 60-ml vials. Snow depth and snow-water equivalent (SWE) were measured along a transect beginning early April each year at 60 points, and snow density at 12 points using a Mount Rose snow sampler. The transect was measured approximately every 3 days until snowmelt was complete. Several weather stations are located within Granger Basin collecting air temperature, humidity, short-wave and net radiation, and windspeed data at half-hour intervals. For further information on instrumentation and methods, see Carey and Quinton (2004, 2005) and McCartney et al. (2006).

Laboratory analysis

For 2002 and 2003, laboratory analyses for δ18O were performed at the National Water Research Institute, Saskatoon, Saskatchewan by CO2 equilibrium. Accuracy of the analysis based on replicate samples was ± 0.05 ‰. For 2006 and 2008, water samples were analyzed for δ 18O at the G.G. Hatch Stable Isotope Laboratory at the University of Ottawa using a Finnigan MAT Delta plus XP + GasBench isotope ratio mass spectrometer. The analytical precision was ± 0.10 ‰.

All water samples were analyzed for major cations and anions using a DIONEX LC25 chromatography oven and CD20 conductivity detector in the Water Chemistry Laboratory at the University of Toronto. Anions evaluated are chloride (Cl−), bromide (Br−), nitrate (NO3 −), phosphate (PO4 3−) and sulfate (SO4 2−). The cations evaluated include sodium (Na+), ammonium (NH4 2+), potassium (K+), magnesium (Mg2+) and calcium (Ca2+). The analytical detection limit for both anions and cations was ±0.04 ppm. DOC samples were run on a Technicon Autoanlyzer via automated persulphate-UV digestion with a phenolphthalein colour reagent.

Hydrograph separation

Two-component hydrograph separations have been widely used to establish the sources of streamwater during snowmelt and in permafrost environments. The method is based on the steady-state mass-balance equations for water and a conservative tracer (Sklash and Farvolden 1979):

$$ {Q_{\mathrm{T}}}={Q_{\mathrm{p}}}+{Q_{\mathrm{e}}} $$
(1)
$$ c_{\text{T}}^{{{{\text{t}}_{\text{i}}}}}{{Q}_{\text{T}}}=c_{\text{p}}^{{{{\text{t}}_{\text{i}}}}}{{Q}_{\text{p}}}+c_{\text{e}}^{{{{\text{t}}_{\text{i}}}}}{{Q}_{\text{e}}} $$
(2)

where Q T is the total runoff, Q p and Q e are the pre-event and event runoff components, and \( c_{\text{p}}^{{{{\text{t}}_{\text{i}}}}},c_{\text{e}}^{{{{\text{t}}_{\text{i}}}}} \) are the respective pre-event and event concentrations of the observed tracer ti. The nature of this technique requires that (1) significant differences exist in the tracer concentrations of the components, (2) tracer concentrations of the individual components \( c_{\text{p}}^{{{{\text{t}}_{\text{i}}}}},c_{\text{e}}^{{{{\text{t}}_{\text{i}}}}} \) are constant in space and time (or variations can be accounted for), (3) additional component contributions are negligible or the tracer concentrations must be similar to that of another component, (4) the tracers mix conservatively and (5) tracer concentrations of the components are not collinear (Hoeg et al. 2000). To estimate uncertainty associated with pre-event water, event water and streamwater values, it is necessary to determine the systematic and analytical errors of the tracer signatures. The Gaussian standard error propagation technique as outlined by Genereux (1998) was applied:

$$ {W_{\mathrm{fp}}}=\sqrt{{{{{\left[ {\frac{{{c_{\mathrm{e}}}-{c_{\mathrm{T}}}}}{{{{{\left( {{c_{\mathrm{p}}}-{c_{\mathrm{e}}}} \right)}}^2}}}{W_{{{{\mathrm{c}}_{\mathrm{p}}}}}}} \right]}}^2}+{{{\left[ {\frac{{{c_{\mathrm{T}}}-{c_{\mathrm{p}}}}}{{{{{\left( {{c_{\mathrm{p}}}-{c_{\mathrm{T}}}} \right)}}^2}}}{W_{{{{\mathrm{c}}_{\mathrm{e}}}}}}} \right]}}^2}+{{{\left[ {\frac{1}{{\left( {{c_{\mathrm{p}}}-{c_{\mathrm{e}}}} \right)}}{W_{{{{\mathrm{c}}_{\mathrm{T}}}}}}} \right]}}^2}}} $$
(3)

where W is uncertainty, c is the tracer concentration, f is the fraction of total discharge and the subscripts p, e and T refer to the pre-event, event and streamwater components, respectively.

Concentrations of event water δ18O (c e) were calculated as the volume-weighted average of snowmelt water for the entire freshet period. The concentration of pre-event (c p) or old water was determined as the average δ18O baseflow value prior to the onset of the freshet hydrograph. Streamflow isotopic concentration c T was taken from grab samples on an approximately daily basis. To estimate uncertainty associated with pre-event (Q p) and event (Q e) fractions of total discharge associated with systematic errors in the isotopic signatures, the Gaussian standard error method of Genereux (1998) was applied. The limitations of this approach are reviewed in the Discussion section.

Results

Discharge

Discharge and streamflow δ18O data for the four study years are shown in Fig. 2. For inter-comparison purposes, the snowmelt period is defined as 20 April to 1 July. There was considerable inter-annual variability in patterns of freshet discharge based primarily on climate conditions during melt. In 2002, discharge increased rapidly from pre-freshet levels < 0.02 to 0.5 m3 s−1 on 16 May, and then remained high over the next 8 days with a peak discharge of ∼ 1.5 m3 s−1. Following this peak, flows gradually declined until 1 June. Two secondary peaks centred on 8 and 15 of June are associated with warm periods that melted the remaining snow within the basin. Superimposed atop this pattern are diel cycles caused by diurnal variability in snowmelt.

Fig. 2
figure 2

Discharge hydrographs, streamflow δ18O (open circles) and snowmelt δ18O (blue circles) concentrations in per mil (‰) for the study years

Full size image

In 2003, there were five distinct increased flow events attributed to enhanced periods of melt. For the first two events on 28 April and 10 May, flows increased from 0.01 to 0.1 m3 s−1 and then declined to pre-event levels as melt was suppressed. During these two events, diurnal fluctuations were not clearly defined. On 20 May, flows again increased and diel cycles became more evident. Peak freshet was observed on 6 June at 0.34 m3 s−1 and is attributed to the release of impounded water behind a snow-dam in the channel. Following 24 June, daily cycles disappeared, as snowmelt was complete. The increased discharge on 17 June was attributed to the one rainfall event during the period, which contributed 9.6 mm of rain.

For 2006, freshet began on 19 May and increased rapidly to its first peak 4.3 m3 s−1on 29 May. This was followed by a second peak on 5 June of 4.2 m3 s−1 and a final one on 12 June at 4.5 m3 s−1. Following this, flows were sustained at relatively high levels compared with 2002 and 2003. A final peak on 1 July is associated with a large (23.4 mm) rainfall event.

In 2008, discharge increased on 20 May with the main melt period from 23 to 29 May. Peak discharge was ∼1.75 m3 s−1 on 26 May, the largest flow observed in the 4 years. Flow declined in early June followed by a secondary increase between 11 and 21 June. During this period, both rain and snow events occurred in Granger Basin.

δ18O patterns

Values of snowmelt δ18O varied among the years, but in all cases were depleted compared with pre-event stream-water concentrations of δ18O (Table 1; Fig. 2). As reported in Carey and Quinton (2004) and Boucher and Carey (2010), there was large variability in both time and space of isotope composition of melt water across Granger Basin. While dealing with this variability is challenging, for the purpose of inter-comparison, event-water contributions were taken as the melt-weighted water samples from a common site near the basin outlet.

Table 1 Volume weighted event and pre-event δ18O concentrations for the study years. ± refers to 1 standard deviation
Full size table

The pattern of streamflow isotopic response was variable among the monitored years, but in all cases showed a deflection of δ18O to more depleted values, particularly at the onset of freshet, signaling a mixing of pre-event (old) groundwater with snowmelt (event/new) water (Fig. 2). In all years, streamwater became more depleted prior to the onset of hydrograph rise, signaling the arrival of meltwater to the stream. Values typically decreased 1–1.5 ‰ δ18O, which was small compared with the snowmelt response of more temperate environments (e.g. Shanley et al. 2002). In 2002, 2006 and 2008, values exhibited only small variance prior and following freshet, with deflections to lighter values associated with periods of early melt or enhanced late melt (in 2008). In contrast, 2003 presented a complex picture as the five small melt events resulted in a highly variable δ18O signature.

Freshet water-balance components

Runoff ranged between 62 mm in 2003 and 147 mm in 2008 for the 20 April to 1 July period (Table 2). While snow-water equivalent varied from 153 to 213 mm, total precipitation during the freshet period was less variable, ranging from 217 mm in 2003 to 237 mm in 2002. Runoff ratio (the fraction of total runoff to SWE + precipitation) ranged from 0.29 in 2003 to 0.66 in 2007. Rainfall from the previous year (August–October) is also presented to indicate antecedent moisture conditions prior to freezeback. Mean annual air temperature from April to June shows that 2002 was notably colder than the other years. However, despite similarity in the average daily temperatures, the pattern in SWE decline varied depending upon the sequencing of warm/cold period, resulting in large differences in hydrograph patterns.

Table 2 Water balance components and melt-period air temperature for the study years
Full size table

Hydrograph separation

For hydrograph separations, concentrations of meltwater δ18O (c e) were determined as a bulk average of meltwater collected in a common lysimeter among years. Concentrations of pre-event water stored within the catchment (c p–e) were taken as the baseflow value prior to any melt in that given year (Table 1). Note that new and old water does not indicate flow paths, but simply water input from melt and water that exists in the catchment prior to melt. New (event or snowmelt) and old (pre-event or groundwater) components of the freshet hydrograph (20 April to 1 July) are presented in Table 3. In all years, pre-event water stored within the catchment dominated the total volume of water delivered to the basin outlet. For 2002, 2006 and 2008, values of event (new) water were surprisingly close: ranging from 22 to 26 % of the total freshet discharge. In contrast, 2003 had the lowest event-water contribution of 10 %. Peak event-water contributions ranged from 26 % (2003) to 60 % (2006) of freshet, which typically occurred at the onset of the freshet hydrograph rise. Standard error was calculated using the method of Genereux (1998) and ranged between 10 and 22 % for all periods of the separation. This, however, is considered an actual underestimate due to the spatial and temporal variability of meltwater input.

Table 3 Two-component hydrograph separation results. c e is event water concentration and c p is pre-event water concentration of δ18O
Full size table

Dissolved ions and DOC

While stable isotopes provide information on source water, dissolved ions and DOC provide information on the flow pathways that water takes within the catchment to the stream. During the four study years, ∼230 water samples were analyzed for major dissolved ions, DOC and δ18O. Binary relationships among analytes for 15 April to 1 July (defined as the snowmelt period) show considerable variability; however, there are certain patterns that are consistent across the data set (Fig. 3).

Fig. 3
figure 3

Bi-variate relations among major dissolved ions (mg L−1), DOC (mg L−1), δ 18 O (‰) and stream discharge (Q)

Full size image

Concentration-discharge relationships reveal that base weathering cations (Mg2+, Ca2+) experience some dilution effect during periods of increased flow (Fig. 3). On a log–log plot, the slope of Q vs Mg2+ and Ca2+ is approximately −0.3, indicating that Granger Basin does not exhibit chemostatic behavior, yet base cations do not fully dilute with increases in flow (Godsey et al. 2009). In contrast, ions and constituents that are more biologically active (K+, NO3 −, DOC) have less well-defined relationship with discharge. In addition, DOC and K+ typically change concentration markedly during storm events as they are ‘flushed’ from the system, and hence may have dramatically different concentrations on rising and falling hydrograph limbs. When comparing δ18O values with discharge, major dissolved ions and DOC, δ18O does not exhibit any strong trends with discharge or base cations, and only a weak negative relation with DOC. The lack of correspondence highlights the complex interaction between water source and flow pathway.

When comparing relationships among major dissolved ions and constituents, a number of patterns are observed related to groundwater flow pathways (Fig. 3). Ions that exhibit dilution effects typically have a linear co-relation (e.g. Mg2+ and Ca2+, Na+ and SO4 2−). In contrast, weathering ions and constituents that vary in streamflow due to the presence of near-surface pathways (DOC and K+) show similar positive co-relations, suggesting their mobilization and mechanism of delivery is similar. δ18O does not have a strong co-relation with any other constituent, reflecting the variable controls of surface and subsurface pathways combined with sources of water on stream isotopic composition.

Temporal patterns of DOC during freshet show the relative timing and influence of near-surface pathways in organic material with those dominated by deeper pathways in the active layer (Fig. 4). As has been reported by Carey (2003) for this catchment and other permafrost-underlain watersheds (Lyon et al. 2010; Petrone et al. 2006; Striegl et al. 2007), DOC peaks on the ascending limb of the freshet hydrograph as near-surface soils saturate and mobilize DOC. However, in 2003, the pattern of DOC delivery is complicated by the timing and nature of freshet. The greatest pulse of DOC is delivered during the first rise in the freshet hydrograph in all years, and declines exponentially thereafter as mobile DOC is removed from the soil and the water tables decline. However, the flushing of the DOC is likely a greater cause for the decline than declining water tables, as evidenced by the 2003 patterns where DOC declined rapidly despite water tables being at or near the ground surface for a sustained period.

Fig. 4
figure 4

Discharge hydrographs and streamflow dissolved organic carbon (DOC; mg L−1) for the study years

Full size image

Discussion

In most years, snowmelt is the dominant hydrochemical and hydrological event in subarctic and permafrost catchments. During melt, areas underlain with permafrost control the runoff response by limiting deep percolation, and confine flow pathways to near-surface horizons, typically through organic soils that rapidly convey water to the stream (Quinton and Marsh 1999; Carey and Woo 2001). These organic soils are capable of transmitting water during melt, even when at 0 °C, and no re-freezing occurs. Unsaturated frozen organic soils allow rapid infiltration and subsequent saturation and downslope movement of water (McCartney et al. 2006), where interaction with water in the thawing soil occurs. Throughout the year, the pathways that runoff takes are strongly controlled by the position of the frost and water table as a positive feedback occurs where more rapid runoff rates (and more runoff from precipitation inputs) occur when the water and frost table is near the surface (Carey and Woo 2001; Quinton et al. 2009). The source of this water, whether it is water that resides within the catchment prior to inputs (old water), or water precipitated on the catchment transferred directly to the stream (new water), has concerned hydrologists for several decades in permafrost catchments, and it typically thought that low storage capacities and high runoff ratios, particularly during melt, would result in new water (derived from melt) dominating the hydrograph.

Results from this study suggest that for Granger Basin, an alpine discontinuous permafrost catchment, that freshet is largely derived from water that existed in the soils prior to melt. Values are low compared with others for permafrost-underlain soils, which typically have larger new water contributions (Cooper et al. 1991; 1993; Gibson et al. 1993; McNamara et al. 1997; Metcalfe and Buttle 2001; Obradovic and Sklash 1986;). Considering the differences in the timing and magnitude of melt, the water-balance components, and discharge, there is limited variability in event-water contributions among years 2002, 2006 and 2008. For these 3 years, runoff ratios were high, and their data suggest that these are related to antecedent wetness of the catchment reflected in the previous fall’s precipitation. In addition, these years had greater total precipitation during the melt period (SWE + precipitation) and freshet was dominated by a single major event with one or two peak melt periods. In contrast, 2003 was characterized by slow melt spread among five main events. This provided increased infiltration opportunity time, lower runoff ratios and significantly lower snowmelt water contribution. An event-water fraction of 10 % is one of the lowest values for snowmelt runoff reported in the literature for either temperate or permafrost basins.

With such large quantities of old water reporting to the stream during freshet, and runoff ratios that are relatively high compared to temperate catchments, there must be considerable storage within the catchment of event water during freshet and a mechanism to displace old water residing over-winter to the stream. In Granger Basin, organic soils cover much of the lower elevation zones with depths ranging from 0.2 to 0.6 m. The porosity of these soils is typically >80 % (Quinton et al. 2008), allowing for several hundred mm of total storage when saturated and considerable storage at field capacity (Quinton et al. 2005). Prior to freeze-back in the fall, organic soils are typically unsaturated—ranging from 18 to 40 % by volume; see Quinton et al. (2005), their Fig. 3— suggesting considerable capacity for infiltration in the basin, as reported by McCartney et al. (2006). In Granger Basin, Carey and Quinton (2004) observed a strong depletion in δ18O for groundwater wells (their Fig. 4) in response to infiltrating meltwater, yet a concomitant displacement of water existing in soils is required to account for the predominantly old-water signal in streamflow. The mechanism of this displacement is not completely clear, yet it is hypothesized that during early stages of melt, there is limited transfer of water to the stream until unsaturated storage capacity is satisfied and infiltrating meltwater supplies sensible heat to bring soils to 0 °C (although not necessarily fully thawed). At this time, lateral runoff pathways begin to activate and mixing with thawing old water occurs. The contribution of old water gradually increases throughout the melt period (Carey and Quinton 2004; Boucher and Carey 2010) as mixing and displacement of old pre-event water occurs.

There is a large degree of uncertainty in the two-component hydrograph separations presented. While the Gaussian error propagation method of Genereux (1998) displays errors between 10 and 20 %, these errors are likely higher as the sampling strategy for measuring meltwater inputs is limited in its spatial scale. Snowmelt isotope data from across Granger Basin (data not shown), while limited, suggests that the spatial variability in meltwater isotopic signature is large, particularly as a result of differential melt timing, (Taylor et al. 2002) and characterizing event-water from one site does not adequately characterize the c e signal. In addition, the event-weighted method for determining c e assumes no catchment storage and that all meltwater is instantaneously routed to the streams. Other methods such as those presented by Laudon et al. (2002), may more appropriately characterize event-water signatures, yet require detailed information on basin storage which is difficult to determine at larger scales.

While multi-year isotope data confirm the results of Carey and Quinton (2004) that old water residing within Granger Basin dominates the freshet hydrograph, dissolved ions and DOC indicate the groundwater flow pathways. Runoff processes have been previously studied in Granger Basin at both the plot and basin scale (Boucher and Carey 2010; Carey and DeBeer 2008; Carey and Quinton 2004, 2005; McCartney et al. 2006; Quinton et al. 2009) and hydrometric evidence indicates that surface flow is largely absent, and that supra-permafrost groundwater, primarily flowing through near-surface organic soils, controls the rate and timing of the snowmelt hydrograph.

Reports of dissolved ion flux from permafrost watersheds are scarce (Petrone et al. 2006), particularly over multiple years (Keller et al. 2010). Petrone et al. (2006) compared three catchments with high, medium and low levels of permafrost coverage in interior Alaska during one season and highlighted how catchments with greater permafrost had higher DOC and lower dissolved ion fluxes; however, total yields were related to the total amount of runoff, which varied among catchments. In Granger Basin, 4 years of data from one catchment show similar results. Deeper groundwater flow pathways through mineral soils from permafrost and permafrost-free areas of the catchment largely control the flux of weathering ions (Ca2+, Mg2+ and Na+) as some dilution occurs during freshet and during large rainfall-runoff events when flow occurs through the near-surface organic horizons (Fig. 3). The level of dilution that occurs with increases in flow is at the upper end of those reported by Godsey et al. (2009) from 59 geologically diverse continental US catchments. DOC and K+ show distinct peaks prior to (in the case of DOC) and nearer the peak (for K+). DOC is not a particularly effective indicator of water flowing through near-surface horizons as it, and other biologically derived and active elements, appear to be ‘flushed’ from soils early during the freshet period, yet it does highlight the onset of freshet prior to the rise in the hydrograph. A combination of stable isotopes, dissolved ions and DOC provide a more integrated picture of groundwater flow pathways operating in Granger Basin.

The multi-year data set presented here conforms to previous conceptual models of runoff generation in subarctic catchments and the influence of different groundwater flow pathways on the quantity and chemistry of streamflow. Inter-annual variability in freshet can be large and is complicated by both the amount of precipitation (as snow and rain) and the timing of freshet. In years such as 2003 with complex melt periods, streamflow chemical response, particularly with regards to isotopes and constituents derived from near-surface flow pathways, are dramatically different than other years. In contrast, base weathering elements show similar dilution effects with flow, as their source to the stream appears to be largely controlled by more permanently connected intra-, sub- and permafrost-free groundwater sources.

Results from headwater catchments provide an informative contrast compared with data from large rivers, which integrate extremely large areas, diverse flow pathways, sources of water and complex chemical interactions. Changes observed in large continental rivers are compelling, yet the evidence of thawing permafrost as the cause is largely circumstantial as there have been few observations that directly link the two trends. Headwater catchment studies, particularly if monitored over long time-scales, provide the opportunity to observe these changes in a constrained system and provide evidence for process-based explanations. This study, while presenting only 4 years of data, highlights the variability and similarities in hydrological and chemical response of a discontinuous permafrost headwater catchment. Intra-annual variability at the headwater scale is such that it will take several decades before definitive links between permafrost thaw and hydrological response are clearly understood in the many diverse permafrost environments. While the role of groundwater response is central to this issue, the lack of physical infrastructure (i.e. complex well and piezometer networks) require inference from streamflow hydrometric and chemical change.

Conclusions

Hydrograph separations, water-balance components and major dissolved ions are presented for four freshet events in a discontinuous permafrost alpine headwater catchment. Results from the hydrograph separation suggest that meltwater accounted for between 10 and 26 % of streamflow during freshet, the vast majority being derived from water displaced from near-surface catchment soils during melt. In years with a single main freshet event, there was little difference in the amount of event-water contribution, yet in 2003 when melt was protracted, event water was markedly lower. The volume of discharge during freshet varied among the years, and was not directly related to snow-water equivalent or meltwater input plus new precipitation during melt. Other factors such as antecedent wetness and total meltwater delivery time are important factors controlling the fraction of snowmelt reaching the stream. Dissolved ions and dissolved organic carbon provide information on the flow pathways of groundwater. In contrast, DOC and biologically active ions (such as K+) reflect near-surface pathways, yet are not closely linked with discharge as their availability to be removed in solution changes during freshet.

The importance of changing groundwater pathways and volumes on streamflow response in permafrost environments is central to understanding changes in large permafrost rivers. However, at the headwater scale, inter-annual variability in water-balance components will complicate the identification of long-term trends. Long-term hydrological observatories in permafrost environments are central to testing hypotheses of the impacts of permafrost thaw on groundwater and hydrological response to change. While thawing permafrost will undoubtedly impact groundwater and streamflow response, changing seasonality and precipitation phase and timing (e.g. Spence et al. 2011) will have confounding effects on the interpretation of data.