HYDROMETEOROLOGY

Snow surveys were conducted in May/June 2001 and 2002 using the terrain index method (Yang and Woo, 1999). Each snow survey transect consisted of 10 snow depth measures at 10-m intervals, with one measure of snow density and SWE at the mid-point. Some rolling hill, gullied terrain and slope aspect units were sampled at 5-m intervals as the 100-m-long standard transect exceeded the length of the feature or the 10-m interval was too coarse to capture variability along the unit. SWE measures were extrapolated along the transect length, averaged at each site, and finally compared by terrain unit and spatially by altitude and specific catchment.

Meteorological conditions in 2001 were recorded at two stations located at 195 and 400 m a.s.l., and in 2002 at five stations ranging between 195 and 550 m a.s.l. (Fig. 1). The additional stations in 2002 were used to improve estimates of lapse rate and rainfall variability across the watersheds. Weather stations recorded temperature at 1.5 m above the ground with an Onset Hobo temperature logger (0.2°C resolution) at 10-min intervals. Precipitation was recorded with a Davis industrial tipping bucket rain gauge (0.2 mm resolution) logged with an Onset event logger. In 2002, the primary weather station (Campmet) was equipped with a Davis solar radiometer (±4%), industrial model anemometer, and tipping bucket rain gauge logged with a Unidata Prologger at 10-min intervals. Air temperature and humidity at Campmet were recorded with an Onset Hobo Pro logger at 1.5 m elevation at the same time intervals.

Discharge measurements were conducted using the velocity-area method, whereby velocity and channel depth measurements were obtained from Zodiac boats using a cableway, and stage was recorded at 10-min intervals with a stilling well equipped with a calibrated Sensym SCX differential pressure transducer vented to the atmosphere and logged with an Onset Hobo logger. Hydrographs were developed for each river using stage-discharge rating curves (Forbes, 2003). Water temperature was measured by suspending an Onset water temperature sensor (0.2°C resolution) and Hobo logger from the cableway in the thalweg.

WATER SAMPLING

Water samples for suspended sediment concentration (SSC) were obtained daily in the Lord Lindsay River and East Tributary in 2001 (n = 70) and twice daily in all three rivers in 2002 (n = 121) during the snowmelt and storm flow periods, and once every 2 to 3 days during recession and baseflow periods. Additional samples were taken during peak flow to determine diurnal variability. Each sample was vacuum filtered for suspended sediment through pre-weighed 0.45 μm cellulose nitrate or polycarbonate filters. The filter papers were stored, dried and weighed in the laboratory.

Attempts to obtain continuous turbidity records from the rivers were not successful due to low SSC and equipment failures. Hourly SSC records were established for all rivers using discharge-SSC rating curves. Daily SSQ records were then compiled by cumulatively summing the hourly SSC data.

HYDROMETEOROLOGICAL ANALYSIS

Hydrological and meteorological relationships were assessed using regression analysis in order to determine the climatic controls on water and sediment outputs. The input variables considered were mean daily incoming short-wave radiation from the Campmet station (Fig. 1) and mean daily air temperature from all available meteorological stations. However, air temperatures from Uppermet were used exclusively for several reasons: (1) data was available from both seasons, thus allowing for interseasonal comparisons; (2) the station elevation (400 m a.s.l.) was representative of the largest proportion of the overall watersheds; and (3) observed relationships between air temperature and hydrological output variables showed only minor variability between stations (Forbes, 2003). The dependent variables considered were total daily runoff and mean daily SSC for each respective river. SSC values were based on manual sample measurements, whereby only the afternoon suspended sediment sample was available in 2001 and a mean value of the morning and afternoon sample was taken in 2002.

Daily hydrological outputs were found to substantially lag daily thermal conditions. Woo (1976) attributed this phenomenon to the changing areal extent and location of basin snow cover relative to the stream gauging stations. In order to allow direct comparisons of hydrometeorological variables, it was necessary to establish a fixed “hydrological day” for each river that captured the daily water and sediment outputs in response to the events of the meteorological day (0000–2400h). The average time of daily peak flow during the snowmelt period was calculated for each river and served as the midpoint for the hydrological day. The hydrological day was also applied to suspended sediment data as SSC and discharge were found to be consistently in-phase (e.g., Braun et al., 2000). The mean time of peak flow occurred at 0140h and 0240h for the West and East Tributaries, respectively, and 0600h for the Lord Lindsay River.

It is important to point out that the application of a fixed hydrological day is an oversimplification of the physical processes that occur across a watershed (Hardy, 1996). However, generalizations made in this study are warranted by the fact that considerably greater error would be introduced to the analysis by not applying any form of hydrological lag time. Over- or under-estimation that may result from applying a fixed hydrological day through the season would occur during daily low flow, a relatively insensitive period of the hydrograph.

HYDROMETEOROLOGICAL CONDITIONS IN 2001 AND 2002

Both field seasons were substantially different due to a number of factors. The winter SWE that preceded the 2002 streamflow season was 78 mm, or, approximately half that of the 132 mm in 2001 (Table 1). Consistent with other arctic studies (e.g., Kane et al., 1991; Woo and Young, 1997; Yang and Woo, 1999), gullies, hollows, and channels collected approximately three times more snow than the broad plain terrain unit that dominates the watershed (Table 1). A more representative terrain index snow survey from 2002 confirmed the findings of Yang and Woo (1999), showing that at large scales, the dominant terrain feature showed the least variability and was most representative of the mean SWE for the watershed (Table 1). Elevation did not appear to have significant effect on SWE (Forbes, 2003). Based on this snow distribution it was possible to use the more limited snow survey data from 2001 (Fig. 1), which was estimated exclusively from transects conducted across broad plain terrain units.

The 2001 season had two major rainfall events, with 29 mm recorded during the largest at the Uppermet station on 12–13 July (Fig. 2a). The 2002 season was characterized by numerous trace rain and wet snow events and an 8-mm rainfall on 7–8 July. Soil moisture surveys and field observations noted the watershed to be highly saturated well into the snowmelt recession period of both seasons, followed by a rapid drying and expansion of the active layer, and brief return to saturated conditions after the major rainfall events. The saturated layer was observed to drop to 80 cm depth by the late summer in 2001. Estimated evapotranspiration rates, based on gravimetric measures from a network of soil samples, were 0.8 and 0.6 mm d⁻¹ in 2001 and 2002, respectively (Forbes, 2003).

Air temperature and incoming solar radiation followed strong diurnal variation through the season (Forbes, 2003). A vertical lapse rate was observed across the watershed, varying both intra- and interannually, and is evident in the comparison of air temperature records between Campmet and Uppermet stations (Fig. 2a). In 2001, an average lapse rate of 1.5°C/100 m was calculated from first flow (10 June) until the end of the snowmelt recession period (8 July), and lowered to 0.3°C/100 m in the last two weeks of study season. In 2002 the average lapse rate was 1.0°C/100 m and remained relatively stable until the end of the snowmelt recession period (1 July). During the final week of observations, the lapse rate was 0.7°C/100 m. The steady decline of the lapse rate through the snowmelt period and into the summer was likely due to waning snow cover and suggests that snowmelt followed an altitudinal retreat.

Daily mean air temperatures from both Campmet and Uppermet were highly correlated with the nearest weather station located at Taloyoak (r² = 0.93, n = 122, p < 0.01; and r² = 0.92, n = 99, p < 0.01, respectively). The timing of precipitation events between the stations was also similar, although the amounts were not as well correlated (r² = 0.64, n = 34, p < 0.01; and, r² = 0.56, n = 34, p < 0.01, respectively). This is a well-documented problem in arctic regions where meteorological records from coastal stations tend to underestimate precipitation relative to interior sites (Woo et al., 1999). Nonetheless, it can be concluded that the watersheds broadly experienced the same weather patterns as Taloyoak on a synoptic scale.

The diurnal pattern of the discharge hydrographs through the spring melt periods suggests that runoff was governed by the atmospheric energy available for snowmelt. Daily discharge amplitude dampened with the exhaustion of the snowpack. River water temperature also followed a diurnal pattern through the snowmelt period with opposite amplitude to discharge. The coldest waters coincided with highest discharge during peak snowmelt (Forbes, 2003). The cycle was dampened during the 2001 peak flow period, most notably for the Lord Lindsay River, indicating an almost complete snowmelt flush.

Subtle climatic differences between years produced pronounced differences in the hydrographs. Specific flow periods have been delineated with regards to the state of the snowpack and discharge response (Fig. 2b). Variable early season temperatures in 2001 caused an initial melt peak (1) to be interrupted by an inter-flood low-flow period (2). Subsequent warming resulted in the rapid rise (3) and recession (4) of the main snowmelt flood peak. Finally, a summer storm period (5) punctuated the baseflow period (6). The 2002 hydrograph was marked by a single prominent flood peak (1) in response to the spring snowmelt and followed by rapid recession (2) to baseflow (3). A week of modest early snowmelt flow preceded the main 2002 peak discharge (14–21 June) due to sustained cool temperatures (Fig. 2a). Maximum discharges were comparable between years: 400 and 360 m³ s⁻¹ for the Lord Lindsay River, and 100 and 90 m³ s⁻¹ for the East Tributary for 2001 and 2002, respectively.

SUSPENDED SEDIMENT TRANSPORT

Suspended sediment concentration (SSC) was found to increase both linearly and non-linearly with discharge throughout the streamflow season (Fig. 3). Although sampling resolution prevented a more detailed analysis, it was evident that high flows were accompanied by variable, but relatively high SSC (10–30 mg L⁻¹), whereas subsequent low flows were characterized by low SSC (<2 mg L⁻¹). Results from 2002 indicate that SSC was generally higher for the morning (peak flow) compared to the afternoon (daily low flow), but during low to moderate flow periods the difference was minimal (∼1 mg L⁻¹).

High suspended sediment transfer only occurred during the short-lived nival peak, suggesting that a threshold discharge was required in order to generate significant sediment transport. As previously mentioned, peak flow rates were comparable between seasons but, due to greater SWE, were sustained for approximately four days in 2001 compared to only one day in 2002. As a result, seasonal sediment yields were almost four times greater in 2001 than 2002 (Fig. 2c). Greater than 80% of total seasonal runoff and 93% of total annual suspended sediment transport occurred during the snowmelt periods, with approximately half the annual sediment being transferred during the maximum flow periods alone. The dominance of the snowmelt period for suspended sediment transport is consistent with other arctic studies (e.g., Woo et al., 1996; McNamara et al., 1998) and emphasizes the importance of the annually recurring snowmelt flood to suspended sediment transport (Braun et al., 2000). Nonetheless, the specific sediment yields recorded are some of the lowest reported from the Arctic, ranging between 0.2 and 1.9 t km⁻²·a⁻¹ (Fig. 2c).

SUSPENDED SEDIMENT-DISCHARGE HYSTERESIS

Suspended sediment-discharge hysteresis was evident both intra- and inter-seasonally with varying intensity in each watershed. Relationships between hydrological variables were analyzed at different times of the season and diurnally when possible (Fig. 4).

EARLY MELT AND RECESSION: MODERATE FLOW

Broadly, periods of moderate flow in 2001 were characterized by higher SSC than similar discharge in 2002 (Fig. 4). Although larger rivers do not typically experience the effects of snow damming that both hinder flow and entrain sediment on smaller streams (e.g., Woo and Sauriol, 1983), snow- and ice-lined channels can act to buffer against fluvial erosion and thaw (Forbes, 1975; Scott, 1978; Woo and McCann, 1994). Deep channel snowpack was a common feature in the watersheds during both seasons of study. However, higher SSC at moderate flow rates in 2001 appears to suggest that rivers were better able to counter snow and channel ice effects and access greater sediment sources. This may be attributed to both the nature of the river breakup and time required to reach peak flow in each year.

Soon after initial flow in 2001, ice began to choke river channels, having either emerged from channel bottoms or calved from banks. Ice jamming (Prowse, 1991) followed and led to brief periods of hydrological back-up. Through the higher flow of the initial peak (15–19 June), ice became mobile and floes were observed to scour channel banks, beds, and other ice slabs (Mackay and Mackay, 1977). When stage dropped, substantial erosion marks were observed along channel banks. By comparison, the ice breakup in 2002 was less pronounced, ice floes were greatly reduced, and those that emerged from channel bottoms and banks generally remained in place. As previously mentioned, the first week of discharge in 2002 was marked by cooler temperatures and reduced flows that limited ice mobility. Thus, much of the 2002 channel ice was allowed to gradually decay before stage substantially rose.

Peak discharge occurred after 22 days of river flow in 2001 and only 13 days in 2002, while the recession was delayed a further four days in 2001 as a result of sustained maximum flows. The extended delays may have allowed for extended channel bed and bank thaw (Clark et al., 1988) and subsequent increased sediment availability to the rivers. Miles (1976) noted the importance of thermal effects on the perimeter of the channel bed, explaining that high water temperatures can facilitate greater sediment erosion and transport. Various bank failures, including processes of subsidence and slumping, can occur along channel banks when materials of high ice content are subjected to melt by river water (Church and Miles, 1982). Some measure of this effect can be obtained from the cumulative thermal energy in the river. River water was measured to have exerted more than 52 “River Melting Degree Days” (RMDD) from initial flow to peak in 2001 compared to 11 RMDD in 2002 (Table 2).

Greater exposure to thermal energy inputs from river water that resulted in reduced channel snow may also help to explain the higher SSC in the East Tributary during the peak flow period of 2001 compared to similar discharge rates during the initial melt peak (Fig. 4). Although discharge rates and effects of river breakup appear to have been substantial enough to clear the majority of channel snowpack on the larger Lord Lindsay River, snow and ice-lined channels were observed to persist longer on the East Tributary. As a result, a different suspended sediment–discharge rating curve was apparent for the river during the early melt period (Fig. 3).

The inter-flood low-flow period in 2001 was marked by moderate sediment transport in both rivers (5–8 mg L⁻¹, n = 6) (Fig. 4). These SSC measurements were not applied to the rating curves as they were in poor agreement with later season baseflow where SSC remained consistently low (<2 mg L⁻¹, n = 14). The moderate levels of SSC during this period were perhaps a residual effect of the river ice breakup during the initial melt peak that mobilized suspended sediment and transported moderate sediment concentrations of between 7 and 9 mg L⁻¹ for four days, only to be abruptly terminated by cold temperatures. Furthermore, a severe wind storm accompanied the colder temperatures during the inter-flood low-flow period and contributed substantial eolian debris to the channels.

PEAK FLOW

Maximum flow occurred in channels with reduced snowpack and resulted in consistent SSC-discharge relationships. During each season, suspended sediment load from the falling limb of the main melt peak showed minimal evidence for hysteresis in the Lord Lindsay River and East Tributary (Fig. 4). As an interesting comparison, the mean daily SSC in the Lord Lindsay River during peak discharge from both seasons were 14.6 and 14.9 mg L⁻¹, respectively.

Pronounced suspended sediment–discharge hysteresis was evident in the West Tributary immediately following peak flow, from the afternoon of 27 June to the morning of 29 June. However, it is important to note that hydrological backup was observed at the gauging station during this period due to peak flow on the main river, and as a result discharge was likely overestimated. Manual measurements indicated discharge to be less than half of stage-discharge rating curve estimates for this time. The narrow range of SSC values during this period prevented the use of a separate rating curve. Instead, linear interpolation from measured SSC values was used to provide a more accurate estimate of SSQ.

The 26 June 2002 flood peak caused exceptionally high SSC on the East Tributary (>30 mg L⁻¹, n = 2). The river water was observed to be unusually turbid for at least 12 hours. Due to the relatively extreme nature of these SSC samples, the values were not included in the SSC-discharge rating curve (Fig. 3). In order to estimate the significance of this event, linear interpolation was used from the point that these turbid conditions were first and last observed (26 June 2020h to 27 June 0820h) with the measured SSC values through this period. These estimates suggest that this event increased the annual East Tributary sediment yield by c. 35%. During this event, the river was observed to have overtopped its banks at the stream gauging station and various locations upstream, potentially mobilizing substantial amounts of floodplain sediment.

RAINFALL RESPONSE

Lower intensity rainfall (<8 mm d⁻¹) produced both minimal discharge and suspended sediment response, confirming the observations of other arctic studies (e.g., Kriet et al., 1992; Braun et al., 2000). Discharge after the 29-mm rainfall event in 2001 increased substantially in both rivers (Fig. 2b), while SSC increased only slightly (∼1–3 mg L⁻¹) (Fig. 4). Summer storms generally occur when materials are thawed and thus more readily eroded (Woo and McCann, 1994). Furthermore, rain events occur across the entire watershed and are potentially capable of accessing sediment sources otherwise protected from fluvial erosion (Braun et al., 2000; Lamoureux, 2000). The minimal sedimentological response across the watersheds suggests that erodable sediment was limited. Sediment exhaustion has been identified in many arctic nival catchments (Lewkowicz and Wolfe, 1994; Woo and McCann, 1994; Braun et al., 2000). The wide range in SSC values during the rainfall event period prevented the use of a separate rating curve. Linear interpolation was used with the measured values from 12 to 17 July to provide an estimate of total SSQ.

METEOROLOGICAL AND HYDROLOGICAL ASSOCIATIONS

Significant linear correlations between daily mean air temperature and daily runoff, and mean SSC, were found through the main snowmelt periods of each streamflow season (Table 3). By contrast, cross correlation between mean daily incoming solar radiation and both discharge and SSC produced weak relationships (r² < 0.27). The strength of the temperature relationships deteriorated with the onset of recession as the snowpack was depleted. The strength of correlations was substantially weaker through the early melt of both seasons, likely due to limited fluvial erosion in snow- and ice-lined channels and low water temperatures. These results suggest that the potential for sediment erosion in many cases far exceeded the supply of materials (Woo and McCann, 1994), preventing strong relationships between hydrometeorological variables.

Cumulative watershed outputs of water and sediment showed a linear response to cumulative mean degree-days (MDD) until SWE was depleted (Fig. 5), suggesting that both were initially linked to variations in snowmelt (e.g., Hardy, 1996). It is interesting to note that the cumulative suspended sediment yield and degree-day plots for both seasons in the Lord Lindsay River showed maximum daily discharge peaks and subsequent sediment transfer peaks to have occurred after 18.4 MDD, despite other substantial meteorological differences. The East Tributary experienced similar sediment transfer peaks in 2002 and the onset of high sediment transfer conditions in 2001 following the same number of MDD. However, the sustained peak flow period consisted of only between 10% and 15% of the MDD in both years, but resulted in approximately 50% of the sediment delivery. Hence, although melt energy was a direct control over snowmelt, cumulative discharge and suspended sediment yield were limited by SWE (Fig. 5). Comparison of cumulative sediment yield between 2001 (SWE = 132 mm) and 2002 (SWE = 78 mm) demonstrates this control is proportional, but two years of field data are insufficient to identify a quantitative relationship.

LIMITATIONS OF AIR TEMPERATURE FOR FORECASTING STREAMFLOW: SNOWPACK EXHAUSTION

Thermal energy appears to be critical in generating streamflow and suspended sediment transfer in large middle arctic catchments, but only until the SWE begins to undergo substantial depletion. Once watershed snowmelt begins to wane, a decoupling occurs between the atmospheric energy input (e.g., air temperature) and hydrological output. Braun et al. (2000) also noted that coherent relationships between atmospheric and hydrological variables in the Sophia River watershed on Cornwallis Island weakened and eventually disappeared as a result of snowpack exhaustion. Although late-lying snowbanks and channel snowpack were both observed and measured with hydrochemical indicators to contribute approximately 15–50% of summer flow in the Lord Lindsay River and tributary watersheds (Forbes, 2003), lower summer discharge was not at levels to support high SSC. In this regard, the hydrological output during summer baseflow was not representative of melt energy input, and for all intensive purposes the watershed was exhausted of snowpack that could contribute appreciably to streamflow and sediment transfer.

There is good reason to believe that snowpack exhaustion is not unique to the two seasons of study. Snowpack exhaustion has been observed to be a recurring phenomenon in studies at the Meecham River (Cogley, 1975) and McMaster Rivers (Woo, 1983) on Cornwallis Island and Hot Weather Creek (Lewkowicz and Wolfe, 1994) on Ellesmere Island. For these nival-regime catchments, snow accumulation during each previous winter acts as the primary control on watershed streamflow and suspended sediment transport (Church, 1988). In our study, only peak flow discharge rates were capable of moving high SSC (>10 mg L⁻¹). Increased SWE in 2001 resulted in a prolonged period of maximum annual discharge and substantive sediment transfer. Thus, annual suspended sediment load in this environment was ultimately a function of total peak discharge through SWE rather that a function of melt energy.

In addition to Braun et al. (2000), the only other comparable investigation of climatic influences on hydrological outputs in arctic regions, not hindered by the presence of snow dams (e.g., Wedel et al., 1977; Woo, 1983), was conducted by Hardy (1996) at the Lake C2 catchment (21 km²) on northern Ellesmere Island. Here, the magnitude of daily discharge was found to be limited by melt energy rather than the supply of snow and appears to argue against a SWE control. However, Hardy (1996) noted that the Lake C2 watershed had substantial areas of perennial snow. Although late-lying snow contributed to relatively minimal discharge and suspended sediment outputs within the catchments on the Boothia, it is possible that perennial snowbanks in high arctic catchments could contribute appreciably to watershed outputs. Marsh and Woo (1981) and Lewkowicz and Young (1990) also showed that the runoff regime in many areas of the High Arctic is dominated by late-lying snow. Snowpack persistence due to colder temperatures could explain why associations between watershed outputs and air temperature at Lake C2 remained strong through the summer streamflow season.

More generally, Sampson (2001) found strong correlations between total annual discharge and the magnitude of the nival peak in the Hayes River, a middle arctic watershed (18,100 km²) ∼300 km south of the Lord Lindsay River, during the 1970s (r² = 0.79) and 1980s (r² = 0.84). Associations between daily mean temperature and discharge were found to be much lower (r² < 0.30). These results further reinforce the importance of snowpack at increasing spatial scales.

The large watersheds in this study experienced long delays to both daily and seasonal peak flow as a result of substantial routing and muted topography. While a single day of warm weather may be capable of generating a rapid runoff response accompanied by relatively high sediment transport on a small stream, larger watersheds require greater hydrological inertia to produce substantial discharge. High rates of discharge and sediment transfer did not occur in our study rivers until the meltwater source areas became highly integrated (Bowling et al., 2003). Thus, the maintenance of peak flow rates at the large scale was a culmination of hydrometeorological processes. Although similar amounts of melt energy (18.4 MDD) were required to reach peak discharge during both seasons, it was catchment SWE that appeared to control the duration of peak flow for all rivers, and hence, annual sediment yield.

RAINFALL-INDUCED SUSPENDED SEDIMENT TRANSPORT

With the depletion of the snowpack, watershed outputs in nival regime arctic watersheds generally only respond to summer rainfall (e.g., Braun et al., 2000). Church (1988) suggested that large rain events in arctic environments have the potential to produce sediment yields larger than the nival peak. This has been supported by exceptionally high sediment loads following a major storm in a glacierized high arctic watershed (Cogley and McCann, 1976) and inferred from a sedimentary record (Lamoureux, 2000), but has yet to be quantified within a nival regime arctic catchment (e.g., Hodgson, 1977; Braun et al., 2000). The fact that a major rainfall event (>25 mm) in this study was unable to produce a substantial sedimentological response can be at least partly attributed to sediment source limitations. However, there is also reason to believe that differences in watershed vegetation cover, topography, and spatial scale may have reduced sediment erosion compared to other high arctic studies.

High arctic catchments are generally poorly vegetated and can subsequently exhibit greater response to rainfall (Woo and Young, 1997). Considerable sediment delivery can occur through wash on poorly vegetated slopes flows, gullying, and localized earth flows (Woo and McCann, 1994; Lamoureux, 2000). The Lord Lindsay River and tributaries have considerably higher vegetation cover (∼80%; Laidler, 2002) than high arctic sites and are less susceptible to these forms of erosion. Thick vegetation cover in the subarctic has been shown to buffer the soil from erosion by both rain drops and running water, while the root system strengthens the soil against mass wasting (Woo and McCann, 1994). Furthermore, Lamoureux (2000) argued that localized mass wasting processes would be unlikely to contribute appreciably to sediment yield from a large catchment.

High rates of rain-induced sediment erosion have also been typically observed in smaller catchments (<100 km²). Kriet et al. (1992) noted that larger rivers such as those on the north slope of Alaska, including the Kuparuk River (8100 km²), have sufficient watershed area to buffer local summer storms. In the same manner, the Lord Lindsay River and tributaries have not only large areas but limited relief (>350 m). The low-slope gradients attenuate surface flow and increase water ponding, which infiltrates or evaporates rather than contributing to runoff and sediment erosion (Bowling et al., 2003).

As a final point, rain-induced sediment erosion is generally dominated by the erosion of channel banks and remobilization of floodplain materials (Cogley and McCann, 1976; Woo and McCann, 1994), processes which require high river stage. In this study, high sediment yield only occurred when rivers were near bankfull. The rain event appears to have been substantial enough to satisfy the watershed storage capacity and generate runoff, but not to elevate stage to the level necessary to initiate channel bank erosion and access substantial amounts of floodplain materials.

IMPLICATIONS FOR CLIMATE CHANGE

Climatic changes will inherently influence hydrological processes at varying response times (Woo and McCann, 1994). Studies like this one are of particular importance to understanding the sensitivity of arctic environments to future climate change as well as toward making inferences of past hydrological conditions.

Of particular concern is human-induced climate change, due to activities that have increased atmospheric content of carbon dioxide and other greenhouse gases. General circulation models (GCMs) offer the most plausible scenarios of future anthropogenic climate change (Woo and McCann, 1994), and although forecasts from the various GCMs are not in complete agreement, the consensus is that the arctic environment will be particularly sensitive and experience amplified warming (Watson et al., 1998; Houghton et al., 2001). Broad increases in precipitation have also been predicted at high latitudes (Vörösmarty, 2001), although forecasted changes to precipitation patterns and intensity are considerably more uncertain than temperature (Woo and McCann, 1994).

In the case of runoff in large middle arctic catchments, results from this study suggest that increased winter snowfall would be of greater influence over watershed outputs than increasing melt season temperatures. Higher winter SWE would extend the duration of the nival peak and period of high sediment transfer. Woo and McCann (1994) have also suggested that eventually, a warmer climate would support more extensive vegetation across the region. This may further reduce sediment erosion rates during rain events and provide a greater buffer against erosion during the snowmelt flood.

Past arctic climates have been interpreted from varved lake records through the link between river discharge, suspended sediment delivery, and sediment deposition (e.g., Hardy et al., 1996; Lamoureux, 1999). In the case of Sanagak Lake, located downstream from the rivers in this study (Fig. 1), the amount of sediment entering the lake each year is primarily the function of the duration of the nival peak, essentially a period of several days. By extension, sediment accumulation in the lake can be viewed as proportional to the cumulative sediment delivery in the river (Fig. 5). While it is possible that heavy rain events (>30 mm) would be capable of producing additional sub-annual laminae in some watersheds, results from this study suggest the sediment delivery response and resultant deposition in the lake would be minimal for a c. 25-mm event (Fig. 5). These results point to the necessity of understanding the specific catchment controls over sediment availability and transport in order to interpret sedimentary records as paleohydrological proxies.

Although more seasons of study would be necessary to quantify the relationship, the close association between meteorological variables in this study and the Taloyoak weather station warrants the comparison between the instrumental climate and sediment record in Sanagak Lake as a proxy for past discharge. In particular, SWE in this study was proportional to total winter snowfall at Taloyoak (164 and 103 cm, for 2001 and 2002, respectively). Therefore, the potential exists to extend the analysis of SWE variability during the past 50 years using the sedimentary record (e.g., Hardy et al., 1996).