hydrology of subarctic canadian shield: soil-filled valleys
TRANSCRIPT
Hydrology of subarctic Canadian shield: soil-filled valleys
Christopher Spence*, Ming-ko Woo
School of Geography and Geology, McMaster University, Hamilton, Ont., Canada L8S 4K1
Received 15 April 2002; accepted 28 April 2003
Abstract
The Canadian Shield landscape includes bedrock uplands and valleys infilled with soil. A site near Yellowknife in subarctic
Canada was studied to elucidate the hydrological behaviour of soil-filled valleys. A suite of hydrological processes was found to
be important in the studied valley, including snowmelt and rainfall, ground frost development, evaporation, infiltration, lateral
inflow from adjacent uplands, and surface and subsurface flows. Valley storage requirements during snowmelt are met by local
meltwater infiltration, but lateral inflows in the summer are needed to satisfy the storage before runoff can be generated from the
valley. The valleys perform three functions of collecting vertical and lateral water inputs, retaining and losing the water held in
storage, and transferring water down the valley and generating local outflow. Unlike channel flows in humid areas, runoff from
the upper valley has to meet the storage demands of the lower valley segments, often causing seepage loss along flow paths to
render the flows intermittent. A fill-and-spill runoff system is proposed in which the valley physiography results in a series of
segments with varying storage conditions. As water is supplied to the valley, each segment has to be filled until its storage
threshold for runoff is exceeded. Then, subsurface or surface flows will be generated, but these flows may be arrested to furnish
water to satisfy the storage requirements of the segments downstream. Such a flow system applies also to other valleys in the
Shield environment.
q 2003 Elsevier B.V. All rights reserved.
Keywords: Runoff generation; Canadian shield; Northwest territories; Water budget; Storage; Subarctic
1. Introduction
Headwater basins in the Canadian Shield comprise
a mosaic of exposed Precambrian bedrock uplands
and soil-filled valley bottoms. Upon reaching the
valleys, runoff produced on the uplands is modified by
the storage and flow delivery mechanisms in the soil-
filled zone (Allan and Roulet, 1994; Buttle and Sami,
1992; Peters et al., 1995). There is an insufficient
understanding of the relationship between catchment
water balance, soil-zone storage and hydrological
linkages to account for the wide variations in the
runoff ratios reported for headwater areas (Branfireun
and Roulet, 1998; Landals and Gill, 1972). Further-
more, the subarctic portion of the Canadian Shield is
subject to intense winter coldness and ground frost
may add hydrological complications (Metcalfe and
Buttle, 2001; Thorne et al., 1994; Wright, 1979). It is
the purpose of this research to study the hydrological
processes in a subarctic Canadian Shield headwater
soil-filled valley and to elucidate the hydrological
0022-1694/03/$ - see front matter q 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0022-1694(03)00175-6
Journal of Hydrology 279 (2003) 151–166
www.elsevier.com/locate/jhydrol
* Corresponding author. Address: Environment Canada, #301
5204 50th Avenue, Yellowknife, Northwest Territories, Canada
X1A 1E2. Fax: þ1-867-873-8185.
E-mail address: [email protected] (C. Spence).
connections in terms of water transfer from the
bedrock uplands and through the soil-filled valleys.
An improved understanding of the headwater hydrol-
ogy will permit realistic representation of the hydro-
logical processes in the modelling of shield
hydrology, and, given the considerable extent of
Precambrian bedrock landscape in Canada and in
northern Europe, have application for many northern
areas.
2. Study site
Pocket Lake basin, located 4 km north of the City
of Yellowknife in Canada’s Northwest Territories
(Fig. 1), has been studied intermittently since the
1970 s when it was identified as a field site as part of
Canada’s contribution to the International Hydrolo-
gical Decade. The physical setting of the area has
been described by Landals and Gill (1972), Reid
(1997) and Spence and Woo (2002). The present study
focuses on a soil-filled valley bottom of a headwater
basin downslope of the bedrock runoff plots investi-
gated by Spence and Woo (2002). The soil-filled
valley is 10,875 m2 in size. A surrounding exposed
bedrock upland 37,948 m2 in area drains into the
valley. Open woodlands of black spruce (Picea
mariana) grow at the edge of the valley along the
soil/bedrock border. Understory vegetation includes
dwarf willow (Salix spp.), Labrador tea (Ledum
Fig. 1. Pocket Lake soil filled valley with location map and inset picture. The valley is outlined in grey and the entire basin in black. Pocket Lake
is at the north end of the valley and is seen as the dark area at the bottom of the picture. Dotted arrows show direction of surface flow. Black lines
with letter labels denote well and piezometer transects. White squares are ablation lines. Stars show locations of bedrock runoff plots. Triangles
are trench locations. Weirs are represented by white boxes with a V notch. The rectangle is the soil/bedrock contact instrumentation. The white
circle is a thermistor string and the black circle is a climate tower.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166152
groenlandicum) and rose (Rosa spp.) bushes. Ground
cover includes moss (Sphagnum spp.), lichen (Cla-
donia spp.) grass (Eriophorum spp.) and sedges
(Carex spp.). The upper part of the valley has a
gradient of 1%, while the lower section descends at
9%. The substrate consists of three distinct layers each
composed of organic soils, silty sands and a basal
layer of cobbles and boulders (Table 1) which pinches
out towards the valley sides. The bedrock hillslopes
rise abruptly from the soil-filled valley floor. Beneath
the soil, a bedrock sill stretches across the lower reach
of the valley. Down valley from the sill the bedrock
and ground surfaces have similar gradients.
The 1961–1990 climate at Yellowknife is charac-
terised by short cool summers with a July daily
average temperature of 16 8C and long cold winters
with a January daily average temperature of 229 8C.
Annual unadjusted precipitation for the same period
averages about 280 mm, with 55% of that falling as
snow (Wedel et al., 1990). Convective cells produce
much of the summer precipitation and so summer
rainfall is quite variable from year to year. As the jet
stream settles over the region in September, con-
ditions become cool and damp. If this shift begins
early, the annual precipitation will be higher than
normal due to the extended wet period.
3. Methods
Geophysical surveys were conducted to define the
surface and bedrock topography of the study site. The
bedrock surface beneath the soil-filled valley was
mapped using a combination of probing by hammer-
ing a steel rod into the sediment until encountering the
bedrock and a pulseEKKO IV ground penetrating
radar with a 400 V transmitter and an antenna centre
frequency of 100 MHz (Spence, 1996). Soil tempera-
ture was recorded at half hourly intervals using a
string of Campbell Scientific 107B thermistors
(accuracy ^ 0.4 8C) located at the foot of the hillslope
(Fig. 1). A steel rod was used to probe for the depth of
the frost table during the spring and early summer.
Daily valley water budget terms, including rainfall,
snowmelt, evapotranspiration, inflow from the bed-
rock upland, surface and subsurface runoff and change
in storage, were obtained between 10 May 2000 and
16 May 2001 with the instrumentation shown in Fig.
1. Here, magnitudes of all water budget components
are expressed as depth per unit valley-area, except
when noted. A meteorological tower on a bedrock
ridge above the valley was equipped with a Meteor-
ological Service of Canada Type B rain gauge to
measure rainfall (accuracy 5%). Rainfall intensity was
measured at the same tower using a tipping-bucket
rain gauge with signals recorded by a Campbell
Scientific CR10X datalogger. The snow water equiv-
alent of the spring snowpack was calculated from
snow density measured with an Eastern Snow
Conference snow sampler and snow depth measured
with an aluminium rod along snow survey transects,
following the method described by Pomeroy and Gray
(1995) (accuracy 15%). The snow survey included
five random survey lines each of which included at
least five snow density and twenty snow depth
samples. Daily snowmelt was calculated by measur-
ing the lowering of the snow surface and the snow
density of the surface snow layer, along four ablation
lines on the bedrock upland and two in the valley,
using the technique described by Heron and Woo
(1978) (accuracy 25%). Infiltration rates were deter-
mined under different ground frost conditions using
double-ring infiltrometers. Evapotranspiration was
estimated using the eddy correlation energy budget
techniques and instrumentation described by Spence
and Rouse (2002) who report an accuracy of 20% with
these methods. Lateral inflow from upslope exposed
bedrock was measured with volumetric measurements
and velocity-area flow calculations at the upslope weir
sites (Fig. 1), with velocity obtained using a Price type
pygmy current meter. These weirs captured runoff
from 88% of the bedrock upland. The error associated
with these runoff measurements is estimated con-
servatively at 20%. It was assumed that additional
inflow from bedrock side slopes outside areas
Table 1
Hydrological characteristics of the soil layers in the headwater
valley
Soil type Depths
(m)
Porosity Hydraulic
conductivity (m/d)
Organic 0–0.35 0.75 10
Silty sands 0.35–1.5 0.5 1
Cobbles and boulders 1.5–1.75 0.35 10
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 153
captured by the weirs could be estimated using runoff
data from the bedrock runoff plots described in
Spence and Woo (2002) because of their similar
drainage areas and physiography. The plots were
reported to measure runoff within 7%. Spence and
Woo (2002) classify their plots into bare and soil
covered. The relative areal extent of each across the
side slopes was delineated from air photos and side
slope inflow calculated as a portion of total lateral
inflow following:
Rbss ¼ðRbabÞ þ ðRscascÞ
abss
ð1Þ
where a is an area from which runoff, R; is generated.
The subscripts b, sc, and bss represent bare plots, soil
covered plots and bedrock side slopes, respectively.
Rb and Rsc are expressed in depth per unit runoff plot-
area and Rbss is expressed in depth per unit area of the
side slopes. Rbss was converted into millimetres over
the area of the valley, av; using:
Ibss ¼Rbssabss
av
ð2Þ
and added to Ibw; inflow measured at the weirs, in
runoff/unit area of the valley, to produce total lateral
inflow, I:
I ¼ Ibw þ Ibss ð3Þ
Peters et al. (1995) identified that runoff from
exposed bedrock enters soil along the bedrock
surface. At the bottom of a bedrock hillslope, daily
volumetric measurements were made at a runoff plot
that consisted of paired weirs, one at the soil-bedrock
contact and the other at the bedrock surface. The
partitioning of runoff into surface flow and flow along
the bedrock surface during both frozen and unfrozen
conditions at this location was assumed to occur
everywhere along the soil-filled valley bedrock
contact. At the outlet of the basin, stage recorded
continuously at a 908 V-notch weir was converted into
discharge by a rating curve obtained from periodic
discharge measurements using the velocity-area
method (accuracy similar to the inflow weirs at
20%). Subsurface flow was collected in two trenches
(accuracy 10%). They were operative only in the
summer as flooding prevented their usage during the
spring. Areas contributing to surface runoff were
mapped based on visible observations. A network of
piezometers along three transects enabled the deter-
mination of the direction of groundwater movement
within the soil-filled valley. Hydraulic conductivity
was calculated using pump tests as described in
Freeze and Cherry (1979).
The water table was measured continuously at two
wells, one at the edge and the other in the centre of the
valley along transect H (Fig. 1). Six additional wells
along transect G, three along transect H and two along
transect E were measured opportunistically. Field
calibrated Campbell Scientific CS615 time domain
reflectometry (TDR) soil moisture sensors were
installed at 0.05 and 0.3 m depths within the middle
of the soil-filled valley and to 0.3 m depth at the valley
edge. Daily change in storage, DS; at each point was
calculated as:
DS ¼ DSu þ DSs
¼ Du½z 2 zwðtÞ� þ Sy½zwðtÞ2 zwðt 2 1Þ� ð4Þ
where DSu and DSs are the changes in soil moisture
fraction in the unsaturated and saturated zones; Du is
change in soil moisture fraction in the unsaturated
zone; Sy is the specific yield of soil, measured here to
be 0.13; z is total soil thickness, zwðtÞ and zwðt 2 1Þ are
height of the water table (measured from the bedrock
upward) for the present ðtÞ and the previous ðt 2 1Þ
time periods. The valley edge was delineated using
the strip of trees visible in Fig. 1. The average valley
change in storage was a prorated calculation based on
the relative areas of valley edge and centre. The
change in storage calculation is expected to have an
accuracy of 25% based on the error of the calibration
of the TDR sensors and the estimate of Sy:
4. Hydrological processes in the soil-filled valley
4.1. Ground frost
The soil-filled valley has only seasonal frost.
Ground thaw proceeded evenly during the spring
and early summer of 2000, with rates that varied
within 20% of the average value, close to the accuracy
of the frost probe measurements. Thaw rates between
29 March and 1 June 2000 averaged 5 mm/d and
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166154
increased to 15 mm/d after 1 June. The frozen layer
disappeared between mid and late June. Surface soil
temperature reached a maximum in June and July, but
the entire column was warmest at the end of August.
The soil was frozen to approximately 1 m depth by the
spring of 2001, but periodic measurements indicated
that ground thaw was negligible during the snowmelt
period.
4.2. Snowmelt
A snow survey on 1 April 2001 measured
160 ^ 25 mm of snow water equivalent in the valley.
Ablation rates between 2 April and 19 April 2001
averaged 2.7 mm/d. Latent heat flux estimated from
tower measurements suggested that an average of
0.5 mm/d of the ablation during this period was
directed to the atmosphere. Meltwater produced in
these two weeks was not observed to runoff or enter
soil storage so it was either refrozen within the
snowpack or in surface depressions as a cold snap
ensued until 23 April 2001. After 24 April 2001
above-freezing air temperatures ripened the entire
snow cover and snowmelt increased to a peak of
16.9 mm on 5 May 2001. The rate of losses to
the atmosphere estimated from tower measurements
increased to an average of 1.9 mm/d after 23 April
2001. The entire snowpack in the valley was gone by
10 May 2001. Total valley snowmelt equalled
117 mm with the remainder, 49 mm, directed to the
atmosphere.
4.3. Rainfall
Rainfall, assumed uniform over the entire basin, in
the summer of 2000 totalled 154 mm (Fig. 2), almost
half of which fell between 14 August and 29 August.
Much of the remaining amount came from three
thunderstorms on 20 and 23 June and 14 July. While
May, July and September were drier than normal, wet
conditions dominated June and August. With an
average May to September rainfall of 141 mm,
Yellowknife is drier than the other Canadian Shield
locales where hydrological studies were undertaken
(e.g. 321 mm at Thompson, Manitoba, 360 mm at the
Experimental Lakes Area in northern Ontario,
358 mm at Schefferville, Quebec and 444 mm at
Muskoka, Ontario) (Metcalfe and Buttle, 2001;
Wright, 1979; Thorne et al., 1994; McDonnell and
Taylor, 1987).
Fig. 2. 10 May–25 September 2000 cumulative rainfall time series at Pocket Lake.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 155
4.4. Infiltration
Double-ring infiltrometer measurements yielded
infiltration rates of about 1 m/d regardless of whether
the soil was frozen or otherwise. This is attributed to
the frozen but unsaturated conditions of the porous
soils prior to spring melt. Such non-limiting infiltra-
tion capacities for frozen soil (Gray et al., 2001)
allowed snowmelt and upslope runoff to percolate
through the frozen layer. Shallow soil moisture
measurements show that despite the ability of the
soil to accept infiltration, a large fraction of rainfall
(,0.6) could be intercepted by the ground vegetation
mat of lichen and moss because of its dryness (Bello
and Arama, 1980) during much of the summer. A lack
of soil moisture response after small (,12 mm)
rainfall events at 0.3 m depth implies that this
intercepted water was directed back to the
atmosphere.
4.5. Inflow from exposed bedrock upland
Summer inflow from the upland was intermittent
(Fig. 3). All inflow entered the valley along the soil-
bedrock interface and this is in agreement with
Peters et al.’s (1995) observation near Muskoka that
most lateral inflow from upslope exposed bedrock
enters Canadian Shield soil zones along the bedrock
surface. In the spring, inflow also travelled along the
bedrock surface as the frozen soil did not have much
ice to seal its pores. Spring inflow at the contact plot
began on 28 April 2001. A lag time between peak
bedrock snowmelt on 18 April and peak lateral
inflow to the valley on 3 May was mostly due to two
periods of sub-freezing temperatures during which
the meltwater was refrozen in the snow or in shallow
depressions on the bedrock surface.
4.6. Evapotranspiration
Between May and July 2000, evapotranspiration
from the soil-filled valley averaged 1.8 mm/d. Evapo-
transpiration exceeded rainfall in May, July and
September and almost equalled the June precipitation.
August was the only month when rainfall exceeded
evapotranspiration, as cool and wet conditions
reduced the evapotranspiration rate to 0.8 mm/d.
The similarity in calculated evapotranspiration and
change in storage during a dry period between 6 and
17 July 2001 shows that in the absence of rainfall
Fig. 3. 10 May–25 September 2000 daily lateral inflow from exposed bedrock to the soil filled valley.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166156
input, evapotranspiration was sustained by moisture
storage in the soil (Fig. 4). This is in agreement with
observations at other subarctic Canadian Shield sites
(Spence and Rouse, 2002).
4.7. Storage
Soil moisture in the unsaturated zone at the valley
edge responded more readily to rainfall events than at
Fig. 4. Cumulative daily evapotranspiration and change in storage during a dry period in the middle of the 2000 growing season.
Fig. 5. Daily soil moisture measurements at different locations and depths in the soil filled valley for July through September 2000.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 157
the centre (see smaller peaks in Fig. 5). Valley edge
soil moisture always increased more than observed
rainfall because it was often augmented by lateral
inflow. In contrast, increases to shallow valley centre
soil moisture averaged only 60% of rainfall inputs. At
0.3 m depth in the centre of the valley, the soil
moisture content reacted only to the largest rainfall
events (.12 mm).
As the summer of 2000 progressed, evapotran-
spiration loss exceeded rainfall and lateral inflow,
leading to a depletion of soil moisture storage. During
dry conditions the water table cross sections across the
valley were flatter than during wet conditions (Fig. 6).
A large rainfall event beginning 23 August 2000
(Fig. 2) generated the largest daily lateral inflow from
the bedrock upland during the summer of 2000 (Fig. 3)
which caused distinct spatial differences in valley
water table response. The water table rose more
rapidly at the sides of the valley and peaked 2 h before
the centre because lateral inflow was quicker than
rainfall percolation, much of the latter also being
intercepted by the ground mat of lichen and moss
(Fig. 7). This same event also caused uneven rises in
the water table along the valley. By 27 August 2000
the water table at transect H rose an average of 0.32 m
but the water table at transect G rose by only 0.1 m
because of higher lateral inflow inputs close to
transect H.
After 29 August, the water table declined, first at
the sides, then in the middle. Throughout the winter of
2000–01, groundwater possibly drained through the
fractures of the bedrock underneath the soil-filled
valley. Immediately before snowmelt (end of March),
the water table profile across the valley was flat and
appeared similar to the dry summer condition. The
first rise of the water table (0.6 m) occurred on 18
April in the middle of the valley, indicating that the
water source was from snowmelt in the valley and
the meltwater was able to infiltrate the frozen soil. The
water table at the sides of the valley did not rise until
28 April when lateral flow began in earnest. Estimates
of daily snowmelt, rainfall and lateral inflow show
that before 28 April, 70% of these inputs entered
valley storage and the remainder was lost to the
atmosphere. The water table reached the topographic
surface on April 29 and remained there until the end
of the study period.
4.8. Subsurface runoff
Pumping tests and direct measurement of flow at
the trenches indicate a hydraulic conductivity of
approximately 1 m/d for the soil, regardless of its
thermal state. Piezometric measurements across
transects G and H showed that water from the bedrock
upland consistently drained towards the centre of
the valley and then down the valley. The presence of
Fig. 6. The water table across and along the soil filled valley during
wet (27 August 2000) and dry (16 August 2000) conditions. The
locations of other transects when crossed are referenced on each
cross section. Information on transect H only covers the western 16
m of its length. The white circles denote locations of piezometers or
wells. The locations of the transect H wells with data illustrated in
Fig. 7 are noted. Refer to Fig. 1 for locations of transects.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166158
a bedrock sill in the lower valley (Fig. 6) restricted
subsurface runoff for much of the summer, a
situation similarly reported by Allan and Roulet
(1994) at another Canadian Shield watershed in
Ontario. Subsurface runoff was first observed at the
lower trench when the water table rose above the
sill on 26 August 2000 in response to 40 mm of
rainfall, and continued until 25 September when the
water table dropped below the elevation of the sill
(Fig. 8).
4.9. Surface runoff
A large storm event generated 58 mm of rainfall
and 66 mm of lateral inflow between 14 August and
27 August. This triggered saturation overland flow at
0800 on 27 August as the water table rose above
ground at the western valley edge at transect
H. Surface runoff followed the valley thalweg, along
a water track similar to that described by McNamara
et al. (1998). A rapid rise in near surface soil moisture
0.5 m downslope of transect G on transect E indicated
that this flow reached that position between 0900 and
0930 (Figs. 5 and 8). The piezometric heads along and
adjacent to the flow path showed that part of the
surface runoff infiltrated the soil as it travelled
downstream, suggesting the influent nature of this
intermittent stream. The cessation of rainfall and a
reduction of inflow from the bedrock upland caused
the water table to recede on 29 August, but the loss to
stream influence continued. The stream-flow segment
retreated up-valley until surface flow ceased
altogether. The overall effect of the surface flow
process is therefore the formation of an intermittent
stream that expanded from and contracted to the
source of saturation overland flow near transect H.
Runoff in the spring of 2001 followed two phases.
Despite 93 mm of precipitation, snowmelt and inflow,
only 8 mm of surface runoff was generated by 1 May.
Most of the spring runoff (396 mm) from the 1.1 ha
valley was produced after the cold spells in early May
and was fed by 377 mm of inflow from the 3.8 ha
bedrock upland (Fig. 9). The large lateral inflow
volume is a result of the relative size difference
between the larger upland and smaller valley (Eq. (2)).
The concordance of daily flow rates between the
surface inflows and surface outflows may imply that
the soil-filled valley served mainly as a conduit for
runoff without significantly altering the flows through
soil storage. However, as the frozen ground had little
effect on infiltration and soil hydraulic conductivity,
inflow from the upland could enter the saturated soil-
filled valley after 29 April, to mix with the water
already residing in the valley soils. An analysis of
Fig. 7. Half hourly measurements of water table at the valley edge and valley centre.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 159
water collected during previous spring melt events at
Pocket Lake [from a study by and using analytical
methods described in Gibson et al. (1998)] revealed
that surface runoff was a chemical mixture of
snowmelt event water and pre-event groundwater.
Table 2 shows that ground ice and groundwater from
the soil-filled valley were relatively enriched with
oxygen-18 (18O) and deuterium (2H) and had
Fig. 8. Inflow from exposed bedrock, the response of soil moisture at the centre of the soil filled valley at transect G and runoff during the 23
August rainfall event.
Fig. 9. 2001 Spring melt cumulative water budget.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166160
significantly higher electrical conductivity ðCÞ when
compared to snow. Surface runoff had isotopic and
chemical values between those of the snow and
ground ice and groundwater, suggesting that at least a
part of the valley contributes its pre-event water to the
spring runoff.
The sources that dictated the timing and volume of
runoff differed even though similar runoff generation
processes occurred in the spring and summer events.
In the spring, over half the valley snowpack was
directed to soil moisture storage and evaporation, so
the timing and magnitude of inflow from the bedrock
upland controlled the outflow hydrograph of the soil-
filled valley. In contrast, there was insignificant
vertical input to replenish the summer storage deficit
which then consumed much of the lateral inflow. This
resulted in a significant delay between lateral inflow
and valley outflow, a phenomenon not exhibited
during the spring melt.
5. Hydrological linkages between upland
and valley
The hydrological importance of the bedrock
uplands to the soil-filled valley is demonstrated by
the water balance considerations for the snowmelt and
the summer periods. In the spring (2 April–16 May)
of 2001, the valley received 117 mm from local
snowmelt, 53 mm of direct rainfall, but an inflow of
390 mm from the upland. This permitted 50 mm of
evaporation and 404 mm of outflow to be generated,
and a calculated increase of soil storage of 106 mm
(cf. measured storage increase was 108 mm) (Table 3).
In the summer (May–September) of 2000, direct
rainfall on the valley was 155 mm and inflow from
the uplands was 106 mm. These amounts sustained an
evapotranspiration loss of 164 mm and an outflow of
34 mm, as well as recharged the soil storage by
64 mm (cf. measured storage change was 73 mm)
(Table 3 and Fig. 10). These values show that inflow
from the uplands constitute a significant portion of
total inputs to the valley.
In the snowmelt season, as exemplified by the
spring of 2001, the bedrock upland snow cover melted
and provided the first contributions to surface runoff in
the basin (Fig. 11). At the same time, melting of the
valley snow cover percolated the dry frozen soil and
replenished storage. A rise of the water table at the
sides of the valley first occurred when many parts of
the upland contributed meltwater in support of the peak
inflow. Valley margins were the only sections of the
valley that produced surface runoff. Initial saturation
overland flow from these sites then followed water
tracks along the valley but it was subject to infiltration
losses, so that there was insufficient flow to reach the
valley outlet. Both the eastern and southern portions of
the valley eventually contributed to outflow, but as
inflows from the upland decreased, these intermittent
streams receded back to the up-valley areas.
The summer of 2000 was mostly dry. Rainfall and
inflow from the uplands were needed to replenish the
soil storage before runoff could begin in the valley.
Little rainfall reached the soil because of interception
loss to the ground vegetation. As inflow from the
upland was along the bedrock-soil interface, it could
avoid interception losses to the valley vegetation and
Table 2
Average values of selected chemical characteristics of water from
the Pocket basin site from 1995 to 2000. Stable isotope values are
presented in standard d notation as deviations per mille from
Vienna—SMOW (standard mean ocean water) such that dsample ¼
1000{ðRsample=RSMOWÞ2 1� where R is 2H/1H and 18O/16O
18O 2H C (mS/cm)
Snow 228 2215 7
Snowmelt runoff 224 2191 106
Ground ice/groundwater 220 2162 330
Table 3
2000 growing season and 2001 spring melt water budgets. All units
are in mm. (Qs and Qg are surface and subsurface outflow, M is
snowmelt, P is precipitation, E is evapotranspiration, I is lateral
inflow and DS is change in storage). The p included with Qg for the
2001 spring melt denotes unavailable because of flooding in the
trenches
Month Qs Qg M P E I DS DS
(calc.)
May 2000 0 0 3 29 1 211 226
June 2000 0 0 52 50 13 19 17
July 2000 0 0 25 40 7 252 28
August 2000 12 6 66 26 74 163 96
September 2000 2 14 9 19 11 245 215
2000 summer 14 20 155 164 106 74 64
2001 spring 404 p 117 53 50 390 108 106
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 161
could replenish soil storage directly. Thus, it was
lateral inflow rather than direct rainfall that controlled
the rise of the water table and the occurrence of
saturation overland flow in the soil-filled valley.
This was evidenced by the storm event of 23 August
2000. At the beginning of the rainstorm, the upland
bedrock areas generated runoff first, as their low
storage capacities were quickly satisfied (Fig. 12).
The valley section contributed runoff only after lateral
inputs enabled saturation at the upper valley and along
the water track. The eastern arm of the valley did not
yield surface runoff during this storm because lateral
inputs from exposed bedrock were not as high relative
to available storage. This attests to the importance of
upland contribution in runoff generation from the soil-
filled valley and confirms that the lower reaches of the
valley acted primarily as a conduit for surface runoff.
6. Hydrological behaviour of the soil-filled valley
Soil-filled valleys are an integral part of the Shield
landscape, performing the triple functions of (1)
Fig. 10. Cumulative water budget of the soil filled valley over the summer of 2000.
Fig. 11. Map of areas contributing surface runoff during the 2001
spring melt. Each shade represents the only areas contributing to
surface runoff on that date.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166162
collecting vertical and lateral water inputs, (2)
retaining and losing the water held in storage, and
(3) transferring water down the valley and generating
local outflow. The interactions among these functions
show that inputs to the valley must fill up the storage
requirement before runoff is spilled to areas down-
stream. Conceptually, valley hydrology then can be
examined in terms of water sources, storage and
spillage processes.
6.1. Water sources, storage and runoff
In view of the smaller surface area of the valley
relative to the adjacent upland catchment, vertical
inputs (rainfall and snowmelt) are small in compari-
son to the lateral inflows. The latter are generated
from the upland which represent a dynamic contribut-
ing zone, the hydrological behaviour of which is the
focus of a future paper.
Available storage, being a function of antecedent
moisture and evapotranspiration loss, determines the
minimum input required to exceed the thresholds for
subsurface and surface flow generation. Variable soil
depths caused by bedrock topography generally offer
the valley edges a shallow soil that has a lower
saturation threshold than the middle of the valley.
The proximity of the edge zones to the bedrock slopes
increases the opportunity for large lateral inflow and
greater likelihood to reach storage capacity than the
other parts of the valley.
Soil thickness and its spatial uniformity is
important for runoff delivery because it determines
whether shield hillslopes remain hydrologically
coupled through the dry periods (Buttle et al., 2000;
Branfireun and Roulet, 1998; Devito et al., 1996).
Both Dorset (Buttle and Peters, 1997; Peters et al.,
1995) and the Experimental Lakes Area (Branfireun
and Roulet, 1998) in the humid temperate zone of
Ontario have similar soil depths and catchment areas
as those at Pocket Lake, yet experience seasonal
runoff ratios twice those at our study basin. The
difference is attributed to the precipitation amount,
with the Ontario sites receiving two and a half to three
times the summer rainfall at Yellowknife. Buttle et al.
(2000) suggested that deeper soils in a shield
environment permit more storage, better hydrologic
connections and, in turn, higher runoff. This does not
apply to the dry landscape around Yellowknife, where
high available storage in deep soils requires a larger
amount of water input to saturate. The flow linkage is
affected by water availability and soil storage status
and the presence of thick soil does not necessarily
imply that hydrological connections are intact.
Flow generation is a function of where, when and
how the soil storage capacities are satisfied. Subsur-
face flow often follows the bedrock surface (Peters
et al., 1995), moves within the soil matrix (Devito et al.,
1996) or drains along macropores, but the saturated
zone is unevenly distributed in the valley and the water
table has to rise above any bedrock sills to generate
subsurface outflow. Saturation overland flow is the
dominant surface runoff mechanism in shield valleys
(Buttle et al., 2000) and flow production depends on
local saturation of the valley soil. The 2001 spring melt
runoff ratio was 0.4 higher than the August 2000 rain
event. This demonstrates the variability of flow
generation due to changing spatial storage demands
in the valley. There is also a feedback between
overland flow and valley soil recharge. Runoff
from an upper valley segment may encounter
Fig. 12. Map of areas contributing surface runoff during the August
2000 rain event. Unlike Fig. 11 the shading indicates the cumulative
expansion of the contributing area.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 163
a non-saturated lower segment and the water will
infiltrate along the flow path until the lower valley
storage is satisfied or until all surface flow is lost to
seepage. In the latter case, the stream becomes
intermittent.
6.2. The fill and spill flow concept of runoff generation
Central to the fill-and-spill concept is the spatially
variable valley storage that needs to be satisfied before
water spills to generate either surface or subsurface
flow. Storage capacity in the valley is spatially variable
owing to topographical diversity, soil heterogeneity
and uneven thickness, and seasonal presence of ground
frost. Soil storage status in the valley, enriched by
rainfall, snowmelt and lateral inflow, but lost to
evaporation and downstream drainage, is temporally
dynamic. Thus, along segments of a soil-filled valley,
filling will continue until (1) the local storage spills over
the threshold created by any bedrock sills to permit
subsurface flow, and (2) the water table rises above the
topographic surface to generate overland flow.
The fill-and-spill flow system differs from the
normal modes of flow in a humid climate where
the channel flow in the valleys is permitted to leave
the catchments with little significant interruption. The
importance of lateral inflows and variable storage
present a system that departs greatly from typical
saturation overland flow processes in humid areas
such that contributing areas do not necessarily grow
upslope from the stream channel. The shield valley
represents a series of storage reservoirs with the flow
Fig 13. An illustration of conceptualised fill-and-spill runoff generation. A is a longitudinal profile of the valley and B is a cross section. P is
precipitation; t is time at step 1, 2 or 3. SSSF is subsurface stormflow and SOF is saturation overland flow. A is the contributing area at t2 or t3:
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166164
cascading down the valley, filling individual segments
to satisfy their deficits until the thresholds are reached;
then spillage resumes to continue the flow down-
stream. Fig. 13 illustrates this idea. At t1; rain begins
and the valley water table is below the topographic
surface. By t2; lateral inflow has entered from bedrock
uplands, prompting a water table rise to the topo-
graphic surface and saturation overland flow at the
valley sides. Storage demands downslope interrupt the
saturation overland flow before it can reach the valley
outlet. As event inputs continue, the storage reservoirs
in the valley between the sides and outlet saturate so
that by t3 an ephemeral stream of saturation overland
flow reaches the outlet and discharges runoff.
6.3. Upscaling of the flow system
The fill-and-spill flow system can also be applied
to the large catchments of the Canadian Shield. For
ephemerally and perennially draining wetlands, Buttle
and Sami (1992) and Branfireun and Roulet (1998)
noted a progression of water table rise down valley
during runoff events, indicating that each segment is
first filled by lateral inflows before saturation overland
flow can continue downstream. During the summer in
a southern Ontario swamp, Devito et al. (1996)
observed that lateral surface inflows exceeded surface
outflow, suggesting that some of the surface flow
seeped underground to address storage demands. Of
particular importance is the frequent presence of lakes
in the Shield that through their well known role in
flow regulation functions, exaggerates the storage and
release functions of the hydrologic system as has been
reported by Fitzgibbon and Dunne (1981) and Spence
(2000). Such mechanisms of fill-and-spill therefore
can be upscaled and we recommend that hydrological
modelling of shield catchments should take account of
the cascading behaviour of runoff and its interaction
with in-valley storage.
7. Conclusions
In this research we examined the hydrological
processes in a subarctic Canadian Shield headwater
soil-filled valley and clarified the hydrological con-
nections in terms of water transfer from the bedrock
uplands to and through the soil-filled valleys.
Evapotranspiration was a key controller of storage
amount in the valley. The importance of evapotran-
spiration has not been clearly identified in the
previous studies on the Canadian Shield that have
focused on the wetter environments in Manitoba,
Ontario and Quebec. Lateral inflows at this site were
also a crucial component of the valley water budget.
Our results demonstrate that water budgets at down-
slope positions may be controlled by upslope
locations where the latter’s contributing areas are
large enough. It was previously known that the state of
the hydrologic connection between uplands and
valleys influences basin runoff magnitude, but the
controls on this connection are now quantified.
Results of this study showed that the status of
available storage at the valley edge relative to the
magnitude of lateral inflow controls this connection
and it has significant implications on how water is
transferred down the valley. Valley edges tend to have
lower storage capacities because of shallower soils
and the highest inputs from their adjacent lateral
inflows, causing saturation overland flow to begin at
the upper or edge locations in valleys. Some of this
saturation overland flow would be lost to infiltration at
downstream locations until storage capacities are
satisfied. Such hydrological linkage functions of the
headwater valleys can be conceptualised as a fill-and-
spill runoff mechanism in which segments of the
valley represent reservoirs that have to be filled above
their capacity before spillage allows flow to continue
downstream. Such mechanisms can be upscaled to
large catchments and its representation should
improve the performance of hydrological models of
the shield environment.
Acknowledgements
We thank Shawne Kokelj, Andrea Czarnecki,
Kerry Walsh and Mark Dahl of Environment Canada,
and Claire Oswald of McMaster University and Iain
Stewart and Derek Steadman for their assistance in the
field. Environment Canada and the Mackenzie
GEWEX Study funded this work. Miramar Giant
Mine kindly provided access to the study site. We
acknowledge Shauna Sigurdson and Jesse Jasper of
Environment Canada for supporting this research.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166 165
References
Allan, C., Roulet, N., 1994. Runoff generation in zero order
Precambrian shield catchments: the stormflow response of a
heterogeneous landscape. Hydrological Processes 8, 369–388.
Bello, R., Arama, A., 1980. Rainfall interception in lichen canopies.
Climatological Bulletin 23, 74–78.
Branfireun, B., Roulet, N., 1998. The baseflow and storm flow
hydrology of a Precambrian shield headwater peatland. Hydro-
logical Processes 12, 57–72.
Buttle, J.M., Sami, K., 1992. Testing the groundwater ridging
hypothesis of streamflow generation during snowmelt in a
forested catchment. Journal of Hydrology 135, 53–72.
Buttle, J.M., Peters, D., 1997. Inferring hydrological processes in a
temperate basin using isotopic and geochemical hydrograph
separation: a re-evaluation. Hydrological Processes 11,
557–573.
Buttle, J.M., Creed, I.F., Pomeroy, J.W., 1998. Advances in
Canadian forest hydrology, 1995–1998. Hydrological Processes
14, 1551–1578.
Devito, K., Hill, A., Roulet, N., 1996. Groundwater surface water
interactions in headwater forested wetlands of the Canadian
Shield. Journal of Hydrology 181, 127–147.
Fitzgibbon, J., Dunne, T., 1981. Land surface and lake storage
during snowmelt runoff in a subarctic drainage system. Arctic
and Alpine Research 13, 277–285.
Freeze, R.A., Cherry, J.A., 1979. Groundwater, Prentice Hall,
Englewood Cliffs.
Gibson, J., Reid, R., Spence, C., 1998. A six year isotopic record of
lake evaporation at a mine site in the Canadian subarctic: results
and validation. Hydrological Processes 12, 1779–1792.
Gray, D.M., Toth, B., Zhao, L., Pomeroy, J.W., Granger, R.J., 2001.
Estimating areal snowmelt infiltration into frozen soils. Hydro-
logical Processes 15, 3095–3111.
Heron, R., Woo, M.K., 1978. Snowmelt computation for a high
arctic site. Proceedings of the 35th Eastern Snow Conference,
Hanover, New Hampshire, 162–172.
Landals, A., Gill, D., 1972. Differences in volume of surface runoff
during the snowmelt period: Yellowknife, NWT, In: The Role
of Snow and Ice in Hydrology, IAHS Publication #107, pp.
927–942.
McDonnell, J., Taylor, C., 1987. Surface and subsurface water
contributions during snowmelt in a small Precambrian shield
watershed Muskoka, Ontario. Atmosphere Ocean 25,
251–266.
McNamara, J.P., Kane, D.L., Hinzman, L.D., 1998. An analysis of
streamflow hydrology in Kuparuk River Basin, arctic Alaska: a
nested watershed approach. Journal of Hydrology 206, 39–57.
Metcalfe, R.A., Buttle, J.M., 2001. Soil partitioning and surface
store controls on spring runoff from a boreal forest peatland
basin in north-central Manitoba, Canada. Hydrological Pro-
cesses 15, 2305–2324.
Peters, D., Buttle, J.M., Taylor, C., LaZerte, B., 1995. Runoff
production in a forested shallow soil Canadian Shield basin.
Water Resources Research 31, 1291–1304.
Pomeroy, J.W., Gray, D.M., 1995. Snowcover: Accumulative,
Relocation and Management. National Hydrology Research
Institute Science Report No. 7. Environment Canada, Saskatoon.
Reid, R., 1997. Evaporation studies at mine tailings ponds in the
Northwest Territories. In: Milburn, D., (Ed.), Proceedings of the
Hydro-ecology Workshop on the Arctic Environmental Strat-
egy—Action on Water. May, 1996, Banff, National Hydrology
Research Institute Symposium #16, National Hydrology
Research Institute, Saskatoon, Canada, pp. 115–133.
Spence, C., 1996. Ground penetrating radar as a tool in an
investigation of shallow groundwater flow near Yellowknife,
NWT, internal report, Water Resources Division, Indian and
Northern Affairs Canada, Yellowknife.
Spence, C., 2000. The effect of storage on runoff from a headwater
subarctic shield basin. Arctic 53, 237–247.
Spence, C., Rouse, W.R., 2002. The energy budget of subarctic
Canadian Shield terrain and its impact on hillslope hydrological
processes. Journal of Hydrometeorology 3, 208–218.
Spence, C., Woo, M.K., 2002. Hydrology of subarctic Canadian
Shield: Bedrock upland. Journal of Hydrology 262, 111–127.
Thorne, G., Laporte, J., Clarke, D., 1994. Infiltration and recharge in
granitic terrains of the Canadian Shield during winter periods.
Proceedings, 10th International Northern Research Basins
Symposium and Workshop, Spitsbergen, Norway, pp. 449–466.
Wedel, J.H., Smart, A., Squires, P., 1990. An overview study of the
Yellowknife River Basin, NWT, Inland Waters Directorate,
Environment Canada,.
Wright, R., 1979. Preliminary results of a study on active layer
hydrology in the discontinuous zone at Schefferville, Nouveau-
Quebec. Geographie Physique et Quaternarie 33, 359–368.
C. Spence, M.-k. Woo / Journal of Hydrology 279 (2003) 151–166166