Water and solute transfer between a prairie wetland and adjacent uplands, 1. Water balance

Journal

t$droloka!$y ELSEVIER Journal of Hydrology 207 (1998) 42-55

Water and solute transfer between a prairie wetland and adjacent uplands, 1. Water balance

Masaki Hayashi a. *, Garth van der Kamp’, Dave L. Rudolph’

“Department of Geology and Geophysics. University of Calgary. Cdgar?;. Alberta T2N IN4. Canada ‘National Hydrology Research Institute, Saskatoon. Saskatchewan S7N 3H5. Cunnda

‘Department of Earth Sciences. University of Waterloo, Waterloo. Ontario N2L 3Gl. Canuda

Received 10 June 1997: accepted 28 January 1998

Abstract

The hydrology and water quality of lakes and wetlands are controlled by the exchange of water and solutes with adjacent uplands. We studied a small catchment in Saskatchewan, Canada, to evaluate the mechanisms of water and solute transfer between the wetland and the surrounding upland. Detailed measurements of hydrologic processes (precipitation, runoff. evapotranspiration, and subsurface flow) and chloride distribution are combined to improve the estimate of the transfer flux. This paper describes hydrologic processes and Part 2 describes the solute transport processes. Large snowmelt runoff occurs in the catchment, which transfers 30-60% of winter precipitation on the upland into the wetland to form a pond in the center. Snowmelt water and summer precipitation infiltrate under the central pond. Infiltration accounts for 75% of water leaving the central pond and evapotranspiration accounts for 25%. Most of the infiltrated water flows laterally in the shallow subsurface to the wet margin of the pond and further to the upland, where it is consumed by evapotranspiration without recharging deep groundwater. The net recharge rate of the aquifer underlying the catchment is only l-3 mm year-‘. Snowmelt runoff transfers water from the upland to the wetland, and shallow subsurface flow transfers water in the opposite direction. When the two processes are combined, they provide the paths for cyclic transport of solutes. 0 1998 Elsevier Science B.V. All rights reserved.

Keywords: Wetlands; Hydrologic cycle; Ground water; Recharge; Glaciated terrains; Semi-arid environment

1. Introduction

The hydrology and water quality of rivers, lakes, and wetlands is strongly controlled by the exchange of water and solute with adjacent banks and uplands. In the northern prairie region of North America, millions of small wetlands exist on the undulating terrain. They are commonly referred to as prairie potholes or sloughs and are a major component of the water resources and ecology of the region. Most prairie

* Corresponding author. Tel.: 001 403 220 2794; fax: 001 403 284 0074: e-mail: [email protected]

0022.1694/98/$19.00 0 1998 Elsevier Science. B.V. All rights reserved. PIZ SOO22-1694(98)00098-5

wetlands are located in closed catchments without integrated drainage network. The wetlands and asso- ciated catchments provide favorable sites for the study of water and solute transfer processes involving surface and subsurface water because each catchment constitutes an independent hydrologic system.

The physiography of the northern prairies is characterized by the cold semi-arid climate and clay-rich soils and underlying glacial deposits. The cold winter causes the accumulation of drifting snow in wetlands and significant snowmelt runoff from surrounding uplands (Woo and Rowsell, 1993). The clay-rich sediments allow wetlands to hold snowmelt water after the

M. Hayashi et al./Journal of Hydrology 207 (1998) 42-55 43

snowmelt period, but high evapotranspiration and low precipitation in the summer cause many wetlands to dry up by late fall (Winter and Rosenberry, 1995). Clearly, snowmelt runoff and evapotranspiration are important processes controlling the water balance, but the slow seepage of groundwater through clay-rich sediments is an equally important process governing the hydrology and water quality of prairie wetlands. Without seepage these wetlands would become saline.

Lissey (1968) showed that the recharge of groundwater in the northern prairie region is commonly con- centrated in wetlands and Lissey coined the term ‘depression-focused recharge’ to describe this phe- nomenon. Other workers (e.g. Zebarth et al., 1989) also showed that infiltration of water is highly focused under recharge wetlands and is almost negligible on surrounding uplands. However, high infiltration rates from ponded water in wetlands do not necessarily mean high rates of groundwater recharge to underlying aquifers. Much cf the infiltrated water flows laterally to the adjacent wet margins and uplands, where it leaves the system through evapotranspiration (Meyboom, 1966a; Mills and Zwarich, 1986). The difference between localized infiltration and aquifer recharge is sometimes overlooked (e.g. Woo and Rowsell, 1993), but it needs to be carefully examined because only the latter accounts for the transfer of water and dissolved ions out of the catchment.

Groundwater inflow and outflow strongly affect the water quality of wetlands. For example, wetlands recharging groundwater (recharge wetlands) have low-salinity water and can be easily distinguished from discharge wetlands that have high-salinity water (Lissey. 1968). By studying the movement of water and solutes at the same time, we can quantify the transfer processes with more confidence.

This two-part paper is an integrated study of water and solute transfer processes in prairie wetlands, which examines the hydrology of surface water, vadose zone, and groundwater. In Part 1, the water balance of a small wetland in Saskatchewan, Canada, is evaluated and all major hydrologic processes are described. We will show that snowmelt runoff drives infiltration under the wetland and groundwater how toward the upland. Groundwater flow is also driven by evapotranspiration on the upland. Outward groundwater flow, combined with snowmelt runoff, provides a path for cyclic transport of solutes between the

wetland and the upland. The link between hydrologic regime and water quality is described in Part 2 (Hayashi et al., 1998).

2. Field site

The wetland is located in the St. Denis National Wildlife Area (106”06’W, 52”02’N), 40 km east of Saskatoon, Saskatchewan, Canada (Fig. 1). It is iden- tified as S109, and is one of the series of wetlands studied by Miller et al. (1985). S109 and 17 similar wetlands occur in a cultivated field approximately 1 km2 in size. The cultivated field is surrounded by a narrow valley. The elevation of S109, which is located at the center of the cultivated field, is 10 to 15 m higher than the valley floor. S109 is a recharge wetland according to Lissey’s (1968) classification.

The area has hummocky topography and is underlain by stratified silty sediments, which are further underlain by glacial till of the Battleford and Floral Formations, which are Middle to Late Wisconsin in age (Christiansen, 1992). Because the boundary between the two formations is not clearly visible and is not hydrogeologically relevant, it is not shown in the geological cross-section (Fig. 2) which is based on Miller (1983) and our bore hole logs. The more relevant feature of the till is its oxidation state. The till is oxidized to a depth of 4-6 m (oxidation front) and is olive brown. The unoxidized till below the oxidation front is dark gray. The average texture of the till is 45% sand, 30% silt, and 25% clay (Miller et al., 1985). Thin (0.1-1.0 m) sand lenses occur frequently in the till, but they appear to be dis- continuous. A thin (0.3 m) but continuous layer of dense clay occurs at a depth of 8-9 m. A thicker (B1.5 m) and continuous sand layer at a depth of 25 m is an aquifer, which drains the groundwater under S109. This setting, an aquifer underlying a thick glacial till, is common in this region and called the ‘Prairie Profile’ (Meyboom, 1966b).

Sedge, spike rush, and other aquatic species grow in the wetland in summer. Silver willow, trembling aspen, and balsam poplar grow along the wetland margin to form a ‘willow ring’ (Meyboom, 1966a and Meyboom, 1966b). The willow ring marks the transition from the wetland to the cultivated field on the upland, which was planted with spring wheat in

44 M. Hayashi et al./Journal of Hydrology 207 (1998) 42-55

0 50m ,, 555.- I I

Fig. 1. Topographic map of the study site. The locations of field instruments are indicated in the map. Elevation is in meters above ?ea level. Arrows indicate overflows.

1993 and 1994, fallowed in 1995, and planted with oilseed in 1996.

In this study, the term ‘wetland refers to a topographic depression having nearly saturated soil most of the year, and the term ‘pond’ refers to the inundated portion of the wetland (Fig. 3). The pond usually forms in spring and disappears after several months. The term ‘upland’ refers to the area surrounding the

wetland, which includes both the cultivated field and the willow ring. Accordingly, the wetland of S109 is the region enclosed by the 552.5 m elevation contour (Fig. 1) with an area of 2400 ml. The area of the catchment based on elevation contours is 24 000 m’ (Fig. 1). However. the effective catchment may be greater in years of large snowmelt because snowmelt water may partially overflow from adjacent small

520 4 0 20 40 60 60 100 120 140 160

distance (m)

LEGEND m oxidized till a sand

silty deposits 0 unoxidized till EB clay

Fig. 2. Geological cross-section along the transect NS in Fig. 1 (modified from Miller, 1983). The locations of new bore holes are shown by vertical lines.

M. Hayashi et ai./Journal of Hydrology 207 (1998) 42-55 45

-upland-i-wetland-,-upland-

willow Ang willow ring 4-j ! : -pond-# I-

Fig. 3. Definitions of the pond. wetland, willow ring, and upland.

depressions. Because it is difficult to quantify the amount of overflow, we use the catchment area 24 000 m’ in water balance calculations. Catchments with this size are common in the prairies and are well suited for the detailed study of water and solute transfer.

The annual mean air temperature at Saskatoon Airport is 2°C with monthly means of -19°C in January and 18°C in July (AES, 1997). The area generally becomes covered by snow in November; therefore, we define a hydrologic year starting on November 1 and ending on October 3 1. A hydrologic year consists of a winter (November-March) and a summer (April-October); for example the winter of 1994 starts in November 1993. The 90-year mean annual precipitation in Saskatoon is 360 mm, of which 84 mm occurs in winter mostly as snow (AES, 1997). The annual evaporation from large lakes in this area is 690-710 mm (Morton, 1983). based on the data from Last Mountain Lake, located 100 km south-east of the study site.

This paper presents the results for the period between July 1993 and October 1996. The study period is characterized by relatively wet conditions and high water level in the pond. Fig. 4 shows the

long-term (1968- 1996) records of summer and winter precipitation in Saskatoon (AES, 1997) and the depth of water in the pond in late April (Millar et al., 1998). The variability of climate shown in Fig. 4 is an important factor in the hydrology of prairie wetlands (Winter and Rosenberry, 1995).

3. Field methods

3.1. Precipitation, evaporation, and pond water level

Summer precipitation in the hydrologic years 1994-1996 was measured by a tipping bucket rain gauge in the catchment (near Tl in Fig. 1). Summer precipitation in 1993 and winter precipitation in 1993-1996 were obtained from the Saskatoon Air- port, 40 km west of the site (AES, 1997). The average difference of monthly precipitation between St. Denis and Saskatoon data for 21 months (April-October of 1994, 1995, and 1996) is 0.8 mm and the standard deviation is 19 mm. The difference may be significant when each individual month is concerned, but insig- nificant for long-term average water balance.

Snow depth was measured at 20 locations along a

g 400 .E s ‘3 .Z Q 200 .cL % h

0 -

Fig. 4. Precipitation in Saskatoon and the depth of the pond in late April for 1968-1996.


north-south transect (NS in Fig. 1) in 1994 and 1996 using a 61 mm diameter aluminum snow survey tube. Snow depth was also measured at 20 locations along a west-east transect (WE in Fig. 1) in 1996. Snow water equivalent was calculated by measuring the volume of each snow sample upon melting.

The water level in the pond in 1993 and 1994 was monitored using a Stevens F-type chart recorder. The water level in 1995 and 1996 was monitored using a pressure transducer (Geokon, 4500ALV-5) placed in the bottom of a 3.8 cm diameter stilling well, drilled 1.8 m into sediments to avoid temperature fluctuation. From the water level data, the volume of water in the pond and the area of the pond were estimated using volume-depth and area-depth functions. These functions were generated from the data of a level survey conducted on a 5 to 10 m grid.

Evaporation from the pond water surface was measured with a partially submerged Class-A pan in the center of the pond. Only the top 5-10 cm of the 25 cm deep pan was kept above the pond water level, and the pan water temperature and pond water temperature were monitored by thermistors (Campbell Scientific, 107) and a data logger (Campbell Scientific, 21X) to confirm that thermal equilibrium was maintained. The pan water level was monitored using a Stevens F-type chart recorder.

The evapotranspiration in the wetland was estimated using the Bowen ratio-energy balance method with a Campbell Scientific Bowen Ratio System (Malek and Bingham, 1993) in the center of the pond. Net radiation was measured by a net radiometer (Radiation & Energy Balance Systems, Q*7). Heat storage in pond water and in the bottom sediments was estimated from temperature and heat capacity of the pond water and sediments. The temperature was measured by thermistors and a data logger. Heat capacity per unit volume of the sediments (3.7 x lo6 J m-’ Km’) was estimated from the measured water, mineral, and organic matter content of sediment samples and the specific heat of the three components (Hillel, 1980, p. 161). Heat conduction through the sediments was estimated from the temperature gradient between two thermistors at depths of 0.1 and 0.4 m into sediments. Thermal diffusivity of the sediments (2.7 x lo-‘m’ s-‘) was estimated by analyzing the amplitude and phase lag of diurnal temperature fluctuations at the two depths (Hillel,

1980, p. 304). These detailed heat measurements were necessary for calculating the energy balance of the pond. The measurement of air temperature and humidity is described by Malek and Bingham ( 1993).

The data were collected at 30-min intervals, and energy balance and average Bowen ratio were calculated for 24-hr intervals to minimize the error in esti- mating heat storage and to avoid unfavorable conditions that may occur if shorter intervals are used (Ohmura, 1982). The daily evapotranspiration was estimated from the 24-hr average data.

3.2. Groundwater

Miller (1983) installed 20 piezometers and water table monitoring wells in 1980. We installed an additional 14 piezometers in October 1993, and four other piezometers and wells at various times between 1993 and 1994. Miller’s piezometers were made of 3.2 cm inside diameter PVC pipes and installed in 15 cm diameter holes. Sand packs, 0.7-1.9 m in length, were installed around screens and were sealed by bentonite pellets and clay to the surface. Most of the new piezometers were made of 1.3 cm inside diameter polyethylene tubes, and some were made of 2.4 to 5.2 cm inside diameter PVC pipes to house pressure transducers. The new piezometers were installed in 15 cm diameter holes drilled by a solid- stem auger. Sand packs, 0.7-3.6 m in length, were installed around screens and the bore holes were sealed by bentonite pellets and chips to the surface. Up to three piezometers were installed in a single hole. The intervals between sand packs were tilled with bentonite pellets to prevent flow between sand packs. The new water table monitoring wells were made of 4 cm inside diameter PVC pipe, and were installed in the bore holes drilled by smaller size (5 to 10 cm diameter) augers. The holes for the wells were filled with sand to I m below the ground surface and sealed with bentonite pellets to the surface.

The length of piezometer screens varied between 0.2 and 0.4 m. but the screen length is unimportant in clay-rich glacial tills that have low hydraulic conductivity. Instead, the length and diameter of sand packs determine the response of piezometers. Longer sand packs were used in deeper piezometers to ensure reasonably quick response. The basic time lag

Table 1


Summary of hydrologic parameters in 1993-1996: total precipitation. winter precipitation. snow water equivalent before the snowmelt period, effective snowmelt runoff, and crops on the upland

Year Period Total pep.

(mm)

Winter pep. (mm)

Snow water eqv. Snowmelt

(mm) (mm) crops

1993 Nov/92-O&93 367 60 1994 Nov/93-Octl94 376 81 1995 Nov/94-Octl95 393 17 1996 Nov/95-Ott/96 390 15

No data 17 Wheat 66 45 Wheat No data 21 Fallow 63 45 Oilseed

(Hvorslev, 1951) for most piezometers was less than of the access holes, they were pounded 20 cm into the 50 h, with the exception of a few of Miller’s undisturbed soil to ensure direct contact between the piezometers. exposed part of the probes and the soil.

All piezometers were monitored manually at one week to one month intervals. A few piezometers were continuously monitored by vibrating-wire pressure transducers (Geokon) and a data logger. Satu- rated hydraulic conductivity was calculated from the basic time lag. measured by standard slug tests (Hvorslev, 195 1) performed on the piezometers.

3.3. Vadose zone

Tensiometers were constructed from 23 mm diameter standard porous cups (Soil Measurement Systems, SP-031), PVC pipes, and acrylic tubes (Marthaler et al., 1983). They were installed by the access hole method (Cassel and Klute, 1986) and were measured with a hand-held pressure transducer (Soil Measurement Systems, SW-030). Twenty tensiometers were installed in three nests (indicated as T 1, T2, and T3 in Fig. 1) at depths ranging from 0.2 to 2.0 m. Tensiometers were monitored at one week intervals from June 1994 to October 1994.

Three nests of six TDR probes were installed within 1 m of the tensiometer nests at depth intervals O-0.3, 0.3-0.5, 0.5-0.7, 0.7-0.9, 1.3-1.5, and 1.9-2.1 m. The apparent dielectric constant, Kapp, of the soil was determined by analyzing the wave forms obtained by a portable reading unit (Soil Moisture Equipment Corp., 6050X1) using a computer program developed by Redman (1996). Soil water content was calculated from Kapp using the formula of Topp et al. (1980). The water content at six depth intervals was integrated to calculate total water storage between the ground surface and 2.1 m. The missing intervals, 0.9-1.3 and 1.5-1.9 m, were estimated by interpolation. TDR probes were monitored at one week intervals from April 1994 to October 1994.

Soil water content was monitored using time domain reflectometry (TDR). TDR probes were made of a pair of stainless steel rods (diameter 6.4 mm, spacing 50 mm). Probes longer than 50 cm were covered with PVC pipes (outside diameter 13 mm) leaving only the bottom 20 cm exposed to the soil to avoid conductive loss of the signal in saline soil. The idea of using PVC to cover the rods is similar to the method proposed by Topp and Davis (1985). The PVC-cased probes were inserted into vertical access holes having the same diameter as the probes to the depth just above the proposed monitoring zone. Once the sharp tips of the probes reached the bottom

Soil temperature was measured at Tl and T3 sites at depths ranging from 0.2 to 2.0 m. An array of seven copper-constantan thermocouples and a portable read out device (Cole-Parmer, 8528-10) were used to measure temperature at one month intervals from October 1993 to March 1994, one week intervals from April 1994 to October 1994, and one month intervals from November 1994 to June 1995.

4. Results

4.1. Precipitation and snow accumulation

Total and winter precipitation in the hydrologic years 1993-1996 are listed in Table 1. Snow depth and water equivalent were measured just before the snowmelt started (March 3 1994 and March 13 1996). Due to drifting and wind scouring during the winter,

48 M. Hayashi et al./fournal of Hydrology 207 (I 998) 42 55

E 1.2

ii ’

-depth (94) --o--water eq. (94) ---water eq. (96)

80 60 40 20 0 20 40 60 80 (m)

-*-.water eq. (96)

Fig. 5. Distribution of snow depths and water equivalent along the transects shown in Fig. I: (a) transect NS: (b) transect WE. Shades indicate the location of the willow ring.

snow was almost absent near the catchment boundary in both 1994 and 1996. In contrast, snow was deep in the wetland, particularly along the willow ring (Fig. 5).

The total snow water equivalent in the catchment is estimated as follows. For 1996, we assume that the catchment is circular in shape and the radial distribu- tions of the northern, eastern, southern, and western quadrants of the catchment are represented by Fig. 5(a) and Fig. 5(b). Integrating the snow water equivalent from Fig. 5(a) and Fig. 5(b), the total snow water equivalent is 1500 m3 in 1996. Only one transect was measured in 1994. Assuming that the radial distribution of snow in the northern and southern halves of the catchment are represented by Fig. 5(a), the integration gives the total snow water equivalent of 1600 m3 in 1994. The same method gives 1600 m3 for 1996, which is not significantly different from

1500 m3 estimated by the quadrant method. The total snow water equivalent divided by the catchment area (24 000 m*) gives an average snow water equivalent listed in Table 1.

4.2. Snowmelt runoff and summer runoff

The snowmelt runoff occurred in several phases and lasted for 25-30 days as indicated by water level rises in the pond, which was dry in the winter (Fig. 6). At the end of the snowmelt period, the volume of water in the pond, estimated from the water level, was 1100 m’ in 1994. 500 m3 in 1995. and 1100 m3 in 1996. This volume, divided by the area of the catchment, is the effective snowmelt runoff per unit area (Table 1). The effective snowmelt runoff in 1993-1996 ranged between 30 and 60% of winter

s sSl.Si,,,,i.,,,,,.,,,,i,,,,,,,,,,,;, ,,I,~-r! J J ASONDJ FMAMJ J ASONDJ FMAMJ J ASONDJ FMAMJ J AS0 1993 1994 1996 1906

Fig. 6. Daily precipitation and the water level in the pond. Dotted lines indicate the division between hydrologic years. Winter months are indicated by bars on top of the graphs. (a) Precipitation. (b) Pond water level.

M. Hayashi et al./Joumal of Hydrology 207 (1998) 42-55 49

40 , A z -E

. 8

30 .

5 co 20 ?? '. :/I . . .

T ??.

a, . E

IO . . kz

01/ 0 IO 20 30 40

precipitation (mm)

Fig. 7. Pond water level change versus precipitation. The straight line shows the slope of 1.0.

precipitation in Saskatoon (Table 1). The same ratio for 1968-1996, estimated from Fig. 4, is 30% on average.

Major summer runoff on the upland occurred only once (July 4 1996) in the study period. During this storm, the water level in the pond rose 210 mm when the precipitation was only 54 mm. In Fig. 7, the water level rise is plotted against precipitation for all storms, except for the one mentioned above, that had more

June 27 29 1 _

than 10 mm of precipitation and occurred while the pond existed in the wetland. The water level rise for these storms is not in great excess of precipitation, indicating that the runoff was minor and it perhaps occurred only on the non-inundated portion of the wetland where the water table was close to the ground surface.

4.3. Evapotranspiration and infiltration in the pond

After the initial rise in the spring, the water level in the pond gradually declined due to evapotranspiration and infiltration, and the pond disappeared before November except for 1996 (Fig. 6). The water level generally declined fast in the afternoon and slow in the morning and evening, resulting in the step-like patterns of water level curve (Fig. 8(b)). To separate the water level curve into infiltration and evapotranspiration, we installed a submerged pan in the pond during a two-week period in the summer of 1994, just a few weeks before the pond disappeared (Fig. 6(b)). The slope of the pan water level curve indicates the rate of pan evaporation and precipitation (Fig. 8(c)).

July 5 7 9 11

178 180 182 184 186 188 190 192

Julian day (1995)

Fig. 8. Water level fluctuation in the pond during the period of evaporation measurement. The grid lines indicate the beginning of each day, 0:OO. (a) Two-hour precipitation. (b) Water level in the pond. (c) Water level in the pan and the residual water level, both relative to the initial levels.

50 M. Hayashi et al./Journal of H_vdrology 207 (1998) 32-55

4

8

16

20

24 L till

Fig. 9. Distribution of saturated hydraulic conductivity. Shaded area indicates the till matrix conductivity. Stratigraphy of the area is schematically shown beside the graph.

The residual curve in Fig. 8(c) was obtained by subtracting the pan water level change from the total water level change. The slope of the residual curve indicates the rate of infiltration. Previous researchers assumed, without clear field evidence, that the infiltration rate was constant day and night and evapotranspiration from the pond caused the step- like pattern (Eisenlohr, 1966; Shjeflo, 1968; Keller et al., 1986; Woo and Rowsell, 1993). Our data in Fig. 8(c) indicate that the step-like pattern is largely due to a high infiltration rate in the afternoon. The cause of high infiltration will be discussed later.

During the two-week observation period, the average daily pan evaporation rate was 3.1 i 0.7 mm (Lindicates one standard deviation). During the same period, the average daily evapotranspiration measured by the Bowen ratio-energy balance method was 3.0 ? 0.7 mm. These values are smaller than the expected rate of daily evaporation from large lakes in July (3.9-5.1 mm) reported by Morton (1983, p. 98). This result suggests that the willow ring, which is 5- 10 m tall, presumably blocks the wind and reduces the turbulent transport of water vapor within the wetland.

4.4. Hydraulic conductivity of glacial till

Saturated hydraulic conductivity measured by slug tests is on the order of 1OA m s-’ near the surface, and

Kat (m ~~‘1 1 O-10 IO" 106

oxidized till

sand

clay

unox.

in deeper piezometers that intersect sand lenses (Fig. 9). The conductivity decreases with depth. The transition from the high to low occurs at a depth of 5-6 m, which coincides with the boundary between oxidized and unoxidized tills. i.e. oxidation front. The lowest hydraulic conductivity in the unoxidized till is approximately lo-” m s-’ (shaded area in Fig. 9). which is likely close to the conductivity of the unfractured clayey matrix of the till. Higher hydraulic conductivity is likely due to a fracture network that readily transmits water (Keller et al.. 1988). Hydraulic conductivity is higher in the oxidized till than in the unoxidized till, perhaps because fractures occur more frequently in the shallow oxidized till. Similar contrast of hydraulic conductivity is observed commonly in the prairies (e.g. Zebarth et al.. 1989; Winter and Rosenberry, 1995).

Due to moderately high clay content, the glacial till and soils derived from the till have high water holding capacity. Hayashi et al. (1997) describe the unsaturated flow characteristics of the till in detail.

4.5. Groundwater JOW

The hydraulic head distribution in the spring of 1994 indicates that groundwater flow direction was outward, from the wetland to the upland (Fig. 10(a)). The horizontal groundwater flow occurs mostly in the oxidized zone of the glacial till, because the hydraulic conductivity in the oxidized zone is several orders of magnitude higher than the unoxidized zone. The outward flow continued during the summer (Fig. 10(b)), but it became negligible in the late fall after the pond disappeared (Fig. 1 O(c)). Contours in the vadose zone are only shown in Fig. IO(b), because tensiometer readings were only taken in the summer and fall of 1994. The head distribution in the spring of 1995 was similar to that in 1994; however, the water table under the upland kept rising in the summer and the flow direction became inward in the fall (Fig. 10(d)). The drastic change of groundwater flow direction occurred perhaps because the upland was fallowed in 1995 so that evapotranspiration from the upland was much reduced.

Dense contour lines at a depth of 8-9 m indicate that vertical flow was impeded by the clay layer (Fig. 2). The vertical hydraulic gradient below the clay layer was nearly constant and downward

M. Hayashi et al./JoumalofHydrology 207(1998) 42-55 51

throughout the year (Fig. 10). At the piezometers under the willow ring (25 in Fig. l), the gradient between the depths of 14.6 and 21.8 m was 0.17, and the harmonic mean hydraulic conductivity was 6.2 x lo-” m s-’ based on slug tests. At the piezometers under the cultivated field (27 in Fig. I), the gradient between 18.2 and 21.9 m was 0.26, and the mean conductivity was 1.7 x lo-” m s-‘. The product of the hydraulic gradient and conductivity estimates the vertical flux into the aquifer; 3.3 mm year-’ under the willow ring and 1.4 mm year-’ under the cultivated field.

4.6. Vadose zone

The hydraulic gradient in the vadose zone was generally upward in the summer and fall of 1994 (Fig. 10(b)), indicating that the infiltration from the surface was consumed by evapotranspiration within the root zone. The upward gradient further indicates that the root zone receives water from below in addition to precipitation on the surface.

The root zone refers to the zone within 2.1 m of the ground surface in this paper. This depth corresponds to the deepest TDR probe. The root system of wheat does not grow deeper than 2 m (Chaudhary and Bhatnagar. 1980; Meyer et al., 1990). Soil water storage in the root zone in 1994 was at its maximum in the spring and decreased in the summer and fall at all three TDR sites. The storage change (AS) between 29 June and 6 November was -110 to -120 mm at all three sites. Because the flow in the root zone was upward during this period, evapotranspiration was in excess of the precipitation (150 mm) by at least 1 lo- 120 mm in this 130-day period. Large excess evapotranspiration in the summer suggests that the net water flux across the ground surface on the upland may be upward even when partial infiltration of winter precipitation is taken into account.

4.7. Average flow jield and short-term variations

On average, the shallow subsurface flow system is characterized by large infiltration under the wetland, lateral flow within the zone of high hydraulic conductivity, and upward flow in the vadose zone under the upland (Fig. 11). The oxidation front marks the boundary between high and low conductivity zones,

h%J, ,+ , ,4 I ; (

0 20 40 60 80 100 120 140 Go distance (m)

560

550

540

530

Summer 1994

548

c 4 i

547

/

0 20 40 60 80 100 120 140 160

560,+) Winter 1994

530 ' I 0 20 40 60 60 100 120 140 160

Fall 1995

530 ' 0 20 40 60 80 100 120 140 2160

Fig. 10. Hydraulic head distribution along the transect AB. Arrows indicate general flow directions. (a) April 28 1994: (b) July 28 1994; (c) December 13 1994; (d) September 18 1995.


Fig. I I. Average subsurface flow patterns.

and also marks the transition from the shallow to deep subsurface flow system. Only a small fraction of infiltration under the wetland enters the deep system and flows vertically to recharge the aquifer (Fig. 11).

On shorter time scales, the subsurface flow between the wetland and the upland has complex variations. For example, the upland on the south side of the wetland has a much steeper slope than the rest of the catchment (Fig. l), which causes the horizontal hydraulic gradient in this side to become inward toward the wetland when the water table under the wetland drops in late fall of dry years. Meyboom (1966a) observed similar reversal of hydraulic gradient in the array of piezometers installed along the steepest slope in the catchment.

5. Discussion

5.1. Injiltration under the pond

The residual curve in Fig. 8(c) indicates that high infiltration occurs under the pond in the afternoon. In general, infiltration is induced by transpiration of plants both inside and outside the pond. However, Fig. 8(c) likely represents that part of infiltration induced by plants growing outside the pond, i.e. the non-inundated part of the wetland (see Fig. 3). for the following reasons.

The evaporation pan measures the flux of vapor leaving the open water surface within the pan. The pond surface has emergent and floating plants, which intercept part of the solar radiation and reduce the vapor flux leaving the water surface. Humidity, temperature, and wind speed over the pan are expected to be similar to that over the rest of the pond. The net radiation is also expected to be similar. Therefore, the sum of transpiration and open water evaporation over a unit area of the pond is expected

to be close to evaporation over a unit area of the pan. The close match between pan evaporation and evapotranspiration measured by the Bowen ratio-energy balance method supports this idea. If the pan evaporation represents evapotranspiration within the pond, the residual curve in Fig. 8(c) represents the amount of water flowing from under the pond to the pond margin. The flow is presumably induced by transpiration of plants in the non-inundated part of the wetland.

Hayashi (1996) measured pan evaporation in a recharge wetland (S120) located 400 m north-east of S109, which was clear of emergent plants and obtained a residual curve almost identical to Fig. 8(c). In this experiment, both the pond surface and the pan surface were covered by floating plants commonly called ‘duck weed’, and the slope of the residual curve clearly indicates the groundwater flow from under the pond to the pond margin. Meyboom ( 1967) also found large diurnal fluctuations of groundwater flow rate from under the pond to the pond margin in prairie wetlands.

Based on the above discussion. we assume that the slope of the residual curve in Fig. 8(c) is equal to the infiltration rate under the pond. The average infiltration rate under the pond in the 2-week period was 9.4 mm day-‘. which accounted for 75% of water leaving the pond. An independent measurement by a chloride mass balance method also showed that inti- tration in the summer accounted for 80% of water leaving the pond (Hayashi et al., 1998). These results are consistent with the study of Millar (197 1) who concluded that a significant quantity of water leaves the prairie ponds through infiltration and subsequent flow toward the perimeter of the ponds.

5.2. lnjiltration capacity and runoff on the upland

The soil under the cultivated field (T3 site) was frozen until late May. as indicated by sub-zero soil


upland (21600 mz) ’ wetland (2400 ml)

Fig. 12. Estimation of annual water balance (mm) in 1994: Pep, precipitation: Evt. evapotranspiration; Snm, snowmelt runoff; Hfl, horizontal groundwater flow: Vfl. vertical groundwater flow.

temperature. The infiltration capacity of frozen soil varies significantly depending on both the soil texture and antecedent moisture condition before freezing. However, the infiltration capacity of frozen soil generally is much lower than that of the same soil in the unfrozen state (Granger et al., 1984). The saturated hydraulic conductivity of the top soil is in the order of lo4 m SC’ or 100 mm day-’ (Fig. 9), which gives the lower limit of the infiltration capacity in the unfrozen state. This is much higher than the expected rate of snowmelt (5-30 mm day-‘, Woo and Rowsell, 1993), and the soil must have reduced infiltration capacity to generate significant snowmelt runoff.

In summer the moderately high infiltration capacity of the soil prevents runoff on the upland, except during unusually intense storm events (e.g. July 4 1996).

5.3. Water balance

The water balance in the catchment is estimated from the field data of 1994 (Fig. 12), which is the only year with a complete data set. Because water levels in shallow piezometers in October 1993 were roughly the same as the water levels in October 1994, change of subsurface water storage in the catchment is assumed negligible. Annual precipitation for the hydrologic year 1994 is 376 mm (Table l), of which 45 mm in the upland becomes effective snowmelt runoff (Table 1) and goes into the wetland. Note that the area of the wetland is one-ninth of the area of the upland. Total input of water to the wetland, precipitation plus effective snowmelt runoff, is 780 mm ( = 45 x 9 + 376).

Based on the pan experiment, evapotranspiration

rate in the pond is 25% of the rate of water level decline in the pond. This gives the total pond evapotranspiration of 240 mm in the summer and fall of 1994. The non-inundated part of the wetland has dense aquatic vegetation and the water table close to the ground surface. Therefore, the evapotranspiration in the non-inundated part is perhaps close to that in the pond. Because the evapotranspiration in the winter and spring is expected to be small, we assume that the total evapotranspiration in the entire wetland is 300 mm.

Subtracting the evapotranspiration from the total input, the amount of water leaving the wetland is 480 mm, of which approximately 2 mm becomes the net groundwater recharge to the aquifer, and the rest is transferred to the upland where willows and crops consume a large amount of water. Evapotranspiration in the upland is the sum of precipitation and the water transferred from the wetland minus runoff and the recharge to the aquifer, hence it is 380 mm.

This suggests that evapotranspiration in the upland planted with wheat is greater than that in the wetland. In 1994, wheat was seeded in middle May and har- vested in early September. Soil matric potential head at 0.4 m below the ground surface at Tl-T3 sites was above -3 m during the growing season. Meyer et al. (1990) reported that wheat growing in clay-rich soil can transpire at a rate close to potential evaporation when soil matric potential head is above -6 m. Total potential evaporation in southern Saskatchewan in June, July, and August is 360-450 mm (Morton, 1983, p. 98). Therefore, evapotranspiration in the upland may be as high as 380 mm. However, our estimates of evapotranspiration are dependent on a num- ber of assumptions, and the comparison between the wetland (300 mm) and upland (380 mm) has a large degree of uncertainty. Further research is required to evaluate the effects of the willow ring and crops on evapotranspiration.

Even though the estimated flux in Fig. 12 has a large degree of uncertainty, a few important points should be noted. First, large infiltration occurs under the wetland but only a small fraction recharges the aquifer. Second, water is transferred from the wetland to the upland by horizontal groundwater flow. The magnitude of horizontal flow is approximately equal to that of effective snowmelt runoff from the upland to the wetland. When snowmelt runoff and horizontal

54 M. Hayashi et ai./Journal of Hydrology 207 (1998) 42-55

groundwater flow are combined, they provide the paths for cyclic transport of solutes within the catchment.

5.4. Effects of farming practices

Farming practices on the upland seem to have significant effects on the hydrology of the wetland. First, snowmelt runoff induces infiltration under the wetland, and thus drives the hydrologic cycle in the catchment. Changes in stubble height may change the amount of snowmelt runoff and hence subsequent hydrologic processes. Second, the horizontal flow of groundwater from the wetland to the upland is largely due to transpiration by willows and crops on the upland. When the cultivated part of the upland had no vegetation in the year of summer fallow, the horizontal flow direction became inward. This may also indicate that the wetland is an important source of water for the crops growing on the upland, and that draining the wetland may result in a lower yield of crops.

6. Conclusions

Large snowmelt runoff occurs in the catchment because soils are frozen and thus have greatly reduced infiltration capacity in the spring. On average, roughly 3060% of winter precipitation in the catchment enters the wetland as snowmelt runoff. Summer storm runoff occurs only during extremely large events. Therefore, the water level in the pond during the summer is almost entirely governed by direct precipitation, evapotranspiration, and infiltration. Infil- tration accounts for 75% of water leaving the pond. The infiltration rate is high during the day and low during the night, reflecting transpiration rate from the non-inundated margin of the pond. Evapotranspira- tion rate within the wetland is less than the evaporation from large lakes in southern Saskatchewan. Trees around the wetland seem to reduce the air flow and turbulent transport of vapor within the wetland.

The high infiltration rate under the wetland does not mean a high recharge rate of the deep aquifer. Much of the infiltrated water flows laterally through a zone of high hydraulic conductivity, due to fractures in the till, and is used up by plants on the upland. The

recharge rate of the aquifer underlying the catchment is only about 1% of annual precipitation. When the upland has transpiring plants, the direction of groundwater flow is outward. In contrast, the direction may become inward when the upland is fallowed. This indicates that the outward flow is driven by transpiration on the upland and the hydrology of prairie wetlands is strongly dependent on land use practices.

Water transfer between the wetland and the upland has a strong cyclical nature. In the spring snowmelt water makes its way from the upland to the wetland over the land surface. During the summer water moves by shallow subsurface flow from the wetland to the upland. where it is consumed by evapotranspiration. For a hectare-scale catchment, such as the one that was studied, the two fluxes are roughly equivalent in magnitude and provide the paths for cyclic transport of dissolved ions.

Acknowledgements

We gratefully acknowledge the Canadian Wildlife Service for the use of the St. Denis National Wildlife Area. We thank Randy Schmidt, Vijay Tumber. and Brad Fahlman for field assistance; Mike English and Warren Eilers for discussion; and three reviewers for constructive comments. The research was supported by the Natural Sciences and Engineering Research Council of Canada, and by a Government of Canada Awards scholarship to the senior author.

References

AES. 1997. Canadian daily climate data Canada. Atmospheric Environment Canada, Downsview, Ontario, Canada

Cassel, D.K., Klute. A., 1986. Water In: Methods of soil analysis, Part 1.

on CD-ROM. Western Service, Environment

potential: tensiometry. Agronomy Monograph

No. 9, 2nd edn, pp. 563-596. Am. Sot. Agron.. Madison, Wisconsin, USA.

Chaudhary. T.N., Bhatnagar, V.K.. 1980. Wheat root distribution. water extraction pattern and grain yield as influenced by time and rate of irrigation, Agr. Water Management, 3, 115-124.

Christiansen, E.A., 1992. Pleistocene stratigraphy of the Saskatoon area, Saskatchewan, Canada: an update, Can. J. Earth Sci., 29. 1167-1778.

Eisenlohr, Wm..% Jr, 1966. Water loss from a natural pond through transpiration by hydrophytes, Water Resour. Res., 2, 443-453.

h4. Hayashi et al./Journal of Hydrology 207 (1998) 42-55 5.5

Granger, R.J., Gray, D.M.. Dyck, G.E., 1984. Snowmelt infiltration to frozen Prairie soils, Can. .I. Earth Sci., 21, 669-677.

Hayashi. M., 1996. Surface-subsurface transport cycle of chloride induced by wetland-focused groundwater recharge. Ph.D. thesis, University of Waterloo, Waterloo, Ontario, Canada.

Hayashi. M.. van der Kamp, G., Rudolph, D.L., 1997. Use of tensiometer response time to determine the hydraulic conductivity of unsaturated soil, Soil Sci., 162, 566-575.

Hayashi. M., van der Kamp, G., Rudolph, D.L.. 1998. Water and solute transfer between a prairie wetland and adjacent uplands, 2. Chloride cycle. J. Hydrol., 207( l-2). 56-67.

Hillel. D.. 1980. Fundamentals of Soil Physics. Academic Press, New York.

Hvorslev. M.J.. 1951. Time lag and soil permeability in groundwater observations. Corps of Engineers, US Army, Waterways Experiment Station. Bull, No. 36.

Keller, C.K.. van der Kamp, G., Cherry. J.A., 1986. Fracture permeability and groundwater flow in clayey till near Saskatoon, Saskatchewan, Can. Geotech. J., 23, 229-240.

Keller, C.K., van der Kamp, G., Cherry, J.A.. 1988. Hydrogeology of two Saskatchewan tills: I. Fractures, bulk permeability, and spatial variability of downward flow, J. Hydrol.. 101, 97-121.

Lissey, A.. 1968. Surficial mapping of groundwater flow systems with application to the Oak River Basin, Manitoba. Ph.D. thesis, University of Saskatchewan. Saskatoon, Saskatchewan, Canada.

Malek, E., Bingham, G.E., 1993. Comparison of the Bowen ratio- energy balance and the water balance methods for the measurement of evapotranspiration, J. Hydrol., 146, 209-220.

Marthaler, H.P., Vogelsanger. W., Richard, F., Wierenga, P.J., 1983. A pressure transducer for field tensiometers, Soil Sci. Sot. Am. J., 47, 624-627.

Meyboom, P., 1966a. Unsteady groundwater flow near a willow ring in hummocky moraine, J. Hydrol., 4, 38-62.

Meyboom, P., 1966b. Groundwater studies in the Assiniboine River drainage basin: Part I. The evaluation of a flow system in south-central Saskatchewan, Geol. Surv. Can. Bull.. 139, 65.

Meyboom. P.. 1967. Mass-transfer studies to determine the groundwater regime of permanent lakes in hummocky moraine of western Canada. J. Hydrol., 5, 117-142.

Meyer, W.S.. Tan. C.S., Bans, H.D.. Smith, R.C.G., 1990. Root growth and water uptake by wheat during drying of undisturbed

and repacked soil in drainage lysimeters, Aust. J. Agr. Res., 41. 253-265.

Millar, J.B., 1971. Shoreline-area ratio as a factor in rate of water loss from small sloughs, J. Hydrol., 14, 259-284.

Millar, J.B., Clark. R.G.. Kiss, P., Stolte, W.S., van der Kamp, G.. 1998. Water-depth data for sixteen ponds in the St. Denis National Wildlife Area, 1968-1997. National Hydrology Research Institute Contribution (in press).

Miller, J.J., 1983. Hydrology of a morainic landscape near St. Denis, Saskatchewan, in relation to the genesis, classification and distribution of soils. M.Sc. thesis. University of Saskatch- ewan, Saskatoon, Saskatchewan, Canada.

Miller, J.J., Acton, D.F., St. Amaud, R.J., 1985. The effect of groundwater on soil formation in a morainal landscape in Saskatchewan, Can. J. Soil Sci.. 65, 293-307.

Mills, J.G.. Zwarich, M.A., 1986. Transient groundwater flow surrounding a recharge slough in a till plain, Can. J. Soil Sci., 66, 121-134.

Morton, FL. 1983. Operational estimates of lake evaporation, J. Hydrol.. 66, 77-100.

Ohmura, A., 1982. Objective criteria for rejecting data for B’owen ratio flux calculations, J. Appl. Meteorol., 21, 595-598.

Redman, J.D.. 1996. WATTDR user’s guide. Internal report. University of Waterloo, Waterloo, Ontario, Canada.

Shjeflo. J.B., 1968. Evapotranspiration and the water budget of prairie potholes in North Dakota. US Geol. Survey Prof. Paper 585-B.

Topp, G.C., Davis. J.L.. 1985. Measurement of soil water content using time-domain-reflectometry (TDR): a field evaluation, Soil Sci. Sot. Am. J., 49. 19-24.

Topp, G.C.. Davis, J.L., Annan, A.P., 1980. Electromagnetic deter- mination of soil water content: measurements in coaxial Irans- mission lines, Water Resour. Res., 16, 574-582.

Winter. T.C., Rosenberry, D.O.. 1995. The interaction of ground water with prairie pothole wetlands in the Cottonwood Lake area, east-central North Dakota, 1979-1990, Wetlands, 15, 193-211.

Woo, M.-K.. Rowsell, R.D., 1993. Hydrology of a prairie slough, J. Hydrol., 146, 175-207.

Zebarth, B.J., de Jong, E., Henry, J.L., 1989. Water flow in hummocky landscape in central Saskatchewan, Canada: II. Saturated flow and groundwater recharge, J. Hydrol.. 110. 181-198.

Documents

Water and solute transfer between a prairie wetland and adjacent uplands, 1. Water balance