THE RESPONSE OF WATER CHEMISTRY T O HYDROLOGICAL CONDITIONS A N D FLOW PATHS UNDER DIFFERENT FOREST COVER TYPES IN SMALL SUB ALPINE CATCHMENTS by Robert O. Hudson B.Sc., University of Calgary 1983 M.Sc., University of Calgary 1986 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Faculty of Forestry) We accept this thesis as conforming to the required standard The University of British Columbia November 1995 © Robert O. Hudson, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Pepartmcnt of rOf €2 67 Table 3.1.3: Average Precipitation Chemistry in mg/1, Results of Two Sample T-tests, Total Chemical Inputs due to Rain and Snow 67 Table 3.1.4: Average grain size profiles for early and late melt season 70 Table 3.1.5: Correlations (r) of Base Chemical Concentrations vs Daily Melt (mm) 87 Table 3.1.6: Regressions of Base Chemical Concentration vs Daily Melt 88 Table 3.2.1: Average Chemical Concentrations and Hydrologic Properties at Standpipes 89 Table 3.2.2: Relationships between average chemical concentration and Hydraulic Conductivity 93 Table 3.2.3: Preliminary Groundwater Groupings Suggested by Figure 3.2.2 95 Table 3.2.4: Summary of Physical Soil Factors and Results of Discriminant Analysis 98 Table 3.2.5: Group numbers assigned to soils on the basis of two factors 99 Table 3.2.6: Discriminant analysis of Soil Texture Groups 99 Table 3.2.7: Final Discriminant Analysis of Soil Texture Groups 100 Table 3.2.8: Summary of Available Cations in Soil; classifications based on soil texture and on available cations 103 Table 3.2.9: Discriminant analysis between soils grouped by Ca availability 103 Table 3.2.10: Table of Dummy Variables of Groundwater Site Characteristics 106 Table 3.2.11: Table of average chemical analyses using different dummy variables 108 Table 3.2.12: Correlation Matrix for Group 1 (streamside sites) 109 Table 3.2.13: Correlation Matrix for Group 2 (hillslope sites) 109 Table 3.2.14: Equations Describing Average Chemical Concentration vs Land Slope (0) and Pipe Depth () for Two Hydrological Zones in Soil 114 Table 3.2.15: Standpipe Groups Based on Hydrologic Zones 122 Table 3.2.16: Sulphate vs Total Head 123 Table 3.2.17: Bicarbonate vs Total Head 124 Table 3.2.18: Calcium vs Total Head 125 Table 3.2.19: Sodium vs Total Head 126 Table 3.2.20: Silica vs Total Head 127 Table 3.2.21: Zone 1, 2 and 4 Head and Silica Concentrations at Low and High Flow 143 Table 3.2.22: Average Chemical Concentrations in Slow and Fast Interflow 144 vii LIST OF TABLES (continued) page Table 3.3.1: Log-normal concentration-discharge relationships 146 Table 3.3.2: Mass flux-discharge relationships 154 Table 3.3.3: Average annual mass flux September 1987-August 1990 (Sep 88 - Aug 90 Edelweiss Creek) 154 Table 3.3.4: Average Annual Net input(+)/output(-) 154 Table 3.3.5: Distribution of Anions and Cations in Streamflow 158 J Table 3.3.6: Summary of Nutrient Budgets for Several Research Watersheds Standardized to 1 metre of Precipitation 160 Table 4.1: Equations to predict head from total streamflow 174 Table 4.2: UBC Model Calibration Options for 240 Creek 176 Table 4.3: Correlation Matrices of Groundwater Heads vs Routed Outflows 177 Table 4.4: UBC Model Calibration Parameters For 240, Dennis and Edelweiss Creeks 178 Table 4.5: Relationships of Groundwater Head vs Groundwater Component Outflows 179 Table 4.6: Details of Calculating Variability of Hydraulic Conductivity, Lower and Upper Groundwater, 240 Creek ~ 180 Table 4.7: True Basin-Wide Hydraulic Conductivity Variability and Ionic Chemical Retardation Factors 184 Table 4.8: Frequency at which simulated chemical concentrations were within 20% of observed values 209 viii LIST OF FIGURES Figure Title page 1.1: UPC Experimental Watersheds. 31 1.2: 240 Creek with Forest Cover 32 1.3: Dennis-Edelweiss Creeks with Forest Cover 33 i 1.4: Area-Elevation Curves for Upper Penticton Creek Study Watersheds 35 1.5: Main Channel Profiles 35 1.6: Three Year Mean Climatographs and Hydrographs 1988-90 38 1.7: Selected Soil Temperature Profiles, 1988-90 40 2.1: 241 Creek weir at medium to high flow 42 2.2: Dennis Creek weir at high flow 42 2.3: Edelweiss Creek weir at medium flow 43 2.4: Penticton-2 P&T monitoring site 43 2.5: Schematic of a Hillside with Groundwater Instrumentation 46 2.6: Schematic of a Piezometer 46 2.7: Bail Test for Pl/241 Creek 47 2.8: Bail Test for P1&P2 lower Edelweiss Creek 47 2.9: Cumulative Gauge Catch and SWE at Cheng and Mission, 1988-89 49 2.10: Mean Annual Precipitation vs Elevation Penticton Creek 1988-90 51 3.1.1: Determination of Asymptote for Exponential Curve Fit of Liquid Content 73 3.1.2: Liquid Water Content of Ripe Snow Under Different Temperature Conditions 73 3.1.3-18: Chemical Profiles in Snowpack at 240, Edelweiss and Dennis Creeks 1990 75-82 3.1.19-22: Chemical Concentrations at the Base of the Snow Pack vs Daily Melt, 240 & Dennis Cr., Edelweiss Cr. Clear-cut, Edelweiss Cr. Frst 89-90 3.2.1: Average Solute Concentrations in Groundwater vs Hydraulic Conductivity 94 3.2.2: Hydraulic Conductivity vs Standpipe Depth 94 3.2.3: 240 Creek Forest Cover with Soil and Stream Profile Sampling Sites 96 3.2.4: Dennis-Edelweiss Creeks Forest Cover with Soil and Stream Profile Sampling Sites 97 3.2.5: Average Cumulative Grain Size Curves 101 3.2.6: Average Grain Size Distribution for Three Soil Groups 101 3.2.7: Exchangeable Calcium and Magnesium in Dennis and 240 Creeks vs Elevation 105 3.2.7: Exchangeable Potassium and Sodium in Dennis and 240 Creeks vs Elevation 105 3.2.9-13: Average Chemical Concentration vs Land Slope for Two Hydrological Zones and Three Forest Cover Types 111-113 ix LIST OF FIGURES (continued) Figure Title 3.2.14: Average Silica Concentration vs Pipe Depth for Two Hydrological Zones and Three Forest Cover Types 3.2.15: Pressure Head and Silica Concentration vs Day, P3/240 Creek 1988-90 3.2.16: Silica Concentration vs Pressure Head on P3/240 Creek 3.2.17: Chemical Concentration vs Total Head for P2&3/240 Creek Comparison Between Specific and Zone 2 Equations 3.2.18: Chemical Concentration vs Total Head for P4/240 Creek Comparison Between Specific and Zone 1,3,4 Equations 3.2.19: Chemical Concentration vs Total Head for P5/240 Creek Comparison Between Specific and Zone 1,3,4 Equations i 3.2.20: Chemical Concentration vs Total Head for P I /Wl 241 Comparison Between Specific and Zone 1,3,4 Equations 3.2.21: Chemical Concentration vs Total Head for P2/W2 (Hillslope) 241 Comparison Between Specific and Zone 2 Equations 3.2.22: Chemical Concentration vs Total Head for P1A Edelweiss Comparison Between Specific and Zone 1,3,4 Equations 3.2.23: Chemical Concentration vs Total Head for Seepage Site Comparison Between Specific and Zone 1,3,4 Equations 3.2.24: Chemical Concentration vs Total Head for P1&2 Hillside Edelweiss Comparison Between Specific and Zone 2 Equations 3.2.25: Chemical Concentration vs Total Head for PI 242 Comparison Between Specific and Zone 1,3,4 Equations 3.2.26: Chemical Concentration vs Total Head for P2 242 Comparison Between Specific and Zone 2 Equations 3.3.1: Bicarbonate Concentration vs Stream Discharge, All Creeks 3.3.1-6: Chemical Concentration vs Specific Discharge 3.3.7: Average Concentrations in Streamflow on 240, 241, Dennis and Edelweiss Creeks vs Mean Elevation H50 3.3.8: Average Concentrations in Streamflow on 240, 241, Dennis and Edelweiss Creeks vs Average Crown Closure 3.3.9-11: 240, Dennis and Edelweiss Creek Profiles 3.3.12: Cluster Analysis of Research Basins using Cations, Cations +. Anions Figure Title 4.1-2 page 166 168 168 171 171 175 Hysteresis Loop of Streamflow During Snowmelt on Dennis Creek vs 240 Creek 1989 and 1990 r " 4.3: Early Snowmelt April 1990 ' 4.4: Snowmelt Waves Spring 1990 4.5: Three Year Average Streamflow and Groundwater Heads vs Time 240 Creek 4.6: Groundwater Head vs Streamflow on 240 Creek 1988-90 4.7: Simple Model of Bicarbonate in Base Flow 240 Creek April-September 1988-90 4.8: Average Basin-wide Upper and Lower Soil Hydraulic Conductivity vs Groundwater Outflow, 240 Creek 183 4.9: Retardation Factors for Ionic Species with respect to True Hydraulic Conductivity 240 Creek 183 4.10: Average Basin-wide Lower Soil Hydraulic Conductivity vs Groundwater Outflow, Dennis Creek 186 4.11: Retardation Factors for Ionic Species with respect to True Hydraulic Conductivity Dennis Creek 186 4.12: Average Basin-wide Lower Soil Hydraulic Conductivity vs Groundwater Outflow, Upper Edelweiss Creek 187 4.13: Retardation Factors for Ionic Species with respect to True Hydraulic Conductivity Upper Edelweiss Creek 187 4.14: Average Basin-wide Lower Soil Hydraulic Conductivity vs Groundwater Outflow, Edelweiss Creek Clear-cut 188 4.15: Retardation Factors for Ionic Species with respect to True Hydraulic Conductivity Edelweiss Creek Clear-cut 188 4.16-38: Calibrated Flows, Simulated and Observed Chemical Concentrations 240, Dennis and Edelweiss Creeks, 1987-90 192-203 4.39: Calibrated Streamflow and Sodium, Lower and Upper Edelweiss Creek April to August 1990 203 4.40: Lower Soil Head and Deep Groundwater Outflow, 240 Creek 1988-90 205 4.41: Groundwater Head and Deep Groundwater Outflow, Dennis Creek 1988-90 205 4.42: Upper and Lower Soil Groundwater Head and Groundwater Outflows Edelweiss Clear-cut 1989-90 206 4.43: Groundwater Head Dennis Creek and Groundwater Outflow Upper Edelweiss Creek 1990 206 4.44-49: Model Verification Graphs 240, Dennis and Edelweiss Creeks 214-216 xi Acknowledgments First I would like to thank the members of my committee. Drs. Doug Golding, Mike Feller, Hamish Kimmins, Leslie Smith and Jack Cheng for their guidance. Thank-you also to Jan deVries, who was on my committee prior to retirement. Special thanks go to Doug Golding for his excellent supervision and to Jack Cheng for getting me involved in this research. Many people assisted with field and lab work. Most notable were Helene Carre, Tim Sampson, Aaron Roth, Rob Wilson, Greg O'Neill, John Tyler and George Nercessian. Also, many thanks to Rita Winkler, Fraser Russell (Kamloops Forest Region), Garth Humphries (Water Survey of Canada), and Gary Kennedy (Weyerhaeuser Canada Limited). The patience and generosity of Terry Rollerson is greatly appreciated, which made it possible to complete this work. The support of my family throughout all this has made this work bearable. Helene and Melissa were particularly patient. The last year has been stressful for all of us but we made it. I am now rewriting the acknowledgements for the fourth time. Perhaps it was arrogance that led me to include acknowledgements in each draft The first fame, I wished it would be the last. Alas, this was not to be. The second time I rewrote this paragraph was after spending the Labour Day long weekend making what I hoped were final revisions, while listening to Heavy Metal and drinking Scotch whiskey. The third time, I was still hopeful about the finality of the revisions, but this time, its true! Its all done, defended, and this really is the final version, and I seem to have a whole new outlook on life. "This new learning amazes me....Explain again how sheeps' bladders may be employed to prevent earthquakes." - Monty Python Monty Python and the Holy Grail 'It is not heresy, and I will not recant.' - White Zombie Super-Charger Heaven 'Well, she turned me into a newt!" - Monty Python Monty Python and the Holy Grail xii 1 CHAPTER 1 : INTRODUCTION The purpose of this research is to investigate the processes involved in the generation of stream water quality and its response to inputs of water from snowmelt or rainfall in four small subalpine drainages in the Upper Penticton Creek (UPC) watershed. The role of hydrologic flow paths in governing water chemistry of streamflow is key to this investigation. This will be done by identifying the components that make up streamflow and studying how the chemistry of those components changes with varying hydrologic conditions. The relationships that are derived from those investigations will then be used as the basis for a water chemistry simulator that can be used (with appropriate development) as a tool to predict the effects of forest harvesting on the chemistry of streamflow and its components. The study drainages are in the Engelmann Spruce-Subalpine Fir (ESSFdc) Biogeoclimatic Zone that was originally described by Krajina (1969). The watersheds have different forest cover types including mature Engelmann spruce - subalpine fir (Picea engelmannii, Abies lasiocarpa), lodgepole pine (Pinus contorta) stands 80-140 years of age, and regenerating clear-cut areas with about 15 years of mixed pine, spruce and fir regeneration. This provides the opportunity to study the effects of forest cover type on streamflow and water quality production in headwater basins. The knowledge gained in this study will provide input to modeling procedures designed to predict the effects that forest manipulation might have on water flow and quality in the dry cold Engelmann Spruce - Subalpine Fir (ESSFdc) subzone of the southern interior of British Columbia. This study was conducted as a part of the Upper Penticton Creek Watershed Experiment, which was initiated to provide guidelines for logging on the Okanagan Highlands to minimize detrimental impacts to the hydrological regime and water quality of streams in that region (Cheng, 1982). Funding for this study was provided initially by the Kamloops Forest Region in the form of a contract to assess baseline water quality conditions for the three experimental watersheds 240, 241 and Dennis Creeks, and was later supplemented by a GREAT award. The 2 terms of the contract were fulfilled in Golding & Hudson (1991) by providing an analysis of water chemistry in streamflow. The current research involved studying hydrological characteristics and water chemistry of groundwater, precipitation (including rainfall, snowfall and melting snow packs) and direct runoff components of interflow seepage and surface runoff as well as streamflow. These components and processes were studied under the different forest cover types that are present in the watershed. The three main experimental watersheds were gauged in 1982-83 to determine baseline conditions of water flow and quality prior to the application of harvesting treatments. At the time of writing this thesis, the watersheds remain in an unlogged state. To study an area with recent disturbance, a small first order drainage with a recent clear-cut occupying a portion of its area was instrumented. A complete description of the instrumentation can be found in Chapter 2, "Methods". The results of the studies of the individual components are then combined in a final analysis to determine the relative importance of the various flow paths in streamflow under each different forest cover type. An existing hydrograph simulator is used to accomplish this, with the result that an empirical, component-based water-chemistry model is developed. 1.1: LITERATURE REVIEW The study of water quality and nutrient balances in forested research basins has been given much attention in the past, and those studies are continuing today. Past studies were concerned primarily with ecological consequences of nutrient cycling and the effects of altering those balances, whereas more recent studies are concerned with the role of hydrologic flow paths in buffering acid precipitation, and in determining the consequences of timber harvesting in community watersheds and in important fisheries streams. 1.1.1: Controls on Water Chemistry Gibbs (1970) suggested that the world's water chemistry is controlled by precipitation, mineral weathering and evaporation-crystallization, and that other mechanisms are minor in comparison. Vitousek (1977) suggests that in humid climates, evaporation-crystallization is 3 replaced by evapotranspiration, and that those control processes affect specific chemical species differently. In the northeastern United States sulphate and chloride are controlled by precipitation and evapotranspiration; sodium, silica, magnesium and calcium are controlled by mineral weathering, while nitrate and potassium are controlled by plant uptake. He also demonstrated that the concentrations of those species tend to decrease with increasing elevation, presumably due to decreased transpiration and weathering rates. Johnson and Reynolds (1977) showed that bedrock type has a major influence on stream chemistry in its weatherability. Plutonic rocks (e.g. granite) are acidic, with anions dominated by sulphate and chloride and are least weatherable, producing low concentrations of base cations and silica. Highly weatherable sedimentary rocks, such as shale, produce the highest concentrations of base cations with the dominant anion being bicarbonate. Watersheds with a combination of plutonic and metamorphic rocks were found to produce stream concentrations of base cations slightly higher than those with only plutonic bedrock. It was also noted that watersheds with sedimentary and metamorphic bedrock types experienced closer to neutral pH and slightly lower outputs of nitrate than those with plutonic bedrock, leading to the idea that soils derived from sedimentary bedrock are better able to buffer acid inputs, and may also be more productive in terms of vegetative growth. The notion of weatherability is also extended to latitude as well as bedrock type. Granitic catchments in New Mexico and the Sierra Nevada Mountains were also studied and found to produce higher levels of base cations than the geologically similar New England sites. Cornwell (1992) reported that cation exports from Alaskan arctic watersheds were the "lowest on record". This leads to the conclusion that the effectiveness of weatherability due to climate is also a factor influencing stream chemistry. Cronan et al. (1978) proposed that ecosystems could be divided into categories based on the dominance of SO4, HCO3 or organic anions in streamflow. Studies conducted at Hubbard Brook (e.g. Likens et al, 1970) showed that sulphate is the dominant anion at that location. Cronan et al. suggest that this could be attributed to heavy acidification on already acidic terrain. 4 Thorne et al. (1988) propose that bicarbonate dominated ecosystems be subdivided into those where HCO3 is produced by the dissolution of CO2 from soil respiration, and others where HCO3 is generated from weathering of carbonate rocks, the latter being better buffered than the former. Studies conducted at Sleepers River in Vermont on carbonate bedrock showed that the greater output of base cations than at Hubbard Brook was due to richer parent material, and that HCO3 was the dominant anion in spite of similar inputs of atmospheric derived sulphate. Paces (1986) studied two forested catchments in central Europe with different degrees of acidification. Outputs of cations occurred from the exchangeable pool in soil at rates that were accelerated by acidification. Replacement of cations by mineral weathering was also accelerated, but the rate of cation depletion was greater than for replacement by weathering. Depletion rates for both catchments were related to the degree of acidification; in the basin with the higher depletion rate, all spruce trees had died. Many studies have been conducted to determine the processes that occur to modify the chemistry of runoff by contact with soil. There is more than one explanation for how this process occurs. Hazlett et al. (1992) in a study at Turkey Lakes watershed in northern Ontario noted high concentrations of hydrogen ion, calcium, sulphate and nitrate in forest floor percolates during the early phases of snowmelt, with a subsequent decrease in concentrations as the melt season progressed. High chemical production at start of snowmelt was attributed to over winter mineralization of sulphur and nitrogen in the forest floor, with the subsequent decline in concentrations attributed to progressive leaching of those chemicals. These trends were also observed in the mineral soil, with the exception of sulphate. The initial pulse of H+ from snowmelt was thought to have mobilized the Ca ions from cation exchange sites in the mineral soil. As snowmelt progressed, pH of forest floor and mineral soil percolates increased, presumably due to neutralization by base cations. There was a corresponding increase in sulphate in mineral soil percolate that was attributed to mineralization of sulphur by microbial action in the mineral soil. However, another explanation, suggested by the results of Nodvin et 5 al. (1988), is that sulphate is desorbed from mineral soil particles as the soil pH increases. These study sites contain deciduous forests that produce copious litter fall in early autumn. Soil freezing did not occur and as noted by Hendershot et al. (1992), this litter is easily mineralized. Thus the process of mineralization in the forest floor under coniferous non-deciduous vegetation and/or in the presence of soil freezing is called into question. McKnight and Bencala (1990) found that in-stream processes can significantly modify concentrations of aluminum, iron and dissolved organic material over short distances of the order of 100 metres. Drake and Ford (1974) demonstrated that seasonal variations in stream water chemistry in the North Saskatchewan and Athabaska Rivers draining the east slopes of the Rocky Mountains can be explained by the mixing of two water bodies; a groundwater body dominated by rock weathering with dilution due to snowmelt. 1.1.2: Evolution of Modern Theories of Runoff Mechanisms The development of modern theory of runoff generation has been reviewed by various authors in some detail (Bonnell. 1993; Dunne, 1978; Kirkby, 1988; Beven and Germann, 1982). Horton's (1933) theory of overland flow induced by precipitation in excess of infiltration has been abandoned for forested watersheds where infiltratioh rates are typically much higher than rates of water input (e.g., Cheng, 1988). Whipkey (1965) proposed that subsurface stormflow occurs where land is sloping, the upper soil is highly permeable and there is an impeding layer in the soil. In the variable source area concept proposed by Hewlett and Hibbert (1967), the channel system expands during storms to tap an expanding subsurface contributing area. The water that reached the channel was pre-event water displaced by piston action of the new water acting on the water in the soil matrix (translatory flow). This concept has been criticized by Freeze (1972) in that matrix flow is too slow to produce rapid runoff by subsurface pathways. As an alternative, the partial area concept was developed in which storm runoff is generated as saturation overland flow over saturated soils adjacent to stream channels. Dunne and Black (1970) were able to artificially induce saturation overland flow at the base of a slope, however it 6 did not occur in response to naturally occurring storms. The variable source area and partial area concepts did not take into account rapid subsurface runoff from soil pipes or macropore channels. Rapid subsurface stormflow via soil pipes or macropores was documented by Weyman (1973), Pilgrim et al (1978) and Mosley (1979), using electrical conductivity and dye tracers to demonstrate that storm runoff consisted at least partially of new water as opposed to old water. On the other hand, studies that used oxygen isotope ratios to divide storm runoff into old and new water (e.g. Sklash & Farvolden, 1979) determined that large proportions of storm runoff consisted of old water. Thus, there was contradictory evidence from these studies that calls into question the validity of runoff theories based solely on the mixing of two essentially fixed-components. More recently, Maule and Stein (1990) used two tracers, oxygen isotopes and silica, to divide stream water into four distinct components; surface, recent subsurface, old + new subsurface and old subsurface. These components can be thought of as analogous to precipitation, interflow/shallow groundwater and deep groundwater. Current thinking includes pipe throughflow (interflow, bypass flow) as an important contributor to runoff. In reality, runoff occurs as a combination of some or all of the above mechanisms depending on the site and the hydrological conditions. The following is a review of some of the more recent research concerning the roles of those runoff mechanisms in generating streamflow and buffering inputs of acidic precipitation. 1.1.3: Current Research on Runoff Pathways and their Role in Stream Water Chemistry. Pipe throughflow is common in British Columbia forest soils and can be readily observed issuing from the cut slopes of forest roads. This phenomenon has been documented by Cheng (1988), Cheng et al (1975), deVries and Chow (1978), Feller and Kimmins (1979) and Hetherington (1982) in coastal B.C., where slopes tend to be steep and rapidly drained. Preferred pathways for rapid subsurface flow tend to develop at the interface between the forest floor and mineral soil, at the interface between soil and compacted till or bedrock, or through macropores 7 created from rotted root channels. These preferred pathways form an interconnected network that bypasses the soil matrix, and form the primary route for rapid runoff in coastal B.C.; surface runoff is extremely rare since infiltration capacities and saturated hydraulic conductivities tend to be about two orders of magnitude greater than maximum rainfall intensifies (Cheng, 1988) and soil freezing is rare. Hetherington (1995) measured rates of rapid subsurface flow through soil pipes at 0.4 X10-3 to 1.6 X10-2 m/s. These figures are 2-3 orders of magnitude greater than typical hydraulic conductivities of forest soils. Several studies have focussed on infiltration and throughflow as influenced by ground frost. Espeby (1990) used both electrical conductivity and I 80 / I6O ratios as tracers to determine the movement of acidic snow melt in a frozen till slope. He found that when the soil was frozen, bypass flow occurred by way of continuous soil pipes. The chemistry of runoff thus produced was almost identical to that of precipitation. Once the soil thawed, the soil matrix became saturated and the runoff that occurred was then altered by soil contact. In contrast to this, Roberge & Plamondon (1987) found that pipe throughflow occurred only when the water table rises to the surface of the mineral soil. The rate of pipe flow was strongly dependent on the groundwater level. Soil pipes were more or less continuous, but only operated for a few days around the peak of snowmelt. It was considered a minor component on the seasonal scale, but important in context of peak runoff. The outflow from pipes was relatively unaltered by contact with soil. Ground frost was limited to the upper 20 cm of mineral soil. Apparently, water infiltrating into dry frozen soil does not freeze into a concrete frost layer, and this may be attributable to the different thermal conductivities of water and dry soil solids. Kane & Stein (1983) conducted infiltration tests on frozen silt-loam soils in Alaska. Water movement into the soil was influenced by the ice content; ice layers in the soil were primarily responsible for inhibiting flow. Hydraulic conductivity can be effectively increased by soil freezing if the soil is dry because it becomes more granular; however, infiltration is greatly reduced if the moisture content is high when freezing occurs. Because ice becomes part of the 8 solid phase of the soil, its presence effectively alters both porosity and permeability. Barry et al. (1990) found that this change in effective porosity due to ground frost had to be accounted for when using the hydrology model HYFOR to simulate water movement in frozen soil at Lac Laflamme (see Roberge & Plamondon, 1987, above). Similarly, Johnsson & Lundin (1991) used a coupled soil water and heat flow model to simulate the effects of soil frost on infiltration and drainage. They concluded that drainage was strongly influenced, though not prevented, by soil freezing. The model underestimated drainage and consequently overestimated overland flow. This behaviour was attributed to preferential macropore flow that bypassed the frozen soil matrix. Prevost et al. (1990) observed that at Lac Laflamme, patches of bare ground began to appear about the same time that the water table rose to near the surface, and surface soils were near saturation. Night-time temperatures fell as low as -10°C and resulted in local development of concrete frost that impedes infiltration. This was found to occur at many sites in late spring, inducing overland flow. Hendershot et al. (1992) found that rapid runoff that occurred as overland flow was significantly altered by contact with the soil. During early snowmelt, runoff was dominated by deep groundwater with relatively high pH buffering capacity. At peak runoff, the stream water chemistry approached that of the less well buffered upper soil, suggesting lateral saturated flow through that zone was the dominant flow component. It was noted that the forest floor and mineral soil affected the chemistry of runoff in different ways, and outflows from the different zones were used to explain chemical changes in streamflow during runoff. Thus it appears that the forest floor has the ability to modify runoff as discussed above, tending to increase acidity, whereas rapid subsurface flow can occur through upper soil matrices that have the ability to buffer acidity and to control basic cations, bicarbonate and sulphate by adsorption-desorption and precipitation-dissolution reactions. On the other hand, the occurrence of subsurface bypass flow via continuous soil pipes can pass runoff to a surface water body relatively unchanged, particularly when the soil matrix is frozen. g One purpose of studying the role of flow pathways in the production of streamwater chemistry is to determine the ability of those flow paths to modify the chemistry of the input In part, the reason for this interest lies in the ability of ecosystems to buffer inputs of acid precipitation. Denning et al. (1992) studied the chemistry of runoff and streamflow in an alpine bedrock and talus dominated watershed in Colorado. Only 5-6% of the area was occupied by forest soils, located at the bottom of the catchment. Despite the limited extent of soil, much of the runoff that was generated from snow and glacier melt passed through the soil, and thus it was effective at buffering potential acid precipitation impacts. Jacks & Paces (1987) documented chemical changes of snowmelt along its pathway in mini-catchments. They found that the chemistry of snowmelt was altered very little by contact with granitic gneiss bedrock. When the water came in contact with organic matter, pH dropped and concentrations of all dissolved substances increased. However, when a combination of outflows from organic layers and bedrock reacted with till soils, pH was effectively neutralized, nitate and ammonium concentrations fell and concentrations of base cations, silica and bicarbonate increased substantially. A watershed in karst terrain was studied to determine the effects of shallow (rapid) and deep (slow) runoff pathways on water chemistry during storms (Mulholland et al.., 1990). Because of the karst terrain, runoff was always dominated by subsurface runoff. During storm runoff, pre-event baseflows were diluted by addition of new water, resulting in high-flow depressions in base cations, bicarbonate and silica, and corresponding increases in sulphate and nitrate. During storms where shallow runoff pathways dominated, the concentrations of base cations and bicarbonate were more than 30% lower than in storms where runoff was dominated by deep flow paths. Near surface flow paths were increasingly more important for more intense storms where antecedent soil moisture was high. Stottlemeyer and Toczydlowski (1991) studied flow paths and stream chemistry during snowmelt near Lake Superior. During early melt, most snowmelt moved down through the soil 10 profile causing the water table to rise, accompanied by a small component of pipe throughflow. Around the period of peak flow, the water table rose to the surface and diurnal lag time was reduced, providing evidence that surface runoff was at least one third of streamflow. During early melt, minor decreases in bicarbonate and base cations occurred accompanied by small pulses of nitrate and ammonium due to the small component of throughflow. During peak melt, base cations declined by 35% of the concentration in base flow, accompanied by large pulses of nitrogen, phosphorus and aluminum that were produced within the forest floor. Many of these studies (e.g. Hendershot et al., 1992, Stottlemeyer and Toczydlowski, 1991, Roberge & Plamondon, 1987) indicated that during the early phases of snowmelt, rapid groundwater recharge occurred such that streamflow was dominated by groundwater. In later phases of snowmelt, and particularly during peak flow, surface runoff and pipe throughflow became more important Prevost et al. (1990) used the variable source area simulator VSAS2 (Bernier, 1983) to simulate streamflow due to snowmelt runoff. They found that the simulator represented the first 2-3 weeks of runoff but severely underestimated peak flow. Overland flow was observed to occur around peak flow. This suggests that the variable source area concept is valid for early snowmelt runoff where streamflow is generated primarily by recharging groundwater, but that peak flows also include direct surface runoff and/or subsurface pipe throughflow that bypasses the soil matrix. 1.1.4: Chemical Changes in Snowpacks During Melt Several studies have been conducted in eastern Canada and New England to study chemical changes in snowpacks during melt. The primary aim of these studies is to determine the severity of the acid shock effect on surface water bodies. Semkin & Jefferies (1986), English et al. (1986) and Hazlett et al. (1992) studied the chemistry of H, NO3 and SO4 at Turkey Lakes watershed in northern Ontario. All studies noted that there was a rapid release of those ions early in the melt period; a laboratory simulation by Colbeck (1981) yielded similar' results. Semkin & Jefferies (1986) noted that more than 50% of those ions were lost from the snowpack 11 during the first 30% of the melt period and observed that all three ions behaved similarly, however Hazlett et al. (1992) documented selective release of ions from the snowpack during early melt. In a similar study conducted in Norway, Johanessen and Henriksen (1978) also reported selective ion release with 50-80% of the total amount of each component measured lost during the first 30% of snowmelt. English et al. (1986) determined that the ions were diverted downslope by ice lenses within the snowpack, resulting in reduced ion loading at the top of the slope and increased loading at the base of the slope. This implies that different flow paths might be affected differently by ion loading. Rascher et al. (1987) studied the same ions as well as base cations and chloride in snowpacks and forest floor leachates in New England with similar results; ions were selectively released from the snowpack over a period of about eight days during early melt resulting in a series of pulses of concentrated ions. The pulses of acidic ions were enhanced by processes that occurred in the forest floor, whereas the calcium pulse was absorbed by the forest floor. Chemical stratification within the snowpack was also observed; concentrations in the lower snowpack strata were lower than the concentrations in bulk precipitation, whereas the surface stratum was enriched in ions relative to bulk precipitation. A reason for this was not suggested. Kelso et al. (1986) studied the response of several headwater lakes near the Turkey Lakes site to the pulse inputs of acidic ions, and found a response that depended on the acid buffering capacity of the lakes. Lakes with low alkalinities showed large pH depressions, whereas other lakes with high alkalinities showed only minor drops in pH, although all lakes experienced sharp drops in alkalinity in response to rapid snowmelt. All lakes had chemical characteristics consistent with alkalinity decreased by acid loading. 1.1.5: Impacts due to Forest Harvesting J The impacts of forest harvesting on nutrient budgets and their subsequent effect on stream water chemistry were first documented by Likens et al. (1970). A sub-basin of Hubbard Brook in New Hampshire was entirely clear-cut and vegetation regrowth suppressed with 12 herbicides for a period of two years following harvesting. This treatment resulted in large increases in base cations, hydrogen ion (i.e., reduced pH), nitrate and chloride, as well as an 8-fold increase in Al 3 + . There was a corresponding decrease in sulphate. Likens et al. (1978) presented a mechanism to explain the increases. The temperature and water content of the forest floor were increased following clear-cutting due to reduced shade and evapotranspiration, resulting in increased rate of organic matter decomposition and nitrification (i.e., conversion of organic N and NH4 to NO3 and NO2). Nitrification produces hydrogen ion, lowering stream pH and mobilizing base cations. The occupation of cation exchange sites by H + causes sulphate to adsorb to the soil, explaining the reduction in sulphate in streamflow (Nodvin et al, 1988). An alternate explanation is that reduction of soil pH increases anion exchange capacity, and causes Al 3 + to appear at cation exchange sites, allowing increased sulphate adsorption (T. Ballard, pers. comm.). As vegetation was allowed to regrow, streamwater ionic concentrations recovered to pre-treatment levels by 7-8 years after harvesting. In comparison, Martin et al. (1984) studied 15 streams in New England (some unlogged, some with recent logging varying from 16 to 100 % of their area) with different results. The study was based on comparative sampling of creeks logged within 2 years of the study, with nearby uncut watersheds with similar characteristics. The authors did not find the kind of impacts to stream chemistry that were found at Hubbard Brook. Some logged streams experienced slight changes in pH, Ca and Mg but not others. Nitrogen in streams rose following clear-cutting only in the vicinity of the White Mountains where Hubbard Brook is located. The failure to detect significant chemical changes was attributed to the following factors; most watersheds were harvested in stages, most were not 100% clear-cut so that nutrient uptake in uncut areas or streamside buffer strips may have moderated the impacts in the harvested areas, and soil types in the study creeks may have buffered the impacts. Martin et al. (1985) stress the role of patch or strip cutting or the use of buffers wider than 9 m on each side of the stream in reducing the impacts to stream chemistry. Streams that had been logged to the bank or with buffer strips that 13 were too narrow suffered increases in water temperature and loading of logging introduced detritus. The increased temperatures resulted in increased mineralization rate of the introduced organic matter. This resulted in an increase in insect populations, but may also have contributed to the increases in nutrient loading following harvesting that were noted above. One obvious problem with the study that was not discussed is that it represents one limited period of time with no pre- or post-treatment water chemistry data on any of the creeks. In fact, Hubbard Brook has been identified by Thorne et al. (1988) as having sulphate influenced terrain due to heavy inputs of acid precipitation on base-poor soils, and this may have made it highly susceptible to logging related changes in stream chemistry. Other watersheds in which the major anion is bicarbonate will tend to be better buffered. Nitrate export is one of the most important indicators of watershed disturbance due to logging. Binkley and Brown (1993) reviewed several North American studies that examined the impacts of forest management practices on nitrate concentrations, as well as temperature, dissolved oxygen and suspended sediment. About 70% of the studies reported that mean annual nitrate concentrations were below 0.5 mg/1 for both harvested and control watersheds. Exceptions were in the red alder (Alnus rubra) and Douglas^ fir (Pseudotsuga menzesii) forests in Oregon where equally high levels of nitrate existed in the control and treatment watersheds, and in the hardwood forests of Hubbard Brook where nitrate levels increased significantly after conventional strip-cut logging. While in many cases, significant increases in mean annual nitrate concentration occurred after logging, the Hubbard Brook case was the only one in which the increase was serious enough to constitute degradation of water quality. However, the use of mean annual concentration as an indicator of disturbance may have been deceiving, since nitrate export tends to exhibit pulse behaviour. Indeed, pulses of nitrate that exceeded the American drinking water standard of 10 mg/1 were observed at the harvested Hubbard Brook watershed and in the alder/fir forests. Vitousek et al. (1982) conducted field and laboratory studies to determine the causes of the above noted differences between sites in terms of nitrate export. 14 Study areas included New England, Indiana, New Mexico and Oregon. The field studies consisted of trenched plot (analogous to clear-cut) and control sites to determine nitrate movement through forest floor and mineral soil horizons, and the laboratory studies determined the nitrogen mineralization potential. The results of these studies showed that nitrogen mineralization potential was highest in the northern hardwood forest of New Hampshire, where root uptake is also more important than at other sites. This would explain the impact that forest harvesting had on nitrate concentrations at Hubbard Brook. Nitrification potential is related to the abundance of bacteria that convert ammonium to nitrate. At sites that lack these types of bacteria, there would be a delay following disturbance in the conversion of ammonium to nitrate. This would therefore explain the general resistance to increases in nitrate export that were noted at most sites in the studies by Martin et al. (1984, 1985). However, those studies are not conclusive concerning the impact of forest harvesting on nitrate concentrations in streams since they only examined watersheds that had been logged two years prior to the study. 1.1.6: Summary In humid temperate areas, water chemistry is controlled by mineral weathering, precipitation chemistry and biological activity. Bedrock type and climate control weathering, with plutonic (e.g., granite) bedrock being the least weatherable, producing the lowest levels of base cations, bicarbonate and silica. In contrast, sedimentary rocks (e.g., shale/carbonate rocks) are most weatherable and metamorphic rocks (e.g., gneiss, schist) are intermediate. Biological activity can produce additional bicarbonate from soil respiration, and alter concentrations of chemical species by plant uptake. Temperature and moisture availability also affect weatherability. Weatherability relates to chemical buffering capacity; for example, granitic terrain in a cold dry climate will be least well buffered, whereas sedimentary terrain in a warm moist climate will be well buffered. Clearly, mountainous terrain in southern Canada and northern United States falls between these extremes. Precipitation affects stream chemistry in its quantity (hence effectiveness of weathering) and its chemistry. Precipitation tends to be acidic; in the absence of air pollution, naturally occurring nitrate is the dominant anion. High levels of 15 sulphate and corresponding increases in hydrogen ion from anthropogenic sources cause acid shocks from snowmelt in areas dominated by granitic terrain. This is particularly important in areas such as the Canadian Shield in central Ontario (e.g. Turkey Lakes) or the White Mountains of New Hampshire (e.g. Hubbard Brook) where bedrock is granitic and there is abundant anthropogenic atmospheric sulphate. At these sites, sulphate has been identified as the major anion involved in cation leaching. Nearby sites with carbonate bedrock (e.g. Sleeper River) are better buffered from acidification. Sedimentary and metamorphic bedrock types are more common in western North America and atmospheric acidification is generally less severe than in the great lakes lowlands, but local sources of air pollution combined with acidic bedrock puts some western ecosystems at risk from acidification (e.g. Feller, 1986). The reviewed studies stress the effect that different flow paths have on water chemistry. The forest floor appears to enhance the production of1 acidic ions (nitrate, sulphate, hydrogen ion) due to biological mineralization processes. In contrast, groundwater flow paths are primarily responsible for pH buffering of precipitation inputs by providing base cations produced by mineral weathering and alkalinity produced by dissolution of carbonate rocks or by soil respiration. These base ions can be mobilized by acidic percolates from the forest floor. Shallower (hence faster) flow paths show a lesser degree of buffering than deeper flow paths, and it has been suggested that this is due to the shorter residence time of water in those shallow paths. Pipe throughflow involves a further reduction in the degree of chemical modification that occurs along the flow path, although different authors report different levels of chemical modification of pipe throughflow by soil contact. The differences appear to be related to whether or not the surrounding soil matrix is frozen, whether or not the pipes are continuous and the flow rate through the pipes. Studies suggest that these different flow paths act individually and in combination to generate the chemistry of streamflow. Forest harvesting has been shown to reduce pH, to increase the production of nitrate and basic ions, and to decrease the output of sulphate. These changes in streamflow chemistry have been attributed to alteration of mineralization and nitrification processes in the forest floor, and 16 to changes in the relative roles of flow pathways in producing runoff. Acid precipitation has been shown to reduce buffering capacities of aquatic ecosystems over time by gradual depletion of base cations and alkalinity, with increasing dominance of sulphate as the main leaching anion. The literature suggests that the changes in stream chemistry that result from forest harvesting and acid precipitation are similar in the case of pH, and cation and nitrogen export, but opposite in the case of sulphate export. Hydrologic flow paths, soil and bedrock mineralogy and weathering rate are important factors that determine the ability of an ecosystem to resist chemical changes due to acidification or forest harvesting. These processes have important implications from the perspectives of water quality (community water supply and fisheries habitat) and forest health. 1.2: SOME SNOWMELT-RUNOFF MODELS The purpose of this section is to review some commonly used snowmelt-runoff models and to discuss their advantages and disadvantages. There are three basic methods for modeling snowmelt: energy balance, temperature index and degree day methods. 1.2.1: Snowmelt 1.2.1a: Energy Balance approach The basic energy balance method is based on the energy available for melt at a point as given by Male and Gray (1981): Qm = Qsn + Qin + Qh + Qe + Qg + Qp - dU/dt where Qm = energy available for melt Qsn = net short-wave radiation absorbed by the snow Qin = net long-wave radiation at the snow surface Qh = sensible or convective heat flux at the snow surface Qe = latent heat flux at the snow surface (+ = condensation) Qg = flux of heat from the ground at the snow-ground interface Qp = flux of heat from rainfall dU/dt = rate of change of energy stored in the snowpack. 17 Most of the above terms operate at the snow surface, although small amounts of short wave radiation may penetrate the pack. The ground heat flux is also small or non-existent so that melt is primarily generated at the snow surface. For an isothermal (melting) snowpack, the rate of change of stored energy is zero. Snowmelt models include a factor for cold content so that at times when melt is occurring, the last term can be neglected. Thus, the energy balance equation during melt can often be simplified as Qm = Qn + Qh + Qe + Qp where Qn is the net all wave radiation. While the energy balance method is generally considered the best approach because it is physically based, in practice it is difficult to measure the various radiation components. Measurement of net longwave radiation requires instrumentation that is difficult to maintain, and the sensible and latent heat fluxes must be calculated from meteorological data. Thus, various authors have developed simplified energy balance approaches. The only radiation component that is easy to measure is the incoming short-wave radiation. Anderson (1968) gives a formula to calculate the net all-wave radiation balance (Qr) as: Qr = Qsn + Qln = Qi(l-a) + Qa - sTsk4 where Qi = incident solar radiation a = albedo of snow surface Qa = incoming long-wave radiation s = the Stephan-Boltzmann constant at 11.71 X 10-8 langleys/day/°KA Tsk = temperature of the snow surface in ° K The albedo of the snowpack depends on the age and liquid water content of the surface snow. Since albedo affects net short-wave radiation, it can be measured directly using a radiometer facing down towards the snowpack. However, the most commonly used approach to assess albedo is the aging curve approach proposed by USACE (1956). The terms Qh and Qe are governed by complex turbulent energy exchanges that operate within the lower 3 metres of the atmosphere, and cannot be measured directly. Sensible heat transfer is governed by the temperature gradient. When air is warmer than the snow, heat is 18 transferred to the snow surface and vice versa. Similarly, latent heat transfer is controlled by the vapour pressure gradient When vapour pressure of the air is greater than that of the snow surface there is a transfer of moisture from the air to the snow, accompanied by a release of latent heat, and vice versa. Both are turbulent processes so their estimation involves wind speed. Anderson (1968) gives equations to calculate those fluxes as follows: Qh = f(u)'(Ta-Ts) Qe = f(u) (ea - es) where Ta and T s = the temperatures of the air and snow surface respectively in °C ea and es = the vapour pressures of the air and snow surface respectively in mb. f(u) and f(u)' = appropriate wind functions for latent and sensible heat transfers. Equations to calculate the input of heat from precipitation are given as Qsnow = P T w C i where, P = amount of precipitation T w = wet bulb temperature cT, C i = specific heats of water and ice Wind functions are derived from testing the equations against snowmelt lysimeter data. Temperature of the snow surface is also an important parameter that is difficult to measure continuously. Price and Dunne (1976) stated that Anderson's (1968) equations ignore the role of atmospheric stability in turbulent heat transfers. They proposed the following modification: Qh = ra C p Dh(Ta - Ts) Qe = hf ra De(0.622/p) (ea - es) where C p = specific heat of air p = atmospheric pressure (mb) ra = density of air hf = latent heat of fusion in cal/g and Dh, De = exchange coefficients for sensible and latent heat in cm/hr. These coefficients vary with atmospheric stability such that under neutral conditions, 19 [ M z / z 0 ) ] 2 under stable conditions (Dh)s = Dh /(1+qRi) and under unstable conditions (Dh)u = Dh(l- qRi) where k = von Karman's constant U z = wind speed at height z in cm/hr q = an empirical constant Ri = Richardson number zo = roughness length in cm due to obstacles estimated by zo = h* S/2S' where h* = obstacle height S = silhouetted area and S' = basal area of the object. Independent research conducted about the same time by Granger et al. (1977) used turbulent heat exchange equations of similar form. Price and Dunne (1976) measured incoming, diffuse and net all-wave radiation as well as wind speed, ambient and wet bulb temperatures at 2m above the surface. It was assumed that the temperature of the snow surface was 0 °C during active melt, and equal to the wet bulb temperature at other times. This approach yielded calculated hourly melts that were not significantly different from measured values. Granger et al. (1977) also measured temperature profiles of soil and snowpack, soil heat flux and seven levels of wind speed, temperature and humidity between the surface and a height of 2m. They determined that the major source of snowmelt energy came from net all-wave radiation. Sauter and McDonnell (1992) applied a simplified energy balance by direct measurement of the same components listed above and concluded that net radiation contributed 94% of the heat flux responsible for melt. Using the equations discussed above, the energy for melt over a given time interval at the snow surface at a point can be modeled, assuming that all the input parameters can be measured. Anderson's (1968) equations formed the basis of a widely used energy balance snowmelt simulator (Anderson, 1976). To predict water generation from the snowpack, it is often paired with Colbeck and Davidson's (1973) approach to gravity flow of water through the 20 snowpack (see 1.2.2a). In a series of papers, Bloschl and Kirnbauer (1991) and Bloschl et al. (1991a, 1991b, 1991c) describe a spatially distributed finite difference snowmelt model based on the coupled heat and mass flow model using the approaches of Anderson (1976) and Colbeck and Davidson (1973). The model was applied in a high alpine basin in central Europe. The model involves spatial distribution of snow accumulation by linear extrapolation of air temperature and precipitation, while wind speed and relative humidity are assumed invariant across the basin. Snow is redistributed down slopes by gravity and wind. Model evaluation demonstrated that topography had a strong influence on initial distribution of snow water equivalent, and therefore affected snowmelt according to the patchiness of the snow cover and the routing pf snowmelt through deep snowpacks. Topography also affected snowmelt due to longwave emissions from surrounding terrain in parts of the study basin. 1.2.1b: Simplified energy balance approaches. The data requirements of energy balance models create problems in operational use. This has led to approaches to either simplify the method by eliminating insignificant or difficult to measure terms or to estimate the energy components from commonly available data (i.e., temperature). Dunne and Leopold (1978) give the following equations for calculating daily melt at a point under different forest cover conditions, based on incident solar radiation, temperature and wind speed: <10% tree cover M = 0.0125 Qi(l-a) + (1 - 0.1C)(0.104T2 - 2.13) + 0.013CTa + 0.00078u2 (0.42T2 + 1.51Td) , 10-60% tree cover M = 0.01(1 - F) Qi (1-a) + 0. 00078u2(l - 0.8F)( 0.42T2 + 1.51Td) + 0.14FT2 60-80% tree cover M = 0. 00078u2(l - 0.8F)( 0.42T2 + 1.51Td) + 0.14 T2 >80% tree cover M = (0.19T2 + 0.17Td) " where Qi = incident solar radiation a = albedo of snow surface C = cloud cover expressed in tenths of the sky F = fractional forest cover Ta = air temperature T2, Td & u2 = air temperature and dew point (°Q and wind speed (km/day) at 2 m above the snowpack 21 Note that as forest cover increases, terms drop out of the equations. Radiation drops out of the equation at above 60% forest cover, but the turbulent energy terms remain in the equation up to 80% forest cover. Thus, direct radiation becomes less important as forest cover increases. USACE (1956) developed equations to predict energy budget components from daily maximum and minimum temperatures. These were adapted by Quick et al. (1995) for use in the simplified energy budget snowmelt section of the UBC watershed simulator. These approximations have been developed specifically for conditions in British Columbia, resulting in the following energy budget components, in which all components are in mm/day: Qsn = (54 - 29 cos 2pN/365) (1 - C) (1 - a) Qln = (-20 + 0.94Tm) (1 - C) + 1.24 T dC Qh = 0.18(p/101)TmV Qe = 0.35(p/101)TdV Qp = K T m P where N is the day of the year, p is the atmospheric pressure in kPa, V is the mean daily wind speed in km/h and P is the precipitation input in mm. In these formulations, Ta is approximated by the minimum daily temperature, T m is the mean daily air temperature (°C) and K is a constant that represents the heat content of the rain. It is assumed that the temperature of the rain is equal to the air temperature. The turbulent heat fluxes are modified by a function of the Richardson number. Note that cloud cover and albedo reduce melt due to incoming solar radiation. This leaves the cloud cover, albedo and wind speed to be estimated from temperature. Cloud cover and wind are related to the daily temperature: (1 - C) = (Tmax - Tmin)/D R V = Vmax - (Vmax - 1) (Tmax - Tmin)/25 where D R is the maximum recorded temperature range for clear sky, and Vmax is the maximum wind speed at elevations below 2000 m. Albedo is modeled by two different aging curves. New snow ages rapidly from an initial albedo of 0.95 to a settled value of 0.65 by a factor of 0.9 per day. After this it ages more slowly according to an exponential decay curve to a minimum value 22 of 0.3. Thus, the energy components are calculated according to a set of approximations based entirely on temperature. The model is distributed by elevation band, allowing for either multiple meteorological sites within or near the watershed or on calibrated precipitation and temperature gradients. 1.2.1c: Temperature Index and Degree-Day Methods This approach is based on correlation of daily snowmelt with mean daily temperature, and as described by Dunne and Leopold (1978), M = Mf (Ta - Tb) where Tb is the base temperature above which melt will occur (often set at 0 °C) and Mf is the melt factor equal to the slope of a regression line of melt on temperature. The base temperature should be equal to the intercept of that regression Une. Obviously, daily melt must be measured periodically using a snowmelt lysimeter to develop the regression coefficient. Anderson (1973) includes a separate formulation for melt during rain periods as follows: M = [(6.1X10-io)(Ta + 273.16)4] - 3.4 + [(2.1X10-3)PrTa] + {1.4f(u)[(ea - 6.11) + (9.5X10-5)paTa]} where Pr = total daily rainfall and pa = average daily atmospheric pressure. Melt due to latent heat transfer during rain is estimated assuming that the snow surface is at 100% relative humidity at 0 °C and the air is at 90% relative humidity at Ta. The first 2 terms are incoming and outgoing longwave radiation, the third term represents rain melt, and the last terms represent turbulent heat transfers. Prevost et al. (1991) tested Anderson's (1973) temperature index based model against Anderson's (1976) energy balance model under a fir canopy in Quebec and reported that the models were equally reliable to predict hourly snowmelt outflow. Bloschl and Kirnbauer (1991) compared the same two models in an alpine catchment in the Austrian Alps and found that the temperature index model tended to underestimate snowmelt rates, whereas the energy balance model provided a good approximation of measured snow water equivalent. The difference between these findings can be explained by the effect of the forest canopy at the Quebec site that effectively negates some of the energy balance terms. 23 The most common method of predicting snowmelt in runoff simulators is the degree-day method (WMO, 1986). The degree-day factor is described by Martinec (1960) and Martinec and Rango (1981). The degree day formulation is expressed in terms of depth (cm) of daily melt: M = l.l(ps/pw)TddAd where (ps/ pw) is the density of snow as a fraction of the density of water, Tad is the number of degree days (°C day) above base temperature and Ad is daily fractional areal snow coverage. 1.2.2: Snowpack Conditions Before water can be released from the snowpack, it must become primed. This involves meeting two conditions; the snowpack must become isothermal at 0 °C, and it must have reached its liquid water holding capacity. Most authors refer to the energy deficit of the snowpack below 0 °C as the cold content, and recognize that the energy available for melt must first satisfy the cold content. All snowmelt models, whether they are energy budget or temperature index based need a method of accounting for snowpack cold content When air temperature goes below zero, this can be considered negative melt Most models involve some accounting of melt deficit based on a cumulative negative melt. That deficit must be accounted for before melt can begin or resume. For example, Quick et al. (1994) use a daily decay factor for each day's cumulative deficit so that only the previous 10 days influence cold content. \ • • • When snowmelt begins in spring and the snowpack is primed, subzero night-time temperatures will cause only the surface layers to freeze, implying that for a primed snowpack, cold content is limited to some arbitrary surface depth. USACE (1956) relates cold content of the top 60 cm of the snowpack to the average of the previous ithree days' temperature, whereas Kim et al. (1986) use a depth of 25 cm. The water holding capacity of the snowpack is usually 3-5% (Leaf, 1966), while Perla et al. (1985) suggest using 5% as the practical lower limit of liquid water content. To account for liquid water in snow, Male and Gray (1981) give daily melt produced from the snowpack as M = Qm/(rh_B) 24 where M = melt water in cm/d hf = latent heat of fusion in kj/kg r = density of water in kg/m3 B = thermal quality of snow as a fraction = 1 - (fractional liquid water holding capacity) For normal melt conditions this reduces to M = Qm/(3335B) 1.2.2a: Water movement through snow. The above formulation ignores the time delay associated with water movement through the snowpack. This may not be significant for daily melt, but for hourly melt, it should be accounted for. Many runoff models (e.g., Dunne et al,, 1976; Bloschl et al., 1991a,b,c) use the principles of Colbeck and Davidson (1973) for vertical water flow through snow under Darcy's Law in which flow is proportional to effective permeability: K(q) = KsSe3 where q is the volumetric water content of the snow, Ks is the total permeability of the snow and Se is the effective saturation of the snowpack. Se can be calculated as Se = (q -qr)/(n-q t) where q r is the residual saturation or water holding capacity as discussed above, usually 3 - 5%, and n is the porosity (= 1 - (snow density - q )). Several equations exist to calculate permeability from grain diameter and porosity, e.g. the Kozeny-Carman equation (Freeze and Cherry, 1979): k = (n3/(l-n2))(d2/180) where k is the permeability and d is the mean grain diameter. The data needed to calculate the effective permeability can be obtained from snow survey techniques, however liquid water content is difficult to measure. Kim et al. (1986) used a function of snow density to estimate liquid water deficiency. (In section 3.1 of this study a method to estimate liquid water content of upper, middle and lower snow layers as exponential functions of temperature indices based on current and the previous day's temperature is developed.) Thus, the yield at the bottom of the snowpack can be derived from the melt at the surface. In cases where infiltration is impeded by 25 concrete frost or soil saturation. Colbeck's (1974) kinematic wave description of lateral flow through a thin layer at the base of the snowpack was used in connection with the vertical flow model to predict surface snowmelt runoff from a slope (Dunne et al., 1976). 1.2.2b: Rain vs snow. All snowmelt models also include some criteria to partition precipitation between rain and snow on the basis of temperature. This is often called the critical temperature. In the UBC model, the critical temperature is a user-defined constant. For example, if 2 °C is used, then above that temperature, all precipitation is rain; below 0 °C all precipitation is snow. Between 0 and 2 °C the relative proportions of rain and snow are determined by interpolation (e.g., at 0 °C it is all snow, at 1 °C it is 50% snow and 50% rain, at 2 °C it is all rain). The Snowmelt Runoff Model (SRM) (Martinec et ah, 1983) depends on remote sensed snow cover data during the accumulation period, and therefore relies on critical temperature only during the snowmelt . period. Martinec and Rango (1986) found from direct observation that critical temperature varied from 3 °C in early April to 0.75 °C in July. > 1.2.3: Runoff Models Snowmelt models are linked to runoff models that route the melt or rain input through the watershed to streamflow. For watershed routing of snowmelt yield, any snowmelt model can be paired with any watershed simulator. For example, Prevost et al. (1990) used Anderson's (1973) temperature index SNOW-17 model as input to VSAS2 (variable source area simulator, Bernier, 1982). Most such simulators, however, are of the conceptual type that represent the watershed as one or more storage zones that have controlled outflows. All routing models have, either an implicit or explicit scheme for routing water through the snowpack to account for delays between melt production and runoff production. One of the simpler models, the SRM (Martinec et al, 1983) uses the degree-day method described above, coupled with a single lumped routing reservoir such that in its simplest terms, it can be represented by a single equation (Martinec and Rango, 1986): 26 CVl = [Fdd C s n (T„ + AT„) Ad „ + C m Pn] (1 - k„ + l ) + O n k„ + l 86400 where Q = mean daily streamflow (m3/s) F d d = degree day factor (cm/°C/day) cr, cs = runoff coefficients for rain and snow respectively T = number of degree days above base temperature (°C day) AT = adjustment to degree days due to temperature lapse rate (°C day) P = precipitation contributing to runoff (cm) A = basin area (m2) k = recession parameter n = day number The runoff coefficients must be determined from the ratios of runoff to precipitation in long term records. The degree-day factor and recession parameter can be treated as calibration parameters. Other calibration parameters include lapse rate and time lag. Areal extent of snow cover (Ad) "is also required input. The SRM model is widely used for water supply and hydropower predictions (WMO, 1986) particularly since satellite observations of snow cover became commonplace in the 1970's (Rango, 1992). To develop the model from a simulation model to a forecasting model involved the development of a set of snow cover depletion curves that relate extent of snow cover to snowmelt rather than time (Rango, 1989). There is a procedure for streamflow updating to improve forecasts. It has been tested worldwide on 52 basins and confidence in its results has led to using it to predict the effects of climate change (i.e., temperature increase) on hydrology (Rango, 1992). The SRM model involves distributed snow accumulated and melt calculations with runoff calculations that are lumped vertically and aerially. Haltiner and Salas (1988) used time series analysis to develop an ARMAX (Autoregressive-Moving Average with Exogenous Inputs) model to predict runoff one day ahead. They show that the model is of similar form to SRM but that the parameters can be estimated analytically rather than empirically. According to Rango and Marrinec (1981) the accuracy of prediction decreases with increasing watershed area using SRM. An effort to overcome this is proposed by Kim et dl. (1986) by using three non-linear recession coefficients instead of one. This in effect is the same as using three different storage 27 zones representing fast medium and slow runoff components, although the authors are careful to point out that the three recessions do not represent conventional runoff responses. The standard degree-day approach is used to estimate snowmelt More complex models involve vertically distributed runoff routing, using a series of storage zones that is intended to represent runoff components such as surface storage/runoff, interflow and groundwater. Examples of these types of conceptual models are TANK (Sugawara et al, 1984) and SSARR (USACE, 1976). Both involve temperature index snowmelt models that are distributed through elevation bands, and a vertically distributed water balance model, although with TANK, both components can be treated as lumped. Vertical distribution of runoff routing is handled by zones that represent interception, forest floor, soil, groundwater, etc. In the TANK model, storage zones are represented by a set of tanks. Each tank behaves like a barrel with a hole in the bottom and one in the side. Water enters the top tank and infiltrates to lower tanks through the bottom outlet Runoff from each tank occurs through the side outlet The top tank represents soil moisture. The bottom outlet has a pipe that raises the outflow level, creating a dead space at the bottom of the tank to represent soil moisture storage. The side outlet is above that level. Whether or not runoff is generated from any tank depends on the height of the side outlet and the rate at which water is added. Outflows from either the side or bottom outlet depend on the size of the hole and the head of water in the tank. At the upper end of model complexity are models such as HSPF (Jphanson et al, 1984) and UBC (Quick et al, 1995). Both use energy balance approaches to estimating snowmelt; UBC uses the temperature based approximations described above, whereas HSPF requires input of solar radiation and wind data, as well as potential evapotranspiration. Both are distributed models; UBC snow section is distributed through elevation bands, whereas HSPF is distributed through land segments that can be arranged according to the user's criteria. The water balance components of HSPF are also distributed both vertically and aerially by specifying water balance parameters for each land segment. UBC is vertically distributed, and the descriptive parameters are cbstfibuted by elevation band but the routing parameters are common to all elevation bands. 28 Nonetheless, a separate water balance is calculated for each band. The water routing of HSPF is perhaps the most physically based of the conceptual models with user specified parameters that govern infiltration to the soil moisture routine, evapotranspiration from soil and groundwater, surface runoff according to Manning's n, and parameters to govern storage and recession of soil and groundwater. The UBC routing model structure is somewhat different from that of HSPF but of similar complexity in the vertical direction. Water that passes through the forest canopy is apportioned between the impermeable portion (to fast runoff) and that which enters the soil moisture model. The water entering soil moisture is apportioned according to the soil moisture deficit. Fast runoff is governed by the impermeable fraction of the elevation band and a parameter that adjusts that fraction according to soil moisture deficit. Daily soil moisture deficit for each elevation band is updated by the actual evapotranspiration and water input, and must be satisfied before any further runoff can occur. Potential evapotranspiration is calculated for each elevation band from temperature, tree cover and specified parameters that account for the monthly variability of evaporation. AET depends on PET and soil moisture deficit. Once runoff is generated, it is routed to groundwater, which accepts water to a fixed limit. Any excess beyond that limit is apportioned to medium runoff (interflow). The groundwater is apportioned to upper and lower groundwater zones according to a user defined parameter. The soil moisture routine is distributed through elevation bands, each producing quantities of four components of runoff. Each type of runoff thus produced goes into a single runoff storage. The fast and medium runoff is subject to "unit hydrograph convolution" which involves cascading the runoff through a series of linear storage reservoirs. The user specifies the number of reservoirs used to accomplish this; the larger the number of reservoirs, the more sluggish the response. The fast runoff generated by impermeable area directly enters this cascade. Medium runoff is routed first through a linear reservoir in which a constant percentage of each days storage is released, and this release is then subject to convolution as described } 29 above. Upper and deep groundwater are each routed through a single linear storage reservoir directly to streamflow. s . Hetherington et al. (1995) had limited success at calibrating HSPF to data from Carnation Creek, and this may be due in part to the excessive data requirements and internal complexity of the model. The UBC model has the appropriate level of complexity to act as a research tool, while the data requirements are simple enough to act as an operational hydrology tool. As will be shown in Chapter 2, the data available at UPC make the UBC model the best choice for use in the current study. 1.3: JUSTIFICATION OF THE CURRENT STUDY Lately there has been a shift in emphasis from field based research to modeling research. This is in part related to the high cost of field studies. Bonell (1993) has identified the need for a new intensive phase in experimental hydrology with emphasis on the linkage between runoff pathways and hydrogeochemical processes, and the need for field researchers and modelers to work more closely together. He also suggests topographically based runoff models are appropriate analytical tools for the simulation of the water chemistry response of catchments. The current research addresses these issues by using an existing runoff simulator to develop an empirical component based water chemistry model. The UBC model is identified as the best choice to accomplish this at UPC. The studies reviewed (section 1.1) treat the evolution of stream chemistry as the mixing of components whose chemistry is more or less constant This research will show that hydrologic conditions affect the chemistry of those components, as well as their mixing ratios; thus, models that treat the chemistry of the components as fixed are not likely to succeed. Most of the studies that focus on the hydrochemical impacts of forest harvesting have been conducted in eastern North America; relatively few have been conducted in western Canada, where the predominant forest cover is of a different type. The process of chemical recovery after harvesting has been studied by Likens et al. (1978) at Hubbard Brook. This is an important concept as it relates to water quality in community watersheds, and therefore warrants 30 study in British Columbia's forested watersheds. This study, while it does not include a pre- and post-treatment comparison, establishes the ground work for development of a model to simulate component based water chemistry and its response to forest harvesting, and suggests recovery relationships between catchments with and without recent harvesting. 1.4: THE STUDY AREA The four study basins are called 240, 241, Dennis (242) and Edelweiss Creeks. A map is given (Figure 1.1) that shows the location of the, study area in the south Okanagan, and shows the locations of the study basins and the weirs and meteorological instrumentation. Large scale (1:15,000) maps of the study watersheds are included (Figures 1.2 and 1.3). Some key factors are given in Table 1.1 that describe the main differences between the study watersheds in terms of their basin morphology and forest cover. The average land slope was calculated according to the following formula: S = CTL/A where CI is the contour interval, L is the total length of contours and A is the drainage area. 240 Creek has a steeper average land slope than the other two,due to the steep side slopes in the Table 1.1: Basin Morphology anc Forest Cover. Creek: Attribute 240 241 242 Edelweiss average land slope 24% 17% 17% 16.4% average slope of main channel 3.4% 9.7% 7.2% 10.2% drainage density (km/km2) 1.04 1.76 1.10 1.46 stream order at weir 2 2 2 1 bifurcation ratio 5 7 4 1 elevation range (m) 1608-2005 1598-2005 1780 - 2136 1733-1892 H50(m) 1772 1734 1908 1794 Predominant aspect S SSW W SW Predominant Forest Cover Lodgepole Pine Lodgepole Pine Spruce-Fir, Pine-Fir Spruce-Fir, immature Average canopy closure 42% 45% 76% 61% % area of exposed bedrock 3.6% <1% 8.1% 0% % area with 50% exposed rock NA . 5.9% NA NA % area with bogs <1% <1% <1% <1% % area with recent disturbance 0% 0% 0% 24.8% drainage area (ha) 520 464 390 43 31 Figure 1.1: UPC Experimental Watersheds Contour Interval 30 metres Approximate Scale 1:41,000 Instrumentation groundwater site precipitation & temperature site • weir 32 Figure 1.2: 240 Creek with Forest Cover 0 500 "~ 1000 1500 " 2000 metres 33 Figure 1.3: Dennis-Edelweiss Creeks with Forest Cover R o c ^ contour interval 10 metres Pine-Fir Spruce-Fir Immature watershed boundary I L _^— perennial stream , - - * ephemeral stream bog A weir 2000 3 4 upper watershed, but this may be offset by the gentler slope in the upper section of the main channel. 241 Creek has a uniform slope dissected by stream channels, whereas Dennis Creek is more bowl shaped with steep slopes flanking Greyback Mountain which forms the highest point in the study area. The median elevation was determined using area-elevation curves (Figure 1.4) that also illustrate the character of land slopes in the watersheds. Figure 1.5 shows the profiles of the main channels. The drainage densities and bifurcation ratios (ratio of number of Is* order to 2nd order streams) reflect the relative hydrograph recession rates of the three streams, and also the lag time of the creeks. Edelweiss Creek is a first order stream with a drainage area about one order of magnitude lower than the others, whereas 240, 241 and Dennis Creeks are second order. Thus Edelweiss Creek will tend to be flashier than the other creeks due to lower channel storage, but its lag time compared to the other creeks will be governed primarily by its drainage density. The drainage densities reported here accounted for all permanent first order and higher streams, and excluded ephemeral zero order streams that carry water during high flow periods, predominantly due to snow melt. Thus drainage densities will actually be higher at high flow with a consequent decrease in basin lag time. The canopy closure figures were derived from forest cover maps, supplemented by field measurements collected using a hand held canopy densiometer, and reflect the canopy densities that existed at the time when the.study was conducted. The spruce-fir forest has a very high canopy density compared with the lodgepole pine canopy. This is due to the large canopy of shade tolerant Engelmann spruce and subalpine fir trees compared with the small canopy of the individual lodgepole pine trees, which are shade intolerant. Average canopy density under the spruce-fir stand was measured at 80%, whereas the average density of the lodgepole pine canopy was 45-50%. Average basin-wide canopy density at Dennis Creek was reduced by sparser lodgepole pine and open areas in the upper part of the watershed. Average canopy density at Edelweiss Creek was the weighted average of about 80% spruce fir and 20% regeneration with an average canopy density of about 15%. Figures 2 and 3 show the distribution of forest cover types at 240, Dennis and Edelweiss Creeks. 35 o o o o o o o o m o i n o m o i o o (LU) |9A8~| BBS SAOqV U0!)CA9|3 U O I ; B A S J 3 U S A J O aAoqv JQ »V 6 9 J V JO lusojsd 36 1.4.1: Geology and Soils The bedrock of the Okanagan Highlands has been described by Holland (1976) as gently dipping gneisses and schists of the Shuswap complex. Parkinson (1985) has further described the bedrock as consisting of granodioritic to granitic orthogneiss, a metamorphic rock formed by recrystallization of igneous rocks. Thus, the chemical composition of the rocks is typical of granites and granodiorites, which consist mainly of feldspars and quartzes (Harker, 1964) with chemical formulae (Ca, Na, K)Al2Si308 and SiC>2 respectively. This type of bedrock has extremely low primary permeability, with secondary permeability related to the fracture spacing. Soils consist of relatively thin morainal blankets and veneers and are likely derived from the local bedrock, with the same mineral composition. Soils have been classified by Smith (1984) as Dystric Brunisols with some Orthic Ferro-Humic Podzols. Soils are formed from weathering of the moraine. Feldspar minerals dissolve incongruently, which means that one of the products of dissolution is a solid secondary (clay) mineral. For example, albite tends to weather as follows: NaAlSi308 +H 2C0 3 +f H 2 0 « Na+ + HC0 3 _ + 2H4Si04 +|Al2Si205(OH)4 where the first term on the left hand side is albite and the last term on the right hand side is the clay mineral kaolinite. Similar equations could be written for other feldspar minerals. The weathering agent is water that is slightly acidic due to absorption of atmospheric CO2. The weathering products are dissolved sodium, bicarbonate, silicic acid and clay. This reaction is common in granitic terrain (Freeze and Cherry, 1979). Lawson (1968) has studied the hydrology of the bedrock at Trapping Creek, an Okanagan Highlands watershed to the north-east of the study area near Big White Mountain with geology similar to that of upper Penticton Creek. Hydraulic conductivities of the bedrock were measured using piezometer tests. The hydraulic conductivity distribution of the bedrock was generalized into two layers; an upper layer with fracture spacing of less than 15 cm extending to a depth of about 11 metres, and a lower layer with fracture spacing of 60 cm extending to an impermeable layer of unfractured bedrock. Several bedrock types were tested 37 including granodiorite, similar to that at upper Penticton Creek. The following empirical equation was given to describe the relationship between hydraulic conductivity in the granodiorite and depth: K = 8.14X10-V 0 1 2 3 M where K is hydraulic conductivity in m/s and d is depth in metres. 1.4.2: Climate and Hydrology. The hydrology of the study watersheds is typical of headwater catchments in the south Okanagan. Streamflow is driven primarily by snow melt that occurs in the spring and early summer, with peak flows occurring in middle to late May. Three-year average monthly mean maximum and minimum temperatures, monthly total precipitation and mean daily streamflows are shown in Figure 1.6. The temperature data are averages taken over five sites in the study area where ambient temperature was measured, and the precipitation data were derived from one site that was considered most representative of the study area (see Chapter 2, "Methods"). The streamflow data were derived from mean daily streamflow data at 240 and Dennis Creeks from 1988 to 1990, except that each year of record was lagged so that the peak flow occurred on the same date (May 17th) each year. The average hydrographs so derived were therefore representative of an average year. Continuous records of streamflow are not available at Edelweiss Creek, but the hydrographs of 240 and Dennis Creeks show the differences in streamflow that result from the differences between the two watersheds. Both hydrographs show a double peak in which the first peak is higher on 240 Creek and the second is higher on Dennis Creek. Dennis Creek experiences peak snowmelt later than 240 Creek as a result of differences in forest cover, aspect and elevation. Individual event peaks occur at the same time due to prevailing meteorological conditions, but the Dennis Creek hydrograph tends to lag behind that of 240 Creek. Peak flow due to snowmelt occurs in late May on 240 Creek, and tends to occur two to three weeks later on Dennis Creek. The late May peak on the hydrographs is influenced by a rain on snow event that occurred on May 27-28 in 1990. The summer peaks that occur due to rainfall are higher on Dennis Creek because of lower evapotranspiration that results from the 38 Figure 1.6: Three Year Mean Climatograph and Hydrographs 1988-90 1 — i — i — i — i — i — i — i — i — i — i — i — T 0 30 60 90 120 150 180 210 240 270 300 330 360 Day of Year (adjusted for timing of peak) 39 higher elevation and different aspect, whereas the faster recession rate on Dennis Creek is due to the higher drainage density. The driest months are February, September and October. Generally, mean daily temperatures fall below freezing in late October before heavier precipitation begins in November. The result of this is that soils are frozen before the snowpack begins to develop. Soil temperatures were monitored under the mature spruce-fir and lodgepole pine canopies and in the regenerating clear-cut. These findings will be discussed in later sections, but it is appropriate to present the results here. Figure 1.4 shows selected soil temperature profiles under each forest cover type for the 1988-89 and 1989-90 seasons. The progression of soil freezing seems to depend on elevation and forest cover. In November the surface soil layer falls below 0°C at all sites and the January profiles show that the freezing front has moved to below the 20 cm depth at all sites. By April, the entire profile is frozen despite deep snowpacks, warmer air temperatures and the onset of snowmelt. At all sites, the entire profile then remains just below 0°C until snow has melted from the site, at which time thawing then begins from the surface down. Generally, the winter soil temperatures are highest at the 240 Creek site which is the lowest elevation, on a WSW aspect under a lodgepole pine canopy. Winter temperatures at the 5 cm depth are comparable at the Dennis and Edelweiss Creek sites, despite the different forest cover (immature regeneration at Edelweiss Creek vs mature spruce-fir at Dennis Creek), probably a result of similar elevation. 40 Figure 1.7: Selected Soil Temperature Profiles, 1988-90 + 02 Nov 1988 ^ 21 Jan 1989 Temperature in degrees C A 05 April 1989 X 12 May 1989 Soil Temp. 240 Cr. Soil Temp. Ed. Cr. -1.0 0.0 1.0 2.0 3.0 4.0 -2.0 -1.0 0.0 1.0 2.0 Soil Temp. Den. Cr. -2.0 -1.0 0.0 1.0 2.0 3.0 + A X 05 Nov 1989 03 Jan 1990 16 April 1990 27 April 1990 Soil Temp. 240 Cr. -2.0 -1.0 0.0 1.0 2.0 + 22 Nov 1989 O 03 Jan 1990 A 14 April 1990 X 09 May 1990 Soil Temp. Ed. Cr. -2.0 -1.0 0.0 1.0 2.0 + 22 Nov 1989 ^ 03 Jan 1990 A 14 April 1990 X 17 June 1990 Soil Temp. Den. Cr. -2.0 -1.0 0.0 1.0 2.0 41 CHAPTER 2: METHODS 2.1: INSTRUMENTATION The locations of the instrumentation sites are given in Figure 1.1. The following includes descriptions of instrumentation to measure streamflow, precipitation, temperature (air and soil) and groundwater. 2.1.1: Streamflow. ; Streamflow at 240 241 and 242 (Dennis) Creeks was monitored by Water Survey of Canada. At each site, flow was measured using a sharp crested rectangular notch weir with a V-notch insert to improve measurement precision at low flow. Stage was measured with a Stevens A35 float actuated stage recorder. At flows greater than 0.315 m /^s the weirs are overtopped and are rated using periodic manual flow measurements. Figures 2.1 and 2.2 show examples of the weirs at low and high flow. Two different V-notch weirs were used on Edelweiss Creek to measure streamflow. Initially, a natural control was enhanced by cementing a V-notch between two large boulders that control the flow. Eventually that weir failed and a different weir was constructed in the fall of 1990 by trenching a straight section of channel and installing a V-notch weir plate and holding it in place by sand bags. For both weirs there was no continuous record of stage; only point measurements were collected. This weir is pictured in Figure 2.3. Discharge was calculated directly from the head h at the weir by the standard formula given by Linsley & Franzini (1979): Q(m3/s) = 1.37 h5/2 tan(8/2) where 9 is the notch angle of the weir and h is the head in metres. The weir that was constructed in November 1990 was considered the more accurate of the two. The formula was verified periodically by current metering, which was also used to measure the discharge at times when i there was no weir. 2.1.2: Precipitation and Temperature. Total precipitation was measured using Belfort type weighing gauges at three sites; Cheng, located below Grayback reservoir at an elevation of 1540 metres, Dennis, located on the Figure 2.1: 241 Creek weir at medium to high flow Figure 2.2: Dennis Creek weir at high flow; note snowpack on N aspect. 43 44 ridge that forms the northern watershed boundary of Edelweiss Creek at an elevation of 1760 metres and Penticton-2 at an elevation of 1770 metres, between Dennis and 241 Creeks just to the northeast of the upper drainage divide of Penticton Creek. As an example, the site Penticton-2 is shown in Figure 2.4. These sites are located roughly on a transect that runs through the study area, however none of these gauges was actually located within the study watersheds. Air temperature was measured at all three sites with Lambrecht chart recording thermographs housed in Stevenson screens. These sites were established in 1983 along with the three main weirs and monitored by Water Survey of Canada. Beginning in the winter of 1988-89, air temperature was also monitored under the forest canopy at 240 and Dennis Creeks, also with Lambrecht thermographs housed in Stevenson screens. In the summer of 1988, four soil temperature measurement sites were established; one each under the forest canopy at 240 and Dennis Creeks, and two in the regenerating clear-cut at Edelweiss Creek, located on north and south aspect slopes. Temperature was measured at depths of 5, 20 and 50 cm below the mineral soil surface using "Soiltest" thermistors that were measured with a digital ohm meter. Equations relating temperature to resistance were established by burying the thermistors in a bucket of soil and calibrating them in a cold room with a freezer section. This was done before installation and again after the probes were removed. 2.1.3: Groundwater. Groundwater was monitored using a combination of water table wells and piezometers. Water table wells were used to observe water table heights, whereas piezometers were used to measure pressure head at the completion interval, and to measure hydraulic conductivity. Both types of instruments were used to collect samples of groundwater for chemical analysis. There were 14 piezometers and 5 water table wells altogether. Groundwater measurement sites were located on hillslope transects near the weirs in each of the four watersheds. Most instruments were installed in the summer of 1988, although the transect in Dennis Creek was not installed until the summer of 1989. A schematic of a groundwater installation is shown in Figure 2.5, consisting of lower and upper sites with a water table well and two or more nested piezometers,. 45 although some sites either lacked the water table well or had only one piezometer. Note that in the example shown, groundwater is recharging at the upper site, and discharging at the lower site. The first standpipes were installed at the upper site at 240 Creek. These were drilled with a truck mounted auger rig, and all others were drilled with a hand auger. Three different borehole sizes were used depending on the auger that was available; 4, 5 and 6 inch diameter. In each v case, the first borehole was drilled as deeply as possible and completed with a piezometer. Subsequent holes were drilled to a lesser depth if the intention was to install a piezometer, or as deeply as possible if a water table well was to be installed. At each site, boreholes were drilled about one metre apart. All standpipes were constructed of 1" i.d. PVC irrigation pipe. Water table wells were prepared by cutting holes at 5 cm intervals along the length of the pipe equal to the borehole depth, wrapped in geotextile and inserted into the borehole. The hole was then backfilled with sand to near the ground surface and sealed with cement at the top. Piezometers ] were slotted at 3 cm intervals on alternate sides of the pipe for a 15 cm completion interval and the completion interval was wrapped in geotextile. The standpipe was then inserted into the borehole and backfilled with 20 cm of sand, sealed with a layer of bentonite, and then backfilled with cement to the surface. A schematic of a completed piezometer is given in Figure 2.6. Hydraulic conductivity was measured using standard piezometer tests as described by Hvorslev (1951). In shallow piezometers slug tests were used and in deeper ones, bail tests. With the bail tests, water was removed from the piezometer with a hand pump. Recovery times of piezometers ranged from minutes (Figure 2.7) to several hours (Figure 2.8). Formulae for calculating hydraulic conductivity from piezometer dimensions and test data are as follows: K = 0.00083/To for 4" boreholes K = 0.00076/T0 for 5" boreholes K = 0.00065/To for 6" boreholes where To = the time at which (H - h(t))/(H - Ho) = 0.37 as shown on the bail test graphs (Figures 2.7 and 2.8). Figure 2.6: Schematic of a Piezometer 47 o o o o o o o o o o o o o o h - <£> l O •«*• CO CM T" (sajnutw) eiuji ( s e j n i n u j ) e u t j i 48 2.2: DATA SYNTHESIS 2.2.1: Precipitation. By far the most important input data in any hydrologic modeling procedure is the precipitation data. The modeling procedures used in this study require accurate daily precipitation data to yield meaningful results. Two problems were identified with regards to the historic precipitation data gathered in the Upper Penticton Creek area: some of the daily precipitation values were missing, and there was the likelihood of undercatch of snowfall due to poor siting of the gauges. Missing data occurred because periodically the clock on one or more of the gauges would stop, yielding only total precipitation between servicing intervals rather than daily values. When this occurred, the missing daily values were estimated for the gauge in question by multiplying the total precipitation for the missing interval by the daily values at another gauge expressed as proportions of the total precipitation at that gauge over the same interval. In most cases, daily values were recorded for at least one of the three gauges. On occasions when all three gauge clocks stopped, daily values at McCulloch were used to fill in the gaps. Thus, all the gaps have been filled with estimated daily precipitation. Of the three Belfort (Universal) precipitation gauges that were operating in the Upper Penticton Creek area during the period selected for these simulations, none were sited according to the criteria laid out by Brakensiek et al. (1979) nor were they equipped with alter shields. Instead, each gauge was sited in a clear-cut. Therefore it is reasonable to expect that the gauges might undercatch during snowfall according to wind speed (Goodison, 1978). Unfortunately, wind speed was not measured at any site in or near the study area prior to August 1991, so that Goodison's correction factors for universal gauges could not be applied to the snowfall data. Thus an alternate method was used to correct the daily precipitation record. Near the gauge at Cheng there is a snow course designated "Grayback Reservoir". The snow course is located east of the road near the southwest corner of the lowest clear-cut above where Penticton Creek crosses the road at an elevation of 1550 m. It is sited to best represent 49 snow accumulation on the ground. The gauge at Cheng is at an elevation of about 1540 m immediately southwest of the snow course on level ground. It is therefore assumed that the actual cumulative snowfall at the two sites should be identical, and therefore the difference between snow water equivalent at the snow course and cumulative snowfall at the gauge should be due entirely to sublimation loss. Therefore, cumulative snowcatch should always exceed SWE of the snow course. The Ministry of Environment operates a snow pillow and snow gauge sited next to each other at Mission Creek north of UPC on the Okanagan Highlands. The site is located in a sheltered area where wind does not interfere with the accurate operation of the gauge, at an elevation of 1770 m. Monthly precipitation totals for the gauge at Mission Creek are well correlated with those at Cheng for the period Sep 87-Aug 90 with an R2 of 78.4%. The cumulative precipitation and SWE of the snow pillow at Mission Creek along with the cumulative precipitation and snow course measurements at Cheng-Gray back Reservoir were graphed for the 1987-88,1988-89 and 1989-90 snow accumulation periods. An example is shown in Figure 2.9. Figure 2.9: Cumulative Gauge Catch and SWE at Cheng and Mission, 1988-89 1988-89 50 At Mission Creek, the cumulative precipitation is consistently greater than the SWE of the snow pillow by 5-20 percent. These differences agree with those given by Storr & Golding (1974) that were due to sublimation loss. The differences between cumulative precipitation and SWE at Cheng-Grayback Reservoir are similar to those at Mission Creek for 1987-88, but not for 1988-89 or 1989-90. Therefore, in the absence of wind speed measurements, it is reasonable to assume that the gauge at Cheng has undercaught snowfall for at least part of the latter two seasons. Daily precipitation values fors the gauge at Cheng were increased for the periods between 7 Nov 88-4 Apr 89 and 3 Jan 90-24 Feb 90 such that the percentage differences between cumulative precipitation and SWE of the snow course measurements roughly match the mean monthly percentage differences calculated for Mission Creek at the same date when the snow course measurements were taken at Grayback. Having done this, correction factors were calculated by dividing the adjusted daily values at Cheng by the original unadjusted values. Adjusted daily precipitation values were between 0 and 30% higher than unadjusted values. Assuming that wind effects at Dennis and Pent-2 are similar to those at Cheng, the same adjustment factors were applied to daily precipitation measurements at those sites. It is important to note that for purposes of this study, accuracy of the total snowfall at the gauge is more important than precision of the daily precipitation record during the snow accumulation period. The gauge at Cheng appears to be the most reliable of the three gauges because of the nearby snow course to verify the measured catch. However, at an elevation of 1540 metres, it is lower than the lowest point in the study area (the weir on 241 Creek at an elevation of about 1615 m). The gauges at Dennis and Penticton-2 are both at an elevation of about 1770 metres. Quarterly precipitation totals based on adjusted data for the period 1987-1990 show that precipitation catch at Dennis is consistently lower than at Cheng for any three month period despite the higher elevation, whereas the same totals at Pent-2 are higher than at Cheng, as expected. The most probable explanation for this is that the exposure of the Dermis site on a clear-cut ridge facing the prevailing wind causes severe undercatch due to wind even under 51 rainfall conditions. Since this is contrary to the above mentioned assumption regarding the similarity of wind effects, it follows that the precipitation from the Dennis site cannot be used. The Pent-2 site, on the other hand, is probably more sheltered from wind because it is on the lee side of the drainage divide. For purposes of modeling, any gauge can be used as long as the measurements are accurate. The UBC model includes assumed precipitation-elevation relationships to forecast precipitation at elevations above or below the gauge in a watershed. A site specific relationship might make the modeling procedure more reliable, and will also help to identify the accuracy of the catch at each gauge. To do this, the daily precipitation record at the Summerland Agricultural Research Station at an elevation of 395 metres was obtained. This site is on the east side of Okanagan Lake to the west-southwest of UPC such that all four gauges lie along a straight line. Mean annual precipitation for 1988-89 was calculated for each site and plotted against elevation (Figure 2.10). The precipitation-elevation relationship calculated using Summerland, Cheng and Pent-2 is PPT(mm) = 0.29(elev(m)) + 224.6 where PPT is the mean annual precipitation in mm. The equation has an R2 of 99.9%. Clearly, the precipitation at Dennis deviates significantly from this relationship. At this time, it seems more likely that this deviation is due to gauge undercatch than to any real effect of topography. The effects of topography on precipitation distribution may be the subject of future research in Upper Penticton Creek; however, the gauge measurements from Dermis Creek will not be used at this time because they seem to be unreliable. 2.2.2: Temperature. Temperature data have been collected at the sites Cheng, Dennis and Penticton-2 since 1983 by recording thermograph. These records suffer from periodic losses of data due to the clocks running down or running out of ink, but during the data collection period at least one was running at all times. Most of the time, all three thermographs were operating. The period 1985-1990 was used to develop regression equations to fill in missing data; Single and multiple 0 400 800 1200 1600 2000 Elevation (metres above sea level) regressions were performed on maximum and minimum temperatures at each site against the other two. The regression equations were used to fill in missing data for that period, although only the period September 1987 to August 1990 was used for modeling. When missing data occurred at only one of the sites, the data were estimated from both the other two sites. When only one site was operational, missing data at the other two sites were estimated from the one site. The average maximum and minimum temperatures for UPC were calculated as the average of maximum and minimum temperatures at Cheng, Pent-2 and Dennis. These composite temperature data were used to describe the climate of the study area (see Figure 1.3), and also to fill in temperature data at the other sites, 240 Creek and Dennis-f (Dennis Creek under the spruce-fir forest canopy). At these sites, temperatures were measured under the forest canopy for comparison with temperatures measured at the open sites. These temperatures were considered more representative of conditions in 240 and 242 Creek watersheds for the purpose of modeling snow accumulation and melt. Because the thermograph at 240 Creek was installed in January 1988 collecting weekly data until October of that year, when it was converted to a 30 day clock, 53 ( and the thermograph at Dennis-f was not installed until May 1989, much data estimation was required for the early part of the modeling period. Maximum and minimum temperatures were intercorrelated between the five sites and UPC to determine the best data sets to fill in missing data at 240 Creek and Dennis-f. For 240 Creek, the maximum temperature was most highly correlated with that at UPC. For minimum temperature at 240 Creek, the highest correlation was with Dennis-f, but since this site could not be used to fill in data, the site with the next highest correlation, Cheng, was used. This led to the following regression equations: 240max = l.Ol(UPCmax) - 2.02 R2 = 97.0% 240min = 0.96(CHENGmin) - 0.65 R2 = 84.5% In the case of Dennis-f, the maximum temperature relationship to other sites was more complex, since there was a time lag associated with warming under the spruce-fir canopy, and also because temperatures under the canopy tended to be warmer in winter and cooler in summer compared to the lodgepole pine canopy. This led to the use of a quantity T-date (temperature correction date) equal to 1 on February 01 and 183 on July 31 with linear interpolation in between those dates. Using this quantity it was found that the maximum temperature at 242 Creek was most highly correlated with the UPC average maximum and the minimum temperature was most highly correlated with that at 240 Creek, leading to the following equations to fill in missing data under the spruce-fir canopy: 242max = - 3.69 + 0.783(UPCmax) + 9.0X10-7(T-date)3 R2=96.7% 242min = - 3.03 + 0.821(240min) + 0.0571(T-date) - 7.48X10-4(T-date)2 + 29.7X10-7(T-date)3 R2=94.3% The above regression equations were used to fill in maximum and minimum temperature data for the sites 240 Creek and 242 Creek (both under the forest canopy) for the period Sep 1987-Aug 1990. 54 2.2.3: Other Data. Many snowmelt-runoff models such as HSPF rely on solar radiation and wind data, and some require net radiation measurements. These data were not collected at Upper Penticton Creek prior to 1991. Other model requirements include evaporation pan measurements to derive potential evapotranspiration data. These requirements make such models impractical in their current form for use in British Columbia since those types of data are rare. Initially, attempts were made to estimate solar radiation, potential evapotranspiration and wind data. Radiation data for UPC was estimated using radiation data from Summerland and relationships between temperature and temperature range of UPC vs Summerland, and potential evapotranspiration was calculated using a temperature based formula. Wind was extremely difficult to estimate due to a total lack of year round wind data at nearby sites, and due to the fact that wind is extremely site specific. This contributed to the decision to use the UBC Watershed Modeling system for hydrologic simulation, developed by the Mountain Hydrology Group, Department of Civil Engineering, University of British Columbia. The UBC model has been described in a previous section, and is described in detail by Quick et al. (1995). 2.3: SAMPLING AND DATA COLLECTION Four types of samples were collected for chemical analysis; streamflow, groundwater, precipitation as snowfall or rainfall and stratified snow pack. Streamflow samples were collected using a DH48 integrated depth sampler just immediately upstream of the weir ponds at each creek and transferred to sample containers. Groundwater samples were collected from wells and piezometers using a bailer that was lowered down the standpipe and then transferred to a sample bottle. The bailer was made out of a section of copper pipe with an end cap and holes drilled in the pipe half way along its length. At the beginning of each field trip the piezometers were purged with a hand pump after measuring the water level in the pipe and prior to sampling. Subsequent samples collected in the same held trip were obtained without purging, but the first bail was discarded to ensure that the water collected was representative of that in the soil. Water level in the standpipe was measured 55 prior to the collection of each sample using a carpenter's tape measure that was lowered down the pipe; depth to water was determined by listening for the splash as the end of the tape hit the water. Precision of the measurement thus obtained was +0.5 cm. Interflow, groundwater seepage and overland flow samples were collected directly at the point of origin in a sample bottle. The interflow and seepage samples were collected periodically from specific outflow points. The outflow sites that were sampled were classified as interflow if outflows occurred only around high flow periods, and as groundwater seepage if the outflow was of a more permanent nature. The outflow rate of interflow was judged as fast, medium or slow at the time of sampling. Snowfall and snowpack samples were collected in wide mouthed jars and allowed to melt. Both groundwater and snow samples were filtered in the field and then transferred to a clean sample bottle. Rainfall samples were collected in Atmospheric Environment Service rain gauges that were installed at the groundwater monitoring sites at ,240, Dennis and Edelweiss Creeks and transferred to sample bottles on each field trip. All samples were stored in 150 ml Nalgene containers such that only a small air bubble was trapped inside the container, tightly sealed and packed in ice in a cooler for the duration of each field trip. Some analyses were performed on the samples immediately upon return to the lab; bicarbonate and pH analysis were always performed immediately, and starring in the spring of 1990, nitrate and ammonium analyses were also performed immediately using methods described later, after which the samples were frozen to await the remaining analyses. Snowpack samples were collected at snow pits in 240, Edelweiss and Dennis Creeks. Snow pits were dug so that the exposed face of the pit faced north to avoid melting of the profile during sampling. The depth of the profile was measured and up to five (usually four) samples were collected from the pit wall such that a profile of evenly spaced measurements was obtained from just below the snow surface to the base of the snow pack. The samples were collected using a soup can with a sharpened rim which was weighed after collection to determine the snow density, and then transferred to a wide mouthed jar. At the same time that the samples were 56 collected for chemical analysis, liquid water content of the snow pack was measured using a dilution method described by Perla et al. (1985). A weak (~0.01N) stock solution of HQ at 0°C is mixed thoroughly into a snow sample and about 100-200 ml of the solution is withdrawn from the mixture. The electrical conductivity of the solution relative to that of the stock solution is a measure of the liquid water content of the snow according to the following: ' where W/M is the liquid water content expressed as a percentage of the sample mass, S is the mass of stock solution, M is the mass of the snow sample, Gx and Gs are the specific conductances of the solution withdrawn from the snow mixture and the stock solution respectively, measured at room temperature. If a stock solution of ~0.01N is used, q equals 1.0133 and k (a factor to account for background conductance of the liquid water in the snow) is less than 0.01. A large 2 litre thermos was used to mix the snow and HQ solution in the field. The stock solution is placed in 500 ml wide mouthed jars, weighed and the solutions and the mixing thermos are buried in the snow pack for long enough to ensure that they reach a temperature of 0°C (i.e., over night if the snow pack is isothermal). Snow samples are removed from the pit wall with a scoop and put in the thermos with a stock solution and shaken for about 1 minute, then the thermos containing the mixture is weighed. About 100 ml of the solution is then drained off using the spout at the bottom of the thermos into the original solution container. The thermos is then dried with towels and the procedure repeated. Perla et al. (1985) state that the precision of the method is about 2%. Field work commenced in the summer of 1987 and regular field trips were conducted through until the summer of 1991. Streamflow sampling began in June 1987, with precipitation sampling commencing that winter by sampling snowfall. Rainfall and groundwater sampling began in May-June 1988, with the groundwater sampling program expanding as more wells and w s M M 57 piezometers were constructed. Field trips lasted anywhere from one to five days depending on the purpose of the trip. The aim of the field sampling was to cover as broad a range of hydrological conditions as possible in all seasons. Sampling was done every one to two weeks in the spring, summer and fall and once or twice per month in the winter. Intensive field sampling was conducted in the spring and early summer of 1990 consisting of four to five day field trips every two weeks whose purpose was to study diurnal responses to snowmelt Two field trips were conducted in each of 1992 and 1994 to collect supplementary data. 2.4: CHEMICAL ANALYSIS Chemical analyses were performed on all samples to determine pH and concentrations of the anions bicarbonate (HC03), sulphate (S04), chloride (Cl), nitrate (NO3) and orthophosphate (HP04, H2PO4), the cations calcium (Ca), sodium (Na), potassium (K), magnesium (Mg) and ammonium (NfLj), and dissolved silica (SiC>2). The methods used were either standard methods as described in "Standard Methods" or were modified to suit the range of concentration found in the samples as described below. Bicarbonate was measured as alkalinity by potentiometric titration. Alkalinity is normally interpreted as bicarbonate (HCO3") or carbonate (CO3") depending on the pH of the sample. Carbonate only exists at pH above about 9. Since the highest pH found in these samples is about 7.5, all the alkalinity can be interpreted as bicarbonate. Initially, a full potentiometric titration was performed on the first samples of streamflow, groundwater and precipitation collected. This procedure involves titrating small increments (1-5 ml) of dilute (in this case, 0.001N) hydrochloric acid into a measured volume of sample. At the end of each increment, the pH is measured. The titration proceeds in this way until a pH of about 4 is reached. A plot of pH/dV vs V is prepared, where dV is the incremental volume of acid added after each step and V is the cumulative volume. The peak on this graph corresponds to the point at which all alkalinity has been absorbed by the acid. After several such trials, it was discovered that the pH endpoint of the titration was 5.0+.0.1 for streamflow and groundwater, and 5.6±0.1 for precipitation. Thereafter, alkalinity titrations were done to the appropriate endpoint, with 58 periodic checks using the full method. The equivalent alkalinity as bicarbonate is then calculated as equal to the equivalent concentration of H + ion addedas follows: [HCO3-] = 61.02 (Va/Vs) mg/1 . where Va is the volume of 0.001N acid added, Vs is the volume of the sample and 61.02 is the equivalent mass of the bicarbonate ion (molar mass/valence). Nitrate, ammonium, chloride, orthophosphate, sulphate and silica were measured by Technicon Autoanalyser II in the Forest Hydrology lab in the Faculty of Forestry at UBC using standard colorimetric methods. Those methods are described by Technicon Industrial Systems (1971). Silica was measured in the 0-10 mg/1 range by reduction of silicomolybdate in acidic solution to "heteropoly blue" by ascorbic acid. The method for ammonium used the Berthelot reaction in the range of 0-0.5 mg/1. The method used for nitrate was based on the standard method that uses hydrazine sulphate to reduce the nitrate to nitrite for the analysis for a range of 0.05-5 mg/1. The method was modified to reduce the detection limit to 0.02 mg/1 by using hydrazine hydrate in place of the hydrazine sulphate. Chloride was measured with a method that depends on the liberation of thiocyanate ion from mercuric thiocyanafe by the formation of soluble mercuric chloride, forming a coloured dye proportional to chloride concentration in the range 0-10 mg/1. These methods so described were performed initially on the samples collected from June 1987 to 4 August 1988. Initially, a great deal of difficulty was experienced at trying to find a working method to measure sulphate. For the samples collected between June and November 1987, selected samples were taken to the B.C. Environment Lab (now Zenon Environmental Inc.) for sulphate analysis. A turbidimetric method was attempted on individual samples but with little success. For the period April to 4 August 1988, selected samples were analyzed in another lab in the Faculty of Forestry using a Technicon TRAACS 800 method. For samples collected from 23 August 1988 and after, some changes were made to the above analyses. Orthophosphate ion was added to the analysis. There are several standard methods available for the Technicon Autoanalyzer to analyze for phosphorus; the method used was a general method for total phosphorus or inorganic orthophosphate as P in the range 0.001-59 1.0 mg/1 based on the reaction of phosphorus in solution with ammonium molybdate and potassium antimonyl tartrate in acid medium to form an antimohy-phospho-molybdate complex. This complex is reduced to a blue coloured dye by ascorbic acid with colour proportional to the phosphorus concentration. The results were converted to concentrations of orthophosphate ion by multiplying by 3.11. A working method for measuring sulphate was obtained from Zenon Environmental Inc. that was modified from the standard methylthymol blue method to give a range of 2-50 mg/1. The modifications essentially amounted to the addition of larger mixing coils to allow full colour to develop. This method was further modified by altering the dilution rate to give a range of 0.2-5.0 mg/1 to match the range of concentrations found in the UPC samples. Difficulties that were experienced previously in maintaining a stable baseline in the chloride analysis at the low end of the method's range were overcome by diluting the colour reagent (made from acidic mercuric thiocyanate and ferric nitrate solutions). By experiment, it was found that if the working reagent was made to 1/5 the recommended strength and the sampling rate was changed from 60 to 40 per hour to develop full colour, the best balance between baseline stability and low end sensitivity was achieved for the UPC samples, most of which were in the range of 0.3-0.5 mg/1 chloride. Furthermore, interference due to colour was significant in this range, often amounting to as much as 50% of the apparent chloride concentration. This was overcome by rerunning the samples with a blank reagent in place of the colour reagent. This blank reagent was made of nitric acid diluted to the same concentration that was present in the colour reagent to duplicate the same pH as the colour reagent This gave a chart deflection from the samples that was due only to colour. The true chloride concentration was obtained by subtracting this value from the apparent chloride concentration. The modifications made to the sulphate and chloride methods by altering the sample dilution and dilution of the colour reagent were effective at lowering the detection limit, thereby improving low end sensitivity. Using the same reasoning, it was found that the detection limit of the nitrate method could also be improved. The colour reagent was made at half strength, to give a detection limit of 0.01 mg/1. The modifications so described were used on all samples collected 60 on and after 23 August 1988. The ammonium and silica methods were not altered because the ranges and detection limits of those methods were appropriate to the UPC samples. The cations Ca + + , Mg + + , K + and Na+ were analyzed using atomic absorption spectrophotometry. The instrument used was a Varian Techtron AA6 unit in the Forest Fire Science lab in the Faculty of Forestry. The most precise measurements are achieved by calibrating the instrument to the narrowest range possible. Since the lowest concentrations of cations are found in precipitation and the highest in groundwater, the samples were divided up into lots where the range of cation concentrations were expected to be similar, and the instrument recalibrated for the optimum range for each lot. The instrument required frequent calibration checks, usually after every 10 samples. The calibration of the instrument is easiest to maintain in the middle of its range. At the low end, the calibration tends to be unstable, and at the high end there is a lot of variability in the output when measuring a sample or using a standard to calibrate the instrument. Because this particular instrument is not automated but is operated manually, the operator must mentally integrate the digital readout Therefore, the reported values are actually the estimated mean of a probability distribution. The variance of that distribution is largest at the upper end of the calibration scale for a given cation, and is generally the largest for calcium. Magnesium was the easiest analysis to perform. The calibration was the most stable for this cation, the variance lowest and for the most part, the sample concentrations were right in the middle of the calibration range (except for precipitation, which was at the low end for all four cations). The most difficult was potassium, since the concentrations, even of groundwater, were at the low end of the calibration scale, and the calibration at that range was non-linear. This meant that a correction factor had to be applied to the readout which was different for each calibration of the instrument The measurement accuracy of each cation is estimated at +0.03, +0.003, +0.03 and +0.02 mg/1 for calcium, magnesium, potassium and sodium respectively. These accuracies were estimated from observing the variance of the digital readout. 61 2.5: DATA ANALYSIS 2.5.1: Statistical Analysis. Various methods were used to analyze the physical and chemical data that were collected. Statistical methods included time series, correlation, regression and discriminant analyses. All statistical analyses were done with Minitab for Windows Release 9 except where noted. The time series analysis consisted of cross correlation to determine lag times between flows on the creeks. Regression analysis was widely used and included stepwise regression, simple and multiple regression, linear, polynomial, log-normal and linearized exponential curve fitting, and multiple linear and curvilinear analysis of covariance consisting of regression with dummy variables and regression comparison. In all cases, the same procedure was used to determine that the best fit had been obtained. The procedure is outlined as follows: 1. The dependent variable was plotted against independent variables to see if there appeared to be a relationship, and the form that the relationship would likely take (e.g., linear vs log-normal). 2. Once the form of the relationship was determined, a set of predictor variables was assembled for entry into the regression equation. For example, if the model looked like a second order multiple regression, then the set of predictor variables included first and second order terms.of each independent variable and interaction terms (the product of two independent variables). If the problem appeared to be one in which there was covariance between two variables but the levels of those variables were site dependent (e.g., chemical concentration vs hydraulic head in groundwater), then a set of dummy variables was used to represent the differences between the sites in the attributes that could be responsible for the different levels. 3. Having completed step 2, stepwise regression was used to select the best set of predictors. The selected parameters were then entered into a full regression analysis to determine if the selection formed a good model. This was done by plotting the resultant equation and examination of the residuals (see step 4). However in practice, stepwise regression often failed to produce a reasonable set of predictors particularly when dummy variables were used, although when it did, it shortened the procedure considerably. If the stepwise regression failed then the full set of 62 predictors was entered into the regression analysis, and their significance in the regression was assessed using their associated t-scores and p values. Insignificant predictors were then removed from the equation one at a time according to the least significant until all remaining predictors had significant t-scores. 4. Since dummy variables use a binary code to represent a given attribute, they were often used in pairs if more than two site characteristics were being represented. In this case, their individual t- scores could not be used to assess their significance since both dummy variables were required to represent one attribute. Significance of paired dummy variables was assessed according to the following: partial F - " S S R ( r e d u c e d > ] / r F MSE(fuU) r = # regression coefficients(full) - # regression coefficients(reduced) d.f. = r, di.e.(full) where the full model contains the specific dummies that are being tested, and the partial model has those dummies removed but contains all other predictors that are used in the full model. 5. Having identified the model using the above steps, the residuals were examined by plotting them against the fitted values. The plots were then examined for patterns as described by Draper & Smith (1981). A random residual plot indicates that the model is the correct one. If the residuals form a curved pattern a higher order model or some transformation is indicated, if there is a trend it means that another independent factor should have been considered, or if the pattern fans out then the variance is not independent of the level of one or more independent variables, indicating that weighted least squares analysis should have been used. In all regression analyses, the above steps were followed to determine the appropriateness of the model. Only the results of the best regression equation are reported along with the standard error, the R^ and the p value associated with the overall F statistic, or if the regression was not significant it was reported as such. In cases where regression is through the 63 origin because the constant was found to be insignificant, R^ is not given but rather, an statistic was calculated as follows: 2> - y ) 2 This statistic was normally reported as R^ because regression results were reported in tables. The assumption is that if the equation is given with no constant, the reported R^ statistic was actually calculated as A Other methods used were regression comparison and multiple discriminant analysis. Regression comparison was done according to Kozak's (1970) method using a program called REGCOMP version 1.0. Discriminant analysis was performed to determine correct classifications of soils and average groundwater chemistry. The significance of the differences between groups was assessed according to an F statistic given by Davis (1973): „ ria + n b - m-1 runb „ ? F = x D ( n a + r ib -2 )m r i a + rib with m and (na+nb~m-l)~ degrees of freedom, where m is the number of independent variables and D2 is the squared distance between groups. 2.5.2: Hydrologic Analysis. As discussed elsewhere, hydrologic analysis was performed using the UBC Watershed Model. This modeling was done to determine the composition of streamflow based on the data that were available. A priori knowledge about groundwater behaviour was used in the calibration to pre-determine approximate values for some of the parameters, and then the model was fine-tuned using the optimization technique described by Quick et al. (1995). The goodness of fit of the modeling is assessed according to the Nash Sutcliffe coefficient of efficiency e!. This is used in place of R2 because R2 is independent of volume and only a measure of how well the 64 calculated hydrograph shape matches that of the observed hydrograph. On the other hand, e! relates to both shape and volume of the calibration: ^(Qobs^Qes t j ) e! = l - ^ i = 1 X (Qob S i -Qobs ) 2 A value of 1 indicates a perfect fit Generally values greater than 0.85 are considered indicative of good fit; efficiencies can be negative. All water chemistry modeling was done on a spreadsheet (Quattro Pro for Windows V 5.0) using output from the UBC model that was imported into the spreadsheet and then parsed. The water chemistry was then modeled using the flow components with equations to predict the chemistry of those components. > 65 CHAPTER 3: RESULTS OF FIELD STUDIES As discussed in the methods section, a full range of inorganic anions and cations was measured in samples of precipitation, groundwater and streamflow. Only those chemicals that are significant from a geochemical standpoint were selected for physically based water chemistry modeling. Freeze & Cherry (1979) have identified bicarbonate, calcium, magnesium, chloride, silica, sodium, sulphate and carbonic acid as major constituents in groundwater. However, of these major chemicals, it was decided that only those whose average concentrations in streamflow in UPC creeks were greater than 1.0 mg/1 would be used. Thus, sulphate, bicarbonate, calcium, sodium and silica were considered in the following sections. In addition, nitrate was included in the precipitation results because it was found in significant concentrations in rain and snow samples, although nitrate concentrations were not modeled in groundwater or streamflow because it usually occurred at a random background level of concentration with occasional pulses occurring during early snowmelt. Nitrate and sulphate are inorganic anionic forms of nitrogen and sulphur which are essential plant nutrients with complex biogeochemical cycles. Calcium is a cation that is also an essential plant nutrient. Sodium and silica on the other hand are not significantly accumulated in plant biomass (Vitousek, 1977). Sodium is a cation, whereas silica is neutral. Sodium, calcium and silica are produced primarily by primary mineral weathering in the soil, whereas sulphate is primarily of atmospheric origin, and nitrate is derived from a combination of atmospheric sources (from nitic acid in precipitation or from nitrogen fixation by plants) and from mineralization of organic nitrogen or oxidation of ammonium by microbes. Bicarbonate, which is by far the most significant anion in groundwater and streamflow is produced primarily by the breakdown of organic matter in the soil by microbial action (Thorne et ai, 1988). These six chemicals provide an interesting basis to study water chemistry because they are geochemically and biologically significant, and anions, cations, and non-ionic species are represented. 66 3.1: PRECIPITATION 3.1.1: Types of Precipitation Samples Precipitation samples were divided into three main categories: rainfall, snow surface grab samples (snowfall), and snow pack. The latter samples were stratified and also involved measurements of snow density, grain size and liquid water content during snow melt as described in Chapter 2 "Methods". Those categories were further subdivided into groups according to the forest cover type under which the samples were collected, namely mature lodgepole pine, mature spruce-fir and immature regeneration. Rainfall and snowfall represent the chemistry of precipitation, whereas snowpack samples are used to determine the physical and chemical changes that occur in the snowpack over time and particularly during snowmelt. Initially, multiple discriminant analysis was used to determine if there were significant differences between rain and snow chemistry under different forest cover types. Groups 1, 2 and 3 represent rainfall while groups 4, 5 and 6 represent snowfall (Table 3.1.1). First, all measured chemical species were considered in the analysis, with the results of that analysis given in Table 3.1.1. Significant differences were found between rainfall under the lodgepole pine canopy, and snowfall in the clear-cut and under the pine forest These differences are given in bold type. The analysis was then repeated using as a subset of the analysis the chemicals NO3, SO4, Cl, Mg, K and S1O2 (Table 3.1.2). This produced significant differences between rainfall and snowfall, but showed that rainfall samples collected at all sites were similar, as were snowfall samples. The discriminant analysis indicated similarity within rainfall samples and within snowfall samples, and a significant difference between rainfall and snowfall chemistry. All rainfall samples and all snowfall samples were lumped together and a series of two sample T-tests was conducted to determine the differences and similarities between rain and snow. Average rain and snow chemistry is given in Table 3.1.3 along with the results of the T-tests. Average chloride, magnesium and potassium are included in that table to demonstrate the differences between rain and snow chemistry. There are significant differences between 67 Table 3.1.1: Discriminant Analysis among precipitation grab samples, all chemicals: Squared Distance Between Groups Group Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 1 0.0000 5.4186 7.2373 6.3835 4.1384 7.4138 2 5.4186 0.0000 1.6077 9.8372 8.0966 16.7366 3 7.2373 1.6077 0.0000 8.9777 8.6848 17.0680 4 6.3835 9.8372 8.9777 0.0000 2.0217 3.6268 5 4.1384 8.0966 8.6848 2.0217 0.0000 2.1805 6 7.4138 16.7366 17.0680 3.6268 2.1805 1 0.0000 Groups: 1-3 = rainfall under spruce-fir, lodgepole pine and immature regen respectively 4-5 = snowfall under spruce-fir, lodgepole pine and immature regen respectively Table 3.1.2: Discriminant Analysis using chemicals NO3, SO4, Cl, Mg, K and S1O2: Squared Distance Between Groups Group Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 1 0.0000 3.5230 6.6004 4.8686 2.8480 4.8358 2 3.5230 0.0000 1.0437 6.1938 4.4952 10.7020 3 6.6004 1.0437 0.0000 9.1174 8.1273 15.8617 4 4.8686 6.1938 9.1174 0.0000 1.6523 2.6390 5 2.8480 4.4952 8.1273 1.6523 0.0000 1.7099 6 4.8358 10.7020 15.8617 2.6390 1.7099 0.0000 Table 3.1.3: Average Precipitation Chemistry in mg/1. Results of Two Sample T-tests, Total Chemical Inputs due to Rain and Snow TYPE NO3 Cl SO4 HCO3 Ca Mg K Na S1O2 Rainfall 0.23 0.56 0.93 0.46 0.37 0.05 0.46 0.32 0.19 Snowfall 0.46 0.39 0.39 0.40 0.27 0.02 0.11 0.21 0.68 Snow Pack 0.40 0.59 0.34 0.14 0.13 0.01 0.30 0.24 0.10 T rain vs snowfall -4.12 1.45 3.34 0.35 1.73 3.26 4.28 1.10 -3.69 p rain vs snowfall 0.0001 0.15 0.0017 0.73 0.089 0.0019 0.0001 0.28 0.0006 Rain kg/ha/yr 0.7 i:8 3.0 1.5 1.2 0.2 1.5 1.0 0.6 Snow kg/ha/yr 1.9 1.6 1.7 1.7 1.1 0.1 0.4 0.9 2.9 Total input kg/ha/yr 2.6 3.4 4.6 3.2 2.3 0.3 1.9 1.9 3.5 concentrations of nitrate, sulphate, magnesium, potassium and silica in rain and snow at the 95% confidence level, and between chloride and calcium at lower confidence levels. This suggests that rain and snow have different origins. The fact that rain is higher in chloride and basic cations suggests the presence of salt crystals as condensation nuclei, whereas the higher levels of silica in snow indicates the presence of dust as freezing nuclei. Rain is higher in sulphate whereas snow is higher in nitrate. This also suggests different origins of rain and snow, although wind data are not available to confirm this. In the forested sites, common witches hair (Alectoria 68 sarmentosa), a nitrogen fixing arborealf lichen, is incorporated into the snow pack. This may help to explain the higher levels of nitrate at those sites and also the fact that higher nitrate levels persist during snowmelt (as shown by the chemistry data for snow packs), but those lichens are not found in snow packs in the regenerating clear-cut where nitrate levels are just as high as under mature forest canopies. This would support the argument that the differences in snow and rainfall are due to different origins. Rainfall appears to originate directly from the ocean, in which case the elevated sulphate levels could be from the air mass passing over the Lower Mainland. The snow probably originates from more continental sources. The average snowpack chemistry is given as a comparison to snow surface samples. Slightly lower levels of sulphate and nitrate occur in snow pack samples indicating gradual leaching of those chemicals over time. Higher average concentrations of chloride, potassium, and sodium are due to enrichment at the base of the snowpack under certain conditions, as discussed later. 3.1.2: Annual Inputs of Chemicals due to Precipitation. Table 3.1.3 also contains a calculation of annual average chemical inputs to the watersheds from a combination of rain and snow. This was based on precipitation and temperature records at the meteorological site Penticton 2 whose elevation roughly matches the H50 mark of the UPC watersheds. It was assumed that if the mean daily temperature was 2 °C or less, precipitation would fall as snow and the average chemistry of snowfall would apply. Conversely, if the mean daily temperature was greater than 2, rainfall would occur. Over the three year period from September 1987 to August 1990 the average annual precipitation was 737 mm. Using the above temperature criterion, 57% fell as snow and the remainder as rain. Thus, the total volumes of rainfall and snowfall per hectare were calculated, and the total chemical input due to rain and snow were derived as the product of the chemical concentrations and the volume of the input The above allocation of chemical inputs based on the presence of rain vs snow is valid only if the chemical differences between rain and snow are due to the phase of the precipitation. 69 The 2 °C temperature criterion is also somewhat arbitrary, since there is generally a transition from snow to rain from 0 to 5 °C. If 0 °C is used instead, only 43% falls as snow, whereas if 5 °C is used, 69% falls as snow. For example, the inputs of nitrate, silica and sulphate could vary from 2.43-2.87, 2.95-3.90 and 4.10-5.15 kg/ha/yr respectively. Thus the total inputs of nitrate, silica and sulphate given in Table 3.1.3 are within 8%, 14% and 11% respectively of the true value. On the other hand, if the difference between rain and snow chemistry is assumed due to the origin of the air mass which delivers the precipitation, which is governed by season rather than temperature, the precipitation could be assumed to have the characteristics of rain from May to September, and that of snow from October to April. This would result in a 50/50 split between rain and snow in terms of the chemical input, resulting in total inputs of nitrate, silica and sulphate of 2.54, 3.21 and 4.87 kg/ha/yr respectively, resulting in about half of the error that was calculated from varying the temperature. It is therefore estimated that the total chemical inputs based on the 2 °C temperature criterion are within 5% of their true value. 3.1.3: Physical Properties of the Snowpack. Physical properties of the snowpack were measured during snowmelt, including profiles of snow density, average grain size and liquid water content. Liquid water content of the snowpack was measured using a dilution method described in chapter 2. The purpose of these measurements was to determine the maximum water holding capacity of a ripe snowpack and to determine the variations in liquid water content with ambient temperature. Some snowmelt models (e.g. UBC (Quick et al, 1994) and HSPF, Johanson et al, 1984) require maximum water holding capacity of the snowpack to be specified as an input parameter. Observations showed that the snow pack could be divided into three distinct layers: the upper snow layer of about 10-15 cm in thickness, the middle layer that consisted of the bulk of the pack and the base of the snow pack, again being about 10-15 cm thick. At the beginning of the snowmelt period, the snow pack ripened to the point where its density (minus the liquid water) was between 300 and 450 kg/m3, and contained a quantity of liquid water which varied with the ambient temperature. The base of the snowpack contained enlarged grains known as depth hoar, which develops in 70 dry snow packs as a result of temperature gradient metamorphosis. This process involves vapour transfer to snow grains causing them to become enlarged and tending towards a spherical shape (Perla and Martinelli, 1976). When the snowpack becomes primed with liquid water in the early phases of snow melt, a different type of metamorphosis occurs. Wakahama (1968) observed that snow immersed in liquid water experienced rapid grain growth, with grain sizes tending towards 2 mm in diameter. This process occurs during snowmelt in UPC snowpacks, with grains enlarging over time and the base layer always having larger grains than the upper and middle layers. Average grain size profiles for early melt season (March to mid-April) and late melt season (mid-April to early June) at 240, Edelweiss and Dennis Creeks are given (Table 3.1.4). The surface layer of the snow pack experiences freezing at night when the temperature drops below freezing, but the middle and basal layers remain primed and at a temperature of 0°C. When this occurs the liquid content of the surface snow layer drops to zero, whereas the layers below the surface will continue to drain to some theoretical lower limit at which the liquid water is bonded too tightly to the snow grains to drain under gravity. Leaf (1966) observed a range of liquid water contents in ripe snow from a minimum of 3-5% by mass of bonded water to a maximum of 12% by mass in surface layers using freezing calorimetry with temperatures ranging up to 15°C. He also noted that snowpacks drained to their minimum water holding capacity within 8-10 hours after sunset. Perla et al. (1985) report typical liquid content of snowpacks in the range 5-15%. The results described below are somewhat different from those of Leaf. Liquid water content as high as 18% by mass was observed in the surface layer during clear weather melt at the clear-cut site with air temperature Table 3.1.4: Average grain size profiles for early and late melt season Mean grain diameter (mm) Level in 240 240 Edel Edel Dennis Dennis Snowpack early late early late early late surface layer 0.7 1.1 1.5 2.1 1.8 0.8 upper middle layer 0.7 2.2 1.2 2.2 1.8 1.8 lower middle layer 0.7 2.1 1.2 2.5 1.8 1.8 basal layer 1.7 3.9 1.8 4.1 1.8 4.7 71 of 13°C on April 16, 1990. The following morning at 9:00 am, the surface layer had 0% liquid water, but lower layers still had a liquid content of around 10%. Thus, the lower layers had not completely drained overnight. This suggests that while the surface layer's liquid water content responds directly to the current ambient temperature, the previous day's melt conditions and the current ambient temperature both influence liquid water content of the lower layers. Furthermore, the base layer was found to hold less liquid water than upper layers; this can be attributed to the larger mean grain size (hence lower total grain surface area) at the base of the snowpack (Table 3.1.4). For these reasons, ambient temperature at the time of sampling was used to predict liquid content of the surface layer, whereas for the middle and base layers, a temperature index was used that was equal to the average of the ambient temperature at the time of sampling and the previous day's maximum. Using these temperature factors as predictors, exponential curves were fitted to the data to predict liquid water content of each of the three snow pack layers. Theoretically, there should be a practical upper limit to the water holding capacity of the snow at which the melt rate, the resultant liquid content and percolation rate are in balance. This is because percolation through snow is analogous to percolation through unsaturated soil; the unsaturated hydraulic conductivity increases with increasing water Content. Thus, the natural growth model was selected for the surface layer, and the logistic model for the lower layers. The natural growth model passes through the origin to represent overnight freezing of the snow surface, whereas the logistic model has a non-zero intercept. Both models have a horizontal asymptote to represent the practical upper limit of liquid content as described above (Sit and PouluvCostello, 1994). The natural growth model is of the form: Y = a(t-e b X) 72 The model can be linearized as follows: ' r ^ - I ^ J h M The logistic model is of the form: where b = -bi 1 + e b - c X and can be linearized as follows: N y - 1 | = b f + b 2 X where b = by c = -b2 In both models the asymptote is at Y = a. The value of the asymptote had to be determined by trial. This involved fitting models with various values of a and selecting the best fit model using a graph of F vs the asymptote (Figure 3.1.1). For the natural growth model, the value of the F statistic of the linearized fit was plotted against a and the correct value of a was taken as the inflection point of the curve. For the logistic model the R2 of the linearized fit was plotted against a and the correct value of a was taken at the peak of the curve. The equations are then solved by simple linear regression with the natural growth model being specified through the origin. This has resulted in the following equations: For surface snow layers: W / M % - 2 5^ 1 - e ° 0 8 4 1 ) \2 = 57, 6% For middle snow layers: W / M % 16 1 + e 1.13-0.265Tav R2 = 29.3% For basal snow layers: W / M % 14 1 + e 1.17-0.211Tav R2 = 31.l% These equations are plotted in Figure 3.1.2. Note that the graphs indicate that the middle and basal layers will drain to a minimum water content of 4% and 3.2% respectively given sufficient (jsAei aoejjns) v.d aAjno pazueaun jo O I I S I J B J S J 74 time without additional melt, which agrees with Leafs results. The asymptotes (i.e., upper limit of liquid content) are 25%, 16% and 14% for surface, middle and basal layers respectively. 3.1.4: Profiles of Snowpack Chemistry. In the snowpacks, the most significant chemical species in terms of their concentrations were nitrate, sulphate, calcium and sodium. To illustrate trends in snowpack chemistry during melt, the 1990 data were selected. The four chemical species of interest were plotted as a series of profiles (that is, concentration was plotted on the X axis and height above the ground surface on the Y axis) for each of 240, Edelweiss and Dennis Creeks for the period between early February and mid June (Figures 3.1.3 to 3.1.10, 3.1.11 to 3.1.14 and 3.1.15 to 3.1.18 respectively). In those graphs, the profile that occurred on each day that measurements were taken is represented by one line. Thus the changes in snowpack chemistry over time can be inferred by comparing each successive profile. On each graph, the average concentration of the chemical in snowfall (i.e., the average value of the concentration in surface dry snow grab samples) is plotted on the graph for comparison. These graphs show that the chemical profile of snow under melt conditions is radically different from that of snow that is not actively melting. In the case of nitrate, sulphate and sodium the concentrations are higher than average in the surface snow layers and lower than average in the lower layers prior to the onset of snowmelt. Similar results were reported by Rascher et al (1987). This will be referred to as a positive profile because the chemical concentration increases with height above the ground surface. Rapid melt that began in mid April at 240 and Edelweiss Creeks caused a reversal of those chemical concentration profiles, presumable due to leaching of those chemicals down through the snowpack. Thus leaching achieved a negative profile in which the chemical concentration at depth was greater than average and at the surface layers, less than average. This is in keeping with other studies (Semkin & Jefferies, 1986, English et al. 1986, Hazlett et al, 1992, Rascher et al.,1987) in which chemical pulses occurred early in the melt period. This reversal did not occur at Dennis Creek, probably because rapid snow melt did not occur until much later. The graphs suggest that the reversal was just beginning on the 18th of June (see Figures 3.1.7 and 3.1.10), at which time the (0 ° Is co 1U Ul UI 5 5 2 o o o o o ro ro ro ro ro ro ro ro ro ro (ui) aoejans punojQ OAoqv IMBISH .4-1 o #5 _ ro 2 o 2 5 0} CD CD a> 0> O) fl) d> < < CL < i m m CD I C OT CB CO CO 3 E O C oo o s c a o c o o O 03 o CD o o o (LU) aoejjns punojQ 9Aoq v JuBiaH 00 CO CN o d d d d d (ui) aoepns punojQ O A o q v W B J B H i — i — i — i — i — i — i — i — r CO CO -3- CN o d d d d d (uu) eoejjns punojQ aAoqv W B J B H (w) aoejjns punojQ aAoqv W&\z>H 79 o CO Q. 9 o c .E Oi I* O 0) v. I— Q- O co .2 CO £ CO ^ = a> 4- CD O »-CO CO Z co . . TJ r- LU * : -M CO £ 3 .SP il 5 5 •c *c - 5 5- < ui "5 0> , CD < 2 2 tHH H a C oo o o "S a c o a ° i d o o oo a (tu) soejjns punaio SAoq v WBHSH 80 CM co o ^ d d d (LU) soBjjns punojQ aAoqv i q S i a H E (N co • » o " - o d d (LU) 336jjns punojQ S A o q v m6|3H L 81 CO (O -V CM O o o o o d (ui) aoepns punojQ aAoqv *u.BiaH OO IO CM O o d o i d d (ui) aoejjns punojg aAoqv mBian 82 0 n OL 1 o B O co cn = ? © © o o ° - «T E c 3 C I S co N CO CN CO ro OD 3 o. a a 2 < < < £ i l l 111 (LU) aoeuns puncug aAoqv JU.6j3H 0 co 0. 1 o CA © S a. E 3 TJ « O © © » E c © D CM CN — 8 d T - ; co c i © § ; s ° E E E g < < < 5 I T + t H f r C O I f c a 1 *-> c « a c o o U r to o o C M d o o (ui) B o e u n s punojo O A o q v JLIBJSH 83 snow liquid water content indicated rapid melt. In contrast, calcium levels in snowpacks are always depleted relative to the average concentration in snowfall at all three sites. Rascher et al. (1987) attributed the depletion of ions from lower snow layers prior to snowmelt to periodic partial melts during the winter. The most likely explanation for the positive trend is that nitrate, sulphate and sodium accrete onto the surface of the snowpack by dry deposition in the absence of melt. An alternate explanation is that those chemicals are slowly leached from the lower layers throughout the winter while chemical enrichment occurs at the surface due to sublimation, but this would require that periodic melt occur during the winter to cause the leaching. Maximum daily temperatures remained below freezing from the beginning of November 1989 until early February 1990 at 240 Creek, so the gradual leaching theory would not explain the strong positive trend with depth in the snowpack profiles observed on February 7th. Temperatures indicate that in late February, some small amount of melt probably occurred, enough to partially ripen the snowpack. Chemical profiles observed at 240 Creek on the 10th of " March showed the effect of leaching; sulphate and nitrate levels had increased in the lower snow layers and had decreased in the upper layer such that the profiles were close to neutral. Sodium appears to have been leached out of the snowpack at that time. In the period between 10 March and 03 April, daytime temperatures warmed up gradually but not enough to cause significant melt, while 14.4 mm of precipitation fell, presumably as snow. On 03 April, the chemical profiles had returned to their previous strong positive non-linear trend. This can be explained by the addition of new snow with chemical composition similar to the average in bulk precipitation, with concentrations subsequently enhanced either by dry deposition or by sublimation. Cundy et al. (1980) suggested that a base temperature of 8°C is required to induce rapid melt. At both 240 and Edelweiss Creeks the reversal in the profiles of nitrate, sulphate and sodium occurred on the 14th-15th of April. This was the time during the melt sequence at which the daily maximum temperature exceeded the base temperature. This suggests that chemicals are quickly leached to the bottom of the snow pack during rapid melt resulting in a strongly negative non-linear profile, but that the positive profile is restored again once the show pack has 84 drained or the temperature stays below the base temperature for several days. Laboratory experiments by Colbeck (1981) produced results that support these findings. Impurities were added to snow and a series of intense freeze-thaw cycles were simulated under controlled conditions. The leachate was collected and its conductivity monitored. It was found that the impurities were leached from the upper layers and concentrated in the lower layers over a period of 2-4 days of intense melt with overnight freezing. Subsequent melt-induced runoff would remove those impurities. Between April 27 and May 11 at 240 Creek, this process is repeated; during cold weather. in late April, snow melt stops and the positive chemical profile is re-established at 240 Creek. A precipitation event of 33.9 mm as snow recorded on April 27 must have had an influence on the profile. On May 03-05, a snowmelt event occurred with maximum temperatures ranging from 8.2°C (May 03 under the canopy at Dennis Creek) to 18.5°C (04 May in the clear-cut). By May 09, maximum temperatures had cooled to about 8°C. In snow profiles measured on 09 and 11 May, there is again evidence of leaching (i.e., negative profiles) in spite of snowfall of 10.5 mm water equivalent on May 06-07, but at a slower rate, presumably due to lower temperatures than those recorded in mid April. Relative rates of leaching were not calculated but inferred qualitatively by inspection of the snow chemistry profile graphs. Snow profiles were not measured at Edelweiss Creek on April 27, and on May 09 and 10 the profile was similar to that measured on April 17 except with lower snow depth. It should be noted that snow pits were usually dug around the same location; at 240 Creek on a 10% slope on westerly aspect above the weir and at Edelweiss on a gently sloping (about 12%) north aspect near the weir. On May 10 to 11 alternate sites were chosen where melt rates were slower (on east aspect near the weir at 240 Creek and at a site upstream of the weir at Edelweiss Creek among a stand of trees about 6 m high). Profiles from these sites were similar to the regular sites but with slightly higher chemical concentrations. This suggests that chemical leaching is slower in areas that are more sheltered and consequently experience slower melt 85 At Dennis Creek the situation was different. The positive chemical profiles (i.e., higher than average concentrations at the surface and lower than average concentrations in middle and lower layers) persisted throughout most of the measurement period, although only one profile was measured during the mid April melt sequence, on April 14. On this date the chemical profiles were close to neutral. Profiles were not measured again until April 27, so it is not known if the negative profile developed on April 15-17 as it did at 240 and Edelweiss Creeks. On April 27, the profiles had begun to revert to the positive trend. Heavy snowfall later that day combined with temperatures generally below the base temperature result in restoration of the strongly positive non-linear chemical profiles observed on May 10. After this time the situation was complicated by rain on snow. Between May 10 and 28, 110.9 mm of precipitation fell. The mean daily temperatures during this period hovered close to 0°C so precipitation was likely a mixture of rain and snow. A snowfall of 27.4 mm water equivalent on May 24-25, followed by a rainfall of 18.4 mm on May 27-28 resulted in the peak flow on Dennis Creek (as well as for 240, 241 and Edelweiss Creeks) for 1990 at 06:45 am on May 29. This rain on snow had little effect on the chemical profiles in snow packs on Dennis Creek, except to increase the concentration of sodium; the profiles remained strongly positive and non-linear. On June 18 rapid melt was underway at Dennis Creek, and chemical profiles were roughly neutral with the exception of sulphate. Negative chemical profiles were not observed at Dennis Creek as they were at other sites in 1990, but were observed in 1989. This could be because temperatures at Dennis Creek during snowmelt were generally warmer in 1989 than in 1990. Soils were frozen at all sites and didn't thaw until snow had melted. At Dennis Creek, soil freezing was patchy in 1990 (observed during snow course measurements, R. Winkler, pers. comm.) but soils remained frozen at the site where temperature was measured, which was adjacent to the snow pit site. Soil temperature was unaffected by rain on snow. Thus, due to the reduced infiltrability of frozen soil, it is reasonable to assume that the base of the snowpack is directly connected to the stream channel by direct runoff as overland flow or subsurface bypass flow. 86 3.1.5: Chemical Concentrations at the Base of the Snowpack Depend on Melt Rate. The above noted chemical behaviour of the snowpack during melt has the potential to affect stream water chemistry, since it is likely that the base of the snowpack is linked to streamflow via direct runoff. Daily snowmelt production was determined following calibration of the UBC watershed model as described elsewhere. Relationships between chemical concentration at the base of the snow pack and daily melt were then investigated. Initially, the investigation involved correlating chemical concentrations at the base of the snow pack at 240, Dennis and Edelweiss Creeks (upper and lower sections) with daily melt and the natural log of daily melt that occurred at each site. Those correlation coefficients are reported as r (Table 3.1.5). The most significant relationships were found between melt and nitrate, sulphate, bicarbonate and sodium. There was essentially no relationship between calcium and daily melt at 240 and Edelweiss Creeks, although a significant relationship did exist for Dennis Creek and for the upper (mature spruce-fir) section of Edelweiss Creek. Silica was not significant at the base of the snowpack at any site. The data were plotted against melt (Figures 3.1.19-22) and inspection of the graphs showed that while a linear fit appeared appropriate for bicarbonate and calcium, and also for sodium at Dennis and upper Edelweiss Creeks, the concentrations of the other chemical species seemed to demonstrate some form of exponential behaviour with respect to daily melt at 240 and lower Edelweiss Creeks. For sodium at 240 Creek a logistic function was chosen, and for nitrate and sulphate at 240 and Edelweiss Creeks, the Chapman-Richard's equation was used (Sit and Poulin-Costello, 1994). For sodium, the value of the horizontal asymptote was initially set at the average concentration in overland flow, and then varied until the best fit was obtained. For nitrate and sulphate, the horizontal asymptote was set initially at the maximum concentration measured at the base of the snow pack and then adjusted until a best fit was obtained. This seems more reasonable than simply using linear regression since there should be an upper limit to the chemical content of the base of the snow pack, and the graphs seem to indicate that this is the case. However, in the cases noted above the plotted data points did not demonstrate a non-87 Table 3.1.5: Correlations (r) of Base Chemical Concentrations vs Daily Melt (mm) Chemical DAILY MELT ln(MELT) (DAILY) DAILY MELT ln(MELT) (DAILY) DAILY MELT ln(MELT) (DAILY) Creek 240 240 Dennis Dennis Edelweiss Edelweiss NO3 0.615 0.559 0.584 0.419 0.687 0,737 Cl 0.616 0.461 -0.091 . -0.498 0.797 0.756 SO4 0.528 0.377 0.706 0.653 0.853 0.888 HCO3 0.779 0.502 NA • NA - 0.476 0.319 Ca 0.174 0.170 0.557 0.458 0.059 0.328 Na 0.611 0.500 0.576 0.484 0.765 0.720 S i 0 2 -0.154 -0.371 -0.342 -0.627 -0.360 NA linear relationship. The results of the regression and curve fitting analysis are given in Table 3.1.6. These relationships provide input into the water chemistry modeling procedure described in Chapter 4. 6 'I • °" cn cn •IH SV S o o o o o II sx a? o\ © C N II ft a; o o o o II CX v£> 00 • E-v ° II II CL ON o o o U LO C N o o II EX | 55 o o< « u S SP o ON oo o o II CS 2 o o o o CO CO to f sv sv •a Q LO © © ! © II c5 CO o o © &5 o II o o © II IS? o o o © ll a 65 LO o LO PL. C •2 bp 33 o IS. o o © + Os o © II . * o o o II CO II i 00 o o © II CO NO o o L l CN CO CO ft c , o w sv c5 cu • o o © a, | 55 00 01 -* II ft ON o © co o U o o o © II £X 55 CM CN ON ft d te u '2 60 33 •+-> o 2 8 cu + © o o © II a 55 CO g i II c« 2 0 2 8 8" X U 2 a> iZ (l/Bui) UOJJBJJUOOUOO (l/Bui) uoijej»uaouoo (I/BUI) UO|}BJJU30UO0 (1/BlU) UOIJBJJU80UO0 CD ™ -o OD QJ <1> tm SZ o ** a. * S co C * ; O CU CO « S g " C O o >» " c o "ro" Q .£2 to E > Je ° o 2 C N $ « o £ O cu — 73 C O U J E i f cu ro a . $ •• o « 5 o cu O) (|/BLU) UOIJBJJUOOUOO (l/Btu) uoi iBJ}ueouoo 91 3.2. GROUNDWATER Groundwater samples were collected from piezometers and water table wells. Measurements of piezometric pressure head in the piezometers and water table wells were taken at the time of sampling and at other times. The chemical concentrations and head measurements from the piezometers represent the condition of groundwater at the completion interval of the piezometers, and from the wells, the average condition of the groundwater from the bottom of the well to the water table. As discussed elsewhere, the analysis focuses on bicarbonate, sulphate, calcium, sodium and silica. Two basic strategies were used to analyze chemical concentrations in groundwater. First, the average chemical concentrations at each standpipe were related to soil and standpipe properties to determine the influence on water chemistry of site characteristics and position in the groundwater flow field. Second, individual standpipes and groups of standpipes that were found to be similar in their water chemistry properties were examined to determine the effect of varying hydrologic conditions (represented by day of the year and pressure head) on groundwater chemistry within each site. Finally, these two approaches were combined to produce overall equations that would predict the concentration of each chemical according to both site characteristics and hydrologic conditions. 3.2.1 General Relationships with Hydraulic Properties and Average Concentrations A summary of average chemical concentration, standpipe depth, hydraulic conductivity (K), soil depth and type, hydrologic conditions and vegetation at each site is given in Table 3.2.1. The measured K of bedrock is in keeping with Lawson's (1968) findings from Trapping Creek (see p.37). Values of K in soil range from ll>7 to 10"5 m/s (typical of forest soils), and the conductivity measured at the seepage site is close to ll>8 m/s, typical of strucureless clay. It was felt that there might be a relationship between concentration of a given chemical in groundwater and either pressure head or day of the year, because it was assumed that residence time of water in the flow system would affect its chemistry (e.g., Jacks & Paces, 1987, Denning et al, 1992, Hendershot et al, 1992). To test this hypothesis, the overall average R . o ' * P-60 H > cu § • a s g (8 s s s I 3 "3 c "S o . I 0) u 3 u OH CD 0) u B « cu "3 c f, .cu to LO LO CN l i cu 60 IH «s Xi 60 5ti =3 * \ . cu, cu 5q 60 j ; M te Si t>. CO I N CO CU . S3 a s > te LO oo io S oo co oo od ' z a CN CO ON 60| 00 00 CO ^ cr ~> So vO ON o © PH T3 a 2 CD o CN LO PH CN CM CN CN r H CN PH PH u CU < low r H PQ PH r H PH 93 Table 3.2.2: Relationships between average chemical concentration and Hydraulic Equation R2 s.e. P [HCO l^ = -37.0 - 3.82 ln(K) 64.3% . 5.720 0.000 fCal =-8.9 - 0.88 ln(K) 63.6% 1.337 0.000 [Nal = -0.145 ln(K) 35.3% 0.571 0.000 [SiO,l = -0.752 ln(K) 59.8% 1.078 0.000 concentration of each chemical was regressed against the hydraulic conductivity in the 10 piezometers on which bail tests were conducted. Significant log-normal relationships between the average concentrations of bicarbonate, calcium, sodium and silica and the natural logarithm of hydraulic conductivity K in m/s were found (Table 3.2.2). These equations show that there is a general increase in chemical concentration as the hydraulic conductivity decreases (Figure 3.2.1). This suggests that where water moves more slowly, chemical concentration is higher because the water is in contact with the porous matrix for longer, but also suggests that other soil properties such as grain size distribution or soil structure may be involved that govern both hydraulic conductivity and groundwater chemistry. Because hydraulic conductivity was measured in situ, high conductivity may also indicate the presence of macropores that allow some water to bypass the soil matrix. This would result in lower concentrations than if all flow occurred under Darcy's Law, due to reduced contact between the water and the mineral soil particles. Soil structure and grain size distribution changes with depth in the soil, and this could affect hydraulic conductivity and thus groundwater chemistry. Initially, hydraulic conductivity K was plotted against standpipe depth (Figure 3.2.2), suggesting three different soil groupings, each of which demonstrates a relationship between standpipe depth and hydraulic conductivity. These groups and the equations that relate hydraulic conductivity to standpipe depth are summarized in Table 3.2.3. The relationship between hydraulic conductivity and depth in the soil can be explained by an increase in soil clay content with depth. There is reason to assume that the seepage site represented by PI and P2/BiOWer should occupy a different group than the other wells and piezometers because it is a permanent groundwater discharge site with bog soil (l/Biu) uonejjueouoo aBejaAV 95 Table 3.2.3: Preliminary Groundwater Groupings Suggested by Figure 3.2.2 Group Sites Equation R2(%) 1 all at 240/241 (incl. bedrock), Pl/242 log(K) = -3.03 (Depth) -10.7 83 2 P2/242, PI/A, PI and P2/B U D D er log(K) = -2.97 (Depth) -7.8 92 3 PI & P2/Biower log(K) = -0:59 (Depth) -16.7 100 and vegetation distinct from all other sites. However, the information given in Table 3.2.1 does not offer any reason for the distinction between groups 1 and 2. A reasonable supposition is that the difference is due to differences in soil grain size distribution between the two suggested groups. This will be supported by findings of a soil survey described in section 3.2.2. Two points can be deduced from the relationships described above, as follows; 1. Because there are inverse relationships between concentration and the natural logarithm of hydraulic conductivity (ln(K)), and between ln(K) and standpipe depth, then it follows that chemical concentrations increase linearly with increasing depth in the soil at a site. 2. Excluding seepage and bedrock sites, Figure 3.2.2 suggests that there are two distinct soil types, with different hydraulic conductivity regimes, and this is expected to affect groundwater chemistry. 3.2.2: Soil Survey. The above analysis suggests that soil type in Dennis Creek and Edelweiss Creek may be different from the soils in the 240 and 241 Creeks, and that these differences could affect water chemistry. However, physical and chemical groundwater data alone cannot confirm this. To overcome this difficulty, a soil survey was conducted in 240, Dennis and Edelweiss Creeks to determine grain size distribution and exchangeable cations. The sampling locations are given on the maps (Figures 3.2.3 and 3.2.4). 3.2.2a: Soil groupings according to grain size distribution From the grain size distribution data, four factors were calculated: the gravel content as a percentage, the percent fines (silt and clay), a measure of the median grain size (d50) and the standard deviation of the percentages that fell in each grain size class (Table 3.2.4). The quantity 96 Figure 3.2.3: 240 Creek Forest Cover with Soil and Stream Profile Sampling Sites Pine-Spruce-Fir Spruce-Fir Pine + 2 Soil Sampling Sites X Stream Profile Sites 0 500 1000 1500 2000 metres 97 Figure 3.2.4: Dennis-Edelweiss Creeks Forest Cover with Soil and Stream Profile Sampling Sites Pine-Fir Spruce-Fir Immature + 1 Soil Sampling Sites X Stream Profile Sites I 0 500 1000 1500 2000 metres 98 Table 3.2.4: Summary of Physical Soil Factors and Results of Discriminant Analysis Sample Comments Initial Gravel Fines d50 s.d. Final Group % % (mm) Group 240 Creek VI • P2-P4 6 8.6 8.3 0.1 10.84 6 V2 rocky/ side hill 3 23.3 12.0 0.3 524 3 V3 rocky/ridge 3 27.8 7.0 0.5 7.40 3 V4 v.rocky/swale 3 19.9 3.9 0.3 8.83 3 V5 ridge 4 19.4 2.0 0.5 9.07 3 V6 v.v.rocky/ridge 2 53.8 1.9 0.2 17.72 . 2 V7 side hill 2 30.0 2.5 0.5 10.26 3 V8 side hill 2 37.8 6.0 0.6 11.30 2 V9 rocky/swale 3 24.3 0.9 0.8 8.67 3 V10 bench 4 24.5 0.3 0.5 12.18 3 Vl l valley flat 3 27.0 1.1 0.6 8.01 3 V12 side hill 2 49.9 1.7 2.0 16.16 2 V13 v.rocky/side hill 3 . 20:7 4.1 0.5 6.41 3 V14 ridge 3 26.5 7.9 0.3 7.85 3 V15 P5 2 32.3 2.4 0.7 9.69 3 V16 P1/P2/241 3 19.5 2.8 0.3 7.74 3 Dennis Creek Dl deep sand deposit 6 9.6 3.1 0.2 15.48 6 D2 gentle hillside 4 10.2 0.3 0.3 14.40 6 D3 gentle hillside 4 11.7 0.8 0.4 13.99 6 D4 gentle hillside 4 10.8 7.6 0.2 10.04 6 D5 swale 6 3.3 5.3 0.2 11.37 6 D6 rocky 3 22.1 17.8 0.2 6.49 3 D7 boulders 3 23.1 12.0 0.3 5.50 3 D8 Pl/242 3 17.5 2.1 0.3 8.55 3 D9 P2/242 6 . 4.9 13.0 0.1 9.59 6 D10 hillside 3 17.1 1.9 0.3 7.15 3 Dll hillside 3 12.2 8.3 0.2 6.63 3 D12 A Edelweiss 3 14.6 6.5 0.3 7.27 3 D13 B Edelweiss 3 J 15.1 6.3 0.3 8.40 3 d50 is actually the grain size at which 50% of the grains are smaller and 50% are larger. This is derived from a cumulative grain size curve. The standard deviation of the percentages in each grain size class was used as a measure of how well sorted the sediment is. A high standard deviation indicates a predominance of one or two grain size classes and is indicative of a relatively high degree of sorting in the deposit from which the soil is derived. Conversely, if the 99 standard deviation is low, it indicates a more uniform distribution of grain sizes indicative of an unsorted deposit. Initially, cluster analysis was used in an attempt to find a pattern in the data based on the four factors described above, but the results pf this analysis were inconclusive. Therefore, ) discriminant analysis was used to group the soil samples. To conduct the analysis, initial groupings were assigned according to obvious differences in gravel content and standard deviation. The initial classification scheme is described in Table 3.2.5, in which the body of the table contains group numbers that were assigned according to the attributes given in the column and row headings. Note that not all groups contained in this table were used to group the samples. This scheme was essentially arbitrary and was used as a starting point for the discriminant analysis. The discriminant analysis was then run, and observations that the analysis assigned to different groups were reclassified. The analysis was then rerun. This process was , repeated one more time, at which the analysis revealed that all observations were correctly classified. Squared distances between groups were examined to determine if all groups were significantly different. The distance matrix that resulted from this grouping scheme is given (Table 3.2.6): Table 3.2.5: Group numbers assigned to soils on the basis of two factors; group numbers were % Gravel s.d. of grain size classes < 9 s.d. of grain size classes > 9 30+ 1 2 10-30 . 3 4 0-10 5 6 Table 3.2.6: Discriminant analysis of Soil Texture Groups Group Group 2 Group 3 Group 4 Group 6 2 0.000 32.043 30.283 60.505 .' 3 32.043 0.000 12.473 . 46.841 4 30.283 12.473 0.000 11.586 6 60.505 46.841 11.586 o;ooo 4 vs 6: F=2.17, Fcrit=9.12 4 vs 3: F=4.60, Fcrit=3.11 not significant significant 100 This step, revealed that there was no significant difference between groups 4 and 6. Thus, these two groups were merged together. The: analysis was rerun and observations reclassified as described above until all observations were correctly classified. The result was three distinct groups that were significantly different from each other with respect to % gravel, % fines, d50 and standard deviation as shown by the following distance matrix (Table 3.2.7): , Table 3.2.7: Final Discriminant Analysis of Soil Texture Groups Squared Distance Between Final Groups Group Group 2 , Group 3 Group 6 2 0.000 30.389 64.341 3 30.389 0.000 34.014 6 64.341 34.014 0.000 2 vs 3: F=16.45, Fcrit= 3.01 significant final grouping For convenience/the designation of the soil groups was changed to reflect the apparent extent of each group according to the samples that were collected. Thus, group 3, the largest group, is designated as Soil Type 1. Group 6, the apparent dominant type in Dennis Creek, is designated as Soil Type 2 and group 2, the smallest group, is designated as Soil Type 3. Average grain size distributions for each of the resultant soil groups were calculated for illustrative purposes. Figure 3.2.5 shows average cumulative grain size curves for each soil type showing the derivation of the d50 median grain size. Figure 3.2.6 shows the average grain size distribution of each soil group. This graph clearly shows that Type 3 has a large proportion (almost 50%) of gravel and a relatively uniform distribution of the other grain sizes. Type 2 typically has a lower gravel content than Type 1 and there is evidence of a higher degree of sorting in the depositional process, leading to a distinct peak in the fine sand range. There is no field evidence to suggest that these soils are glaciofluvial in origin, but this analysis suggests a greater degree of water sorting during the deposition of Type 2 than Type 1. It also appears that Type 2 soil is associated with the spruce-fir forest, whereas the Type 1 soil is associated with the predominantly pine forest cover type. This is evident when comparing the sampling locations to the forest cover types (Figures 3.2.1 & 2). The Type 1 soil would tend to be better drained than the Type 2 soil due to its higher gravel content and its lower degree of sorting. Pine is known to occupy well jajouiBia pajeoipui UBIU jeuy % 102 drained soils, while spruce and fir favour wetter sites. Thus, soil type and forest cover type are inter-related. 3.2.2b: Soil groupings according to available cations The available cation analysis yielded different groupings from the grain size analysis (Table 3.2.8). When the same soil groupings that defined soil texture types were used to run a discriminant analysis, the distances between groups were not significant This suggests that soil chemistry is not related to grain size distribution. Inspection of the chemistry data shows that available cations are higher among samples collected at valley bottom and streamside sites and in swales than among samples collected on ridges and on hillslopes. Also, available cations appears to be related to slope angle, being highest at streamside sites that are located at the foot of steep slopes. This could be due to downslope leaching of cations and subsequent accumulation at the bottom of the slope. This pattern is evident among samples from 240 Creek but less clear at Dennis Creek; however, available cations are much lower at Dennis Creek than at 240 Creek and this is likely to explain the differences in groundwater and streamflow chemistry between those two creeks. It is likely that the difference is due to both forest cover type and soil parent material since biological activity and soil development are inter-related. Soil types were grouped on the basis of calcium availability as follows: Type 11 3% Ca Using this classification system, discriminant analysis was run using Ca, Mg, K, and Na availability as factors. The three groups were all significantly different. An attempt was made to subdivide type 11 into two groups using 0.5% available calcium as a dividing line to distinguish topographic differences in Dennis Creek, but the two sub-groups were not significantly different from each other. Thus the final groupings can be described by the squared distance matrix (Table 3.2.9). 103 Table 3.2.8: Summary of Available Cations in Soil; classifications based on soil texture and Sample Comments/ Topography Slope Class Ca(%) Mg(%) K(%) Na(%) Soil Texture Type Soil Chem Type 240 Creek VI P2-P4 low 1.1 0.9 1.1 0.2 2 12 V2 rocky/side hill mod. 2.3 1.9 3.1 0.1 1 12 V3 rocky/ridge low 1.2 1.1 0.1 : 0.3 1 12 V4 v.rocky/swale low 1.1 0.4 0.8 0.1 1 12 V5 ridge low 1.0 0.6 0.8 0.1 1 12 V6 v. v.rocky/ridge low 0.6 .0.3 0.3 0.1 3 11 V7 side hill low 0.7 0.2 0.4 0.1 1 - 11 V8 side hill low 0.3 0.3 0.3 0.1 3 11 V9 rocky/valley bottom mod 4.5 3.4 1.1 0.1 1 13 V10 bench flat 0.6 0.4 1.0 0.1 1 11 Vl l valley flat low 6.1 5.8 1.0 0.3 1 13 V12 side hill mod 0.9 0.4 0.3 0.1 3 11 V13 v.rocky/ side hill mod 0.2 0.2 0.3 0.1 1 11 V14 ridge low 1.3 0.6 0.4 0.1 1 12 V15 P5 high 8.9 8.2 2.9 0.4 1 13 V16 P1/P2/241 low 2.0 1.1 0.4 0.2 1 12 Dennis Creek Dl deep sand deposit low 1.7 0.7 0.4 0.1 2 12 D2 gentle hillside low 1.3 0.6 0.4 0.2 2 12 D3 gentle hillside low 0.3 0.3 0.6 0.1 2 11 D4 gentle hillside low 0.5 0.7 0.7 0.2 2 11 D5 " — swale low 0.6 1.0 0.8 0.3 2 11 D6 rocky low 0.1 0.6 1.2 0.1 1 11 DZ boulders mod 0.9 0.6 . 0.8 0.1 1 11 -D8 Pl/242 mod 0,9 0.7 0.8 0.1 1 11 D9 P2/242 low 0.2 0.2 0.7 0.1 2 11 D10 hillside mod 2.1 1.2 0.5 0.3 1 12 Dll hillside mod 0.3 0.6 0.6 0.2 1 11 D12 A Edel/streamside mod 0.5 0.2 1.2 0.1 1 11 D13 I B Edel/hillslope mod 0.7 0.5 1.2 0.1 1 11 Table 3.2.9: Discriminant analysis between soil types grouped by Ca availability Type Type 11 Type 12 Type 13 1 . 0.000 5.958 93.801 2 5.958 0.000 82.022 3 93.801 82.022 0.000 1 vs 2: F=7.77; Fcrit=5.8 significant 104 There is a definite relationship between available cations and elevation at 240 Creek, although the relationship is less clear for Dennis Creek (Figures 3.2.7,8). The available cation vs elevation relationships for 240 Creek is likely a result of the effectiveness of weathering as suggested by Johnson & Reynolds (1977), whereas at Dennis Creek, this effect is combined with the effect of the variability in soil/forest cover type. 3.2.3: Determination of Between Site Variability due to Soil and Forest Cover The preceding sections suggest that site conditions of soil and forest cover govern between site variability of groundwater chemistry. Regression analysis with dummy variables was used (as described below) to analyze the effect of soil type and forest cover type on average groundwater chemistry. Figure 3.2.2 shows three distinct groups in terms of hydraulic conductivity of the soil-bedrock. The seepage sites that are characterized by Pl&P2/Bi o w e r comprise relatively small proportions of the watersheds (<1%) and do not support trees; therefore the seepage site is not a forest cover type and does not warrant inclusion in this analysis. The bedrock site (P4/240 Creek) is also excluded because bedrock is not a soil type. The soil grain size analysis suggests that the soil texture (hence, hydraulic conductivity) at Edelweiss Creek is more similar to 240 and 241 Creek soils than Dennis Creek soils, contrary to the grouping shown in Figure 3.2.2. The different groupings that resulted from the soil texture and cation analyses led to two possible dummy variable classifications for soil conditions as follows: soil type by grain size S= 1 for unsorted morainal blankets and veneers (Soill) 0 for sorted morainal blankets (Soil2) soil type by hydrology: HZ = 0 for recharging hillslope sites 1 for discharge/recharge streamside sites Dummy variables were also used to classify forest cover type: forest cover (F1,F2)= (1,0) for mature lodgepole pine (0,1) for mature spruce-fir (0,0) for immature regeneration v 105 E 3 l l £ w CO > § 1 1 ] cn > g> JS o f 9 C T3 (0 C .2 Ul c .. c 00 CO 01 Q co e 2 3 © o O o Q E 1 m % pot E CO J£ fi> niss um Cre> CO o T > O CM Q. CO _ + o + + + 0 + + *h + o i—1—i—1—r m o w 8 U 8 C O J C O O o o CM + + + CO (ft £ 1 5 LU . a CD | | 1 1 1+ I* O IQ 8 CO p CO (%) U0| 3|qB36UBM0X3 (%) uoi aiqeaBueuoxg 106 Table 3.2.10: Table of Dummy Variables of Groundwater Site Characteristics, and Land Soil by Grain Size Soil by Hydrology Forest Cover Standpipe SI HZ Slope(%) Fl F2 P2/240 1 0 12 1 0 P3/240 1 0 12 1 0 P5/240 1 1 49 1 0 Pl/241 1 1 24 1 0 P2/241 1 0 10 1 0 Wl/241 1 1 24 1 0 W2/241 1 0 10 1 0 W3/241 1 0 10 1 0 Pl/242 1 1 35 0 1 P2/242 0 0 12 0 1 Pl/A 1 1 28 0 0 Pl/Bu 1 0 30 0 ,0 P2/Bu 1 0 30 0 0 Note that the available cation analysis in soil suggested that land slope was also a factor in governing soil chemistry. Thus, slope should accompany the dummy variable HZ. A complete list of the possible dummy variables used to classify the soil as well as the forest cover variables is given in Table 3.2.10. For the purpose of modeling, the most useful parameters to predict average soil chemistry variables are those that can either be measured or estimated during a field reconnaissance, or derived from model fitting parameters. Parameters that can be measured in the field include soil type and terrain classification, forest cover type, and soil depth. However, soil depths are not known precisely at most sites. In many cases, wells and/or piezometers were drilled to an impeding layer such as hard pan (B upper), layers of large rocks (241, A) or to bedrock (240). At other sites, such as 242 Creek or B lower, drilling stopped at a maximum depth of 1.8 meters and the soil depth was estimated. The site that contains P2-P4 at 240 Creek is the only site where soil depth is known with certainty. Thus standpipe depth is a better variable to predict average chemical concentration since it is known precisely and it is a measure of the position in the groundwater flow field. 107 Hydraulic conductivity is influenced by the depth of the completion interval and the soil type as shown in Figure 3.2.2, the hydraulic conductivity generally decreasing with depth for a given soil type. This occurs because of changes in soil structure and clay content with depth. Hydraulic conductivity may also be influenced by forest cover type, due to either recent disturbance, or to coincidental differences in soils at the different sites. At forested sites and for a given depth, hydraulic conductivities are highest in the area that was recently clear-cut, intermediate under the mature spruce-fir forest, and lowest under the mature lodgepole pine. It is unlikely that forest cover could have such an effect on hydraulic conductivity; however, subsequent analyses will show that forest cover influences groundwater chemistry. The average chemical concentrations were regressed against properties of the soil and the standpipes themselves, using the dummy variables to represent soil and forest cover types. Two separate analyses were conducted to determine which soil type dummy variable is the most appropriate. Because of the demonstrated interaction between pipe depth and hydraulic conductivity, concentrations were regressed against standpipe depth, depth^ , ln(K), an interaction term between depth and ln(K) and dummy variables to represent forest cover type (F1,F2). In the first analysis, the dummy variable to represent soil type according to grain size (S) was introduced into the equation, and in the second analysis, the dummy variable to represent soil type according to hydrological activity (HZ) as well as land slope at the site were introduced ( Table 3.2.11). A comparison of the R2 values shows that soil classification using position on slope (i.e. recharge vs recharge/discharge sites) and slope angle provides better prediction of average chemical concentrations in groundwater than soil type based on grain size distribution. However, not all variables were used as predictors for all chemicals considered here. Sulphate depended only on slope angle and forest cover type, bicarbonate depended on hydraulic conductivity, slope angle and forest cover type, calcium depended on standpipe depth, hydraulic conductivity, hydrological zone, slope angle and forest cover type, sodium depended only on hydrological zone and slope angle, and silica depended on depth and hydraulic 108 Table 3.2.11: Regression analyses of average chemical data using different dummy variables Equation using grain size type R2 Equation using soil chemistry type R2 [SO4] = - 2.11 - 0.225 ln(K) 50.7 [SO4] = 0.018 Slope + 0.61 Fl + 0.50 F2 61.7 [HCO3] = -1.33 ln(K) - 7.37 Fl -6.69F2 76.2 [HCO3] = - 2.24 ln(K) + 0.133 Slope - 7.28 Fl - 7.42 F2 - 15.0 94.0 [Ca] = - 0.663 ln(K) - 1.83 Fl - 1.92 F2 - 4.94 82.1 [Ca] = - 0.38 D-ln(K) - 2.19 depth2 - 0.91 HZ + 0.04 Slope -1.03 Fl - 0.78 F2 97.0 [Na] = -23.3 -1.93 ln(K) + 1.50 D-ln(K) + 19.6 depth 53.4 [Na] = - 0.378 HZ + 0.0512 Slope + 0;76 82.9 [Si0 2] = - 0.715 ln(K) + 0.63 D-ln(K) + 8.67 depth 48.4 [SiOoj = - 0.715 ln(K) + 0.63 D-ln(K) + 8.67 depth 48.4 conductivity and was independent of soil or forest cover types. However on closer examination the equations governing calcium and sodium concentrations suggest a negative relationship between the dummy variable HZ and chemical concentration. A value of 1 for HZ indicates recharge/discharge (i.e. streamside) site. Both the average groundwater chemistry data and the soil chemistry data indicate that these sites tend to exhibit higher chemical concentrations than recharging hillslope sites, all other factors being equal. Thus the equations in Table 3.2.11 do not represent the effect of seasonal groundwater discharge on chemical concentration. This is probably because the sample size of groundwater wells and piezometers is too small and the discharge/recharge effect is masked by other factors, specifically land slope. Slope always has a positive relationship with chemical concentration. This is because among the groundwater monitoring sites, the steepest slopes were found adjacent to streams, reflecting the fact that all main stream channels are incised to some extent. There is no reason to believe that in the upper parts of 240 and Dennis Creeks, groundwater chemical concentrations will be higher than on moderate mid slope sections; in fact, the opposite should be true. Thus, it is concluded that the incised streamside sites and the hillslope sites represent two distinct hydrological zones that should be treated separately. To address this issue the forested groundwater sites in soil (i.e., excluding seepage and bedrock) were divided into two groups; group 1 represents the standpipes located at incised stream-side sites, and group 2 represents those standpipes located on hillsides above the group 1 109 Table 3.2.12: Correlation Matrix for Group 1 (streamside sites) s o 4 HCO3 Ca Na S i 0 2 ln(K) Slope Depth Fl H C 0 3 0.676 Ca 0.661 0.980 Na 0.846 0.865 . 0.777 S i 0 2 -0.263 -0.415 -0.559 -0.076 ln(K) -0.684 -0.178 -0.237 -0.315 0.043 Slope 0.961 0.828 0.785 0.956 -0.225 -0.530 Depth -0.093 0.286 0.453 -0.201 -0.840 0.013 -0.089 Fl 0.134 -0.432 -0.551 0.059 0.815 -0.304 0.043 -0.927 F2 0.286 0.221 0.407 -0.082 -0.846 -0.460 0.160 0.830 -0.612 Table 3.2.13: Correlation Matrix for Group 2 (hillslope sites) SO4 HCO3 Ca Na S i 0 2 ln(K) Slope Depth Fl HCG-3 -0.201 Ca 0.094 0.951 Na -0.282 0.960 0.881 S i 0 2 -0.229 0.774 0.726 0.829 ln(K) -0.859 0.040 -0.219 0.220 -0.002 Slope -0.611 0.881 0.705 0.912 0.676 0.512 Depth -0.256 0.662 0.567 0.569 0.676 -0.191 0.555 Fl 0.714 -0.615 -0.378 -0.620 -0.610 -0.394 -0.772 -0.838 F2 -0.190 -0.198 -0.300 -0.244 0.086 -0.187 -0.171 0.592 -0.488 sites. Because of the small sample size of these subgroups, only one independent variable can be used to predict chemical concentrations, along with the dummy variables that will predict the effect that forest cover type has on groundwater chemistry. To determine the best variable to use as a predictor, average chemical concentrations were correlated with ln(K), standpipe depth and land slope as well as the dummy variables Fl and F2. Correlation matrices for each of the subgroups are given in Tables 3.2.12 and 3.2.13 respectively. In most cases, average chemical concentrations were most strongly correlated with land slope. Exceptions to this are silica in group 1, which was most strongly correlated with standpipe depth, and sulphate in group 2 which was most strongly correlated with ln(K). The matrices also show roughly 50% correlation between ln(K) and slope. In group 1, the correlation is negative indicating a decrease in hydraulic conductivity as slope increases, and in group 2 it is positive, indicating an increase in hydraulic conductivity with slope. This is likely because of downslope movement of clay sized particles as the soils developed and the fact that those clay fractions will be eroded from the 110 hillslopes and deposited at the streamsides sites. Steeper streamside soils would accumulate more clays over time than gentler slopes because of more rapid deposition, whereas steeper hillside soils would tend to lose more clay fractions due to more rapid erosion. However, chemical concentrations are positively correlated with slope in both groups with the exception of sulphate in group 2 and silica in group 1. This is likely a result of leaching of those chemical from sources higher up the hillsides, since all groundwater sites are located on lower slopes. In this case, steeper slopes would result in more rapid leaching into the groundwater flow field on the lower slopes. For each group, average chemical concentrations were regressed against land slope and the dummy variables Fl and F2. Both first and second order models in slope were considered. The resultant best fit equations for both groups were reduced by solving for the dummy variables (Table 3.2.14). The symbol 0 is used to represent the land slope in degrees. Standpipe depth was also used as a predictor for silica, represented by the symbol G>. Those relationships are represented by graphs (Figures 3.2.9 to 3^ 2.14). These graphs show that in each of the two hydrological zones represented here there is a positive relationship between land slope and chemical concentration, and that the slope of that relationship differs from group 1 to group 2. The graphs appear to demonstrate that forest cover has an effect on groundwater chemistry, particularly in group 2. However, the subdivision of the represented standpipes into two subgroups has resulted in very small sample sizes such that it is difficult to separate the effect of forest cover from that of land slope. In some cases (e.g., group 2 site with mature spruce-fir, group 2 site with regen.) the reduced regression lines aire drawn through 2 data points. The significance of the dummy variables in these cases suggests parallel lines between three forest cover types based on a significant relationship between chemical concentration and land slope on only one of those forest cover types. Thus, the dashed lines shown on the graphs (Figures 3.2.9-14) are lines suggested by the significant dummy variables. The importance of the two hydrological zones in groundwater within forested soils (incised streamside zone vs hillslope 0) ° u N c — O n &S SI 2 " ° CD CU > Q. < o CO CD CL k. CD > O o cu c: o N o Q . CO o o CM o o o 00 (l/Bui) uouejjueouoo ejeuoqjeoia OBBJSAV c O CO •JJ; CD CO c is o C N 0) — (0 o r c o C O Q . Sl§ t>«5 <0 CD O CO © * © > CD k-< Q- SZ O h-o> C O T J CN CO CD 1 -o CO CO c CL o CL CO CD CD CD c CD CD CU 1_ >-il £ 6 = " ra 2 E 13 O L-U c ; _ J CO — CM CM CM Ct- CL > fl) 3 3 3 3 $ s s s e s O CD CD CD CD o LO o 9 V co C o ro co c o ° cn. O o ^ P O O If g W o w 2 «- u_ £ £ co fl) u t>i o K ^ CO "O « T3 ' CO C 0) ^ (0 CO 3 .SP > U_ c C D D ) C O C O c Q . co ra o 8 2 o. co CL 3 e CL 2 O C O 1 C O ra o _i C N C L 3 2 O HH 112 o * ; "I H I o O N 5i i n o CO IT) csi o o (l/Biu) uoijejjuaouoo cumpos eBejeAY/ c . 2 w 4 - i CO *-> o C N 0) _ U (B C U ° o> o o E l = • 0 o >» 75 i O o CO £ (0 -o > (A c l t • CO to E CO cu CO > O o (A 2 o l i -CO CO CO Q . 3 o cu c O-o C L CO o> 73 O ti. «> I — 8 a Q . CO CO ra CO I— 3 E E M M I N C L C L C L 3 3 3 o o o CD CD CD CD HI to r o T 1 1 o i — i — i — i — r O O p co csi *-(l/Bui) UO!»BJJU80UOO uinpiBO oBejOAY/ o ~ " g i l I I o O " CO p o o (l/Bui) u o i i B J j u e o u o o BOMS eBejeAV (0 o CL CO > o o = (0 O q> '„ c re o <- N c — co re a o If re -0 — ' — >*. 5 ) i a o fl I* ^ CO 5 t 01 co -o « 1 re CO 5 13 LIS c o N CO •g co X3 O C L o CO o ZJ c CO cn co _ CD ZJ CO E E C L w CM CM CM a. a O O O O O H + + o o " I Si a I o O ^ CO o csi T o o o o o CO (l/Bui) UOUBJJUOOUOO earns OBBJOAV 114 Table 3.2.14: Equations Describing Average Chemical Concentration vs Land Slope (0) and Pipe Depth () for Two Hydrological Zones in Soil Chemical Equation Site R2 s.e. P Sulphate [S04] = 0.031 0 Group 1 Streamside zone 89.9 0.1536 0.000 fS04] = 0.180 0 -1.04 Group 2 Lodgepole pine 88.2 0.1165 0.025 fS04] = 0.180 0 -1.51 Group 2 Spruce-Fir 88.2 0.1165 0.025 fS04l = 0.180 0-3.11 Group 2 Immature Regen 88.2 0.1165 0.025 Bicarbonate [HCO3I = 0.379 0 Group 1 Streamside zone 58.1 2.275 0.000 [HCO3I = 2.16 0 -13.8 Group 2 Lodgepole pine 95.4 1.021 0.004 [HCO3] = 2.16 0 - 16.5 Group 2 Spruce-Fir 95.4 1.021 0.004 fHC0 3] = 2.16 0 - 27.1 Group 2 Immature Regen 95.4 1.021 0.004 Calcium fCa] = 0.1 0-0.00072 0 2 Group 1 Streamside zone 75.9 0.4556 0.003 fCa] = 0.609 0 - 4.50 Group 2 Lodgepole pine 92.4 0.2677 0.011 [Cal = 0.609 0 - 5.57 Group 2 Spruce-Fir 92.4 0.2677 0.011 [Cal = 0.609 0 - 9.13 Group 2 Immature Regen 92.4 0.2677 0.011 Sodium [Na] = 0.074 © - 0.337 Group 1 Lodgepole pine 99.6 0.0567 0.000 [Na] = 0.074 0 - 0.683 Group 1 Spruce-Fir 99.6 0.0567 0.000 [Nal = 0.074 0 Group 1 Immature Regen 99.6 0.0567 0.000 [Na] = 0.176 0 - 0.49 Group 2 Lodgepole pine 96.0 0.0904 0.003 [Nal = 0.176 0 - 0.76 Group 2 Spruce-Fir 96.0 0.0904 0.003 [Nal = 0-176 0 - 1.48 Group 2 Immature Regen 96.0 0.0904 0.003 Silica [Si02l = 8.32 VF) was added as a third analysis. The groundwater chemistry data from wells as well as piezometers at 241 Creek were included in the analysis. The standpipes there are much shallower than at other plots due to the shallow soils at the site chosen to measure groundwater. Because of this, it was not possible to drill nested piezometers, so water samples were collected from both the wells and piezometers. However, unlike the piezometer samples, the well samples represent the average chemical concentration of the saturated zone above the bottom of the standpipe, at the average pressure head for the same profile. The pressure head at the bottom of each well was assumed proportional to the ratio of the depth of the well to the depth of the corresponding piezometer multiplied by the pressure head at that time in the piezometer. The pressure head at the water table is always zero. Samples and measurements were collected simultaneously from the well and piezometer at each of sites 1 and 2. Thus, the average pressure head ¥ for wells Wl and W2 at 241 creek were calculated as: Vwi = 0.92CFP1)/2 ^ W 2 = 1.17CPp2)/2 1 1 8 Since there is no piezometer sited next to W3, a different approach was used to estimate the average pressure head at the sampling times. The average pressure heads as calculated above for Wl and W2 were regressed against the corresponding depth of the water table below the surface H for the following relationship: ¥ = 24.6 + 0.205(H), R2=47.1% where H is a negative value in metres below the surface. This equation was used to estimate the average ¥ values in metres for the samples collected from W3. In most cases, the relationship of chemical concentration vs day of the year is the most significant of the three analyses. There are some exceptions, particularly where the relationship with D is not significant. It seems reasonable to conclude that chemical concentration in groundwater is more directly related to residence time than to pressure head, however these factors are interrelated. Where pressure head is high, groundwater flow rates are higher because of increased hydrostatic pressure and because groundwater flow enters upper soil layers where hydraulic conductivities are higher than at depth. Thus, residence times should decrease with increasing pressure head. Both groundwater levels and chemical concentrations are driven by inputs of new water from snowmelt which itself follows a predictable pattern according to the day of the year. Because both pressure head and chemical concentration each have associated variabilities with input of new water, those variabilities may be additive in the relationship of concentration vs pressure head. This can be illustrated by using P3/240 as an example. Figure 3.2.15 shows a plot of pressure head measured in P3/240 vs adjusted day of the year for the period 1988-90 with a regression line that represents the average relationship between those two quantities for the three year period. Note that the piezometer is normally dry from late October to early March. The same graph presents a plot of silica concentration vs adjusted day with a third order polynomial relating concentration to time, showing the inverse relationship to that of pressure head. The resultant linear relationship between silica concentration and ¥ on P3 is shown in Figure 3.2.16. 119 0 o o o o 01 T* O CF) CO ( | /6Ui) UOIJBJJUSOUOO BO||!S (l/DW) uo!)ej)U93uo3 eojus »• to CM T- o d d d d d d d OK/Cd ie (LU) pB3H ajnssojd 120 3.2.5: Timing of Sampling and its Effect on Chemistry vs Time/Head Relationships. Any equations that predict groundwater chemistry variability as a function of adjusted day of the year or of pressure head are only valid within the range of the observations. Groundwater samples were collected as early as April 4th and as late as October 7th, although many of the piezometers and wells were only active for part of that period. In particular, the shallower standpipes such as P3/240 Creek, the upper sites at Edelweiss Creek and all sites at 241 Creek are active only in moderate to high flow conditions. These standpipes sample the more transient upper groundwater, whereas pipes that sample deeper groundwater and pipes that sample discharging groundwater are always active. Also, sampling of groundwater began at different times in different sites. Sampling began in 1988 at 240 Creek, in 1989 at 241 and Edelweiss Creeks. At these sites, the range of conditions represented by the sampling was considered a sufficient basis to establish representative groundwater chemistry relationships. Samples were not collected from groundwater at 242 Creek before April of 1990. Sampling began at P2 on April 4*h and at PI on May 10*h. The soil usually freezes at most sites in the study area, but in the winter of 1989-90, soil freezing was patchy due to heavy snow early in the season. Peak runoff occurred on May 29m due to a rainfall event of 14.5 mm that fell on the previous day following 108.5 mm of precipitation over three weeks, most of which fell as rain. The rainfall had little effect on the snowpack because its temperature was close to 0 degrees C. Without significant snowmelt prior to this event to add new water to the soil, the relatively high chemical concentrations in groundwater collected during this event (particularly at PI, which is a groundwater discharge site) could be due to the rapid displacement of old water by the addition of a large volume of rainfall onto soil with very wet antecedent conditions. A similar increase in concentrations occurred at P2 during the event of May 28 -^30^ but to a much smaller extent Only one sample was collected at each of PI and P2 in 1991, in August Therefore, it was necessary to return to 242 Creek in the Spring of 1992 to collect more samples to ensure that the resulting water chemistry relationships were representative of average conditions. In 1992, there was a below average snowpack with early (May 6th) peak flows, 121 compared to the above average snowfall and late peak flows of 1990. The combination of these two seasons' data should therefore represent near average conditions. While D was generally a better predictor of concentration than *P, *F is actually the more reliable predictor. Since groundwater was not sampled during the winter, winter groundwater chemistry would have to be extrapolated from fitted curves. The polynomial used to model silica concentration in Figure 3.2.15 could not be used for this purpose. However, the full range of groundwater heads that occurred during the study was captured by the sampling program. Therefore, the use of equations that model chemical concentration on some measure of groundwater storage such as *P (e.g. Figure 3.2.16) does not involve extrapolation. For this reason, it was decided that groundwater head should be used to model within site variability. 3.2.6: Groundwater Chemistry Model Based on Site Properties and Groundwater Storage. 3.2.6a: Model development Using the findings of the analysis of average chemical concentrations vs standpipe and site properties, the standpipes were grouped according to hydrological zone, forest cover type and slope class. Regression comparison analysis (Kozak, 1970) was used to verify that in most cases, standpipes within each group suggested by hydrological zone, slope class and vegetation type shared a common regression equation of chemical concentration vs pressure head, and to separate groups because of dissimilarity. W3 was eliminated from the hillside group at 241 Creek because it was not similar to any other standpipe. Within zone 2 (hillslopes), P2&W2 at 241 Creek were found to be significantly different from P2&P3 at 240 Creek. This resulted in 10 groups (Table 3.2.15). , In the analysis of average groundwater chemistry described above, average chemical concentrations in groundwater were found to vary significantly with land slope and with hydraulic conductivity. The intent of this analysis was to use some measure of the groundwater level as the main predictor variable, and to incorporate land slope or hydraulic conductivity and dummy variables to represent hydrological zone and forest cover type in a general regression 122 Table 3.2.15: Standpipe Groups Based on Hydrologic Zones Group Slope HZ1 HZ2 SL1 SL2 Fl F2 Description P2&3/240 Zone 2 12 0 0 0 1 1 0 Hillside Lodgepole Pine P4/240 Zone 4 12 0 1 0 1 1 0 Bedrock Lodgepole Pine P5/240 Zone 1 49 1 0 1 1 1 0 Dis/recharge Lodgepole Pine P1&W1/241 Zonel 24 1 0 1 0 1 0 Dis/recharge Lodgepole Pine P2&W2/241 Zone 2 10 0 0 0 1 1 0 Hillside Lodgepole Pine P1A Zone 1 28 1 0 1 0 0 0 Dis/recharge immature regen. Pl&2/Bi o w e r Zone 3 20 1 1 1 0 1 1 Seepage bog vegetation Pl&2/BUX(I)er Zone 2 20 0 0 1 0 0 0 Hillside immature regen. Pl/242 Zone 1 35 1 0 1 1 0 1 Dis/recharge Spruce-Fir P2/242 Zone 2 12 0 0 0 1 0 1 Hillside Spruce-Fir equation lumping all the groundwater sites together. Groundwater level alone without the dummy variables cannot be used as a predictor of chemical concentrations because deeper standpipes experience higher water levels such that the between standpipe variability masks the within standpipe variability. Initially, pressure head was used to predict groundwater level. Because some of the groups involve two standpipes of different depths, the relationships between concentrations and pressure head within those groups were masked by differences in average pressure head. That is, deeper standpipes would have higher pressure head for a given groundwater level. Thus total head proved to be a better predictor variable because total head is independent of standpipe depth. At each site, total head was calculated using the position of the bedrock surface derived from the estimated soil depth as datum. It was felt that this quantity would provide a realistic measure of the groundwater storage. Two separate analyses were conducted. Chemical concentrations were then regressed against total head, dummy variables that represent hydrological zone and vegetation type and interaction terms, and land slope in one analysis and hydraulic conductivity in the other. The interaction terms are the products of slope and head, hydraulic conductivity and head and dummy variables and head. Thus, dummy variables affect the intercepts of relationships whereas interaction terms affect the slopes of the relationships between chemical concentration and total head from group to group. In Tables 3.2.16-20, the results of regression analysis are given with reduced equations representing chemical concentration vs total head for each group. Those relationships are plotted in Figures 3.2.17 to 123 Table 3.2.16: Sulphate vs Total Head a: Using Slope: Zones 1,3,4 model is: [S04] = 1.39 + 0.0139 S'H - 0.856 HZ1 - 0.565 HZ2 s = 0.3505, R2 = 50.0%, p = 0.000 Zone 2 model is: [SO4] = 0.400 S-H - 8.65 H + 0.0417 Slope + 0.860 Fl - 0.182 F2 + 3.55 Fl'H + 4.09 F2'H s = 0.2028, R2 = 65.8%, p = 0.000 Group Slope HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 12 0 , 0 1 0 [SO4] = 1.36 - 0.300 (H) P4/240 Bedrock 12 0 1 1 0 rS04l = 0.83 + 0.167 (H) P5/240 Zone 1 49 1 0 1 0 [SO4I = 0.53 + 0.681 (H) P1&W1/241 Zone 1 24 1 0 1 0 [SO4] = 0.53 + .333 (H) Hillside 241 Zone 2 10 0 0 1 0 [SO4] = 1.28 - 1.100 (H) P1A Zone 1 28 1 0 0 0 [SO4I = 0.53 + 0.39 (H) Seepage 20 1 1 1 1 [SO4I = 0.278 (H) - 0.03 Hillside Edel B Zone 2 20 0 0 0 0 rS04l = 0.83 - 0.650 (H) Pl/242 Zone 1 35 1 0 0 1 [SO4I = 0.53 + 0.487 (H) P2/242 Zone 2 12 0 0 0 1 [SO4] = 0.32 + .240 (H) b: Using Hydraulic Conductivity: Zones 1,3,4 model is: [SO4] = 1.86 - 0.0554 Hin(K) -1.49 HZ1 -1.67 HZ2 s = 0.3510, R2 = 50.5%, p = 0.000 Zone 2 model is: [SO4] = - 8.39 H - 0.0647 ln(K) - 0.648 Hln(K) + 0.446 Fl - 0.519 F2 - 0.723 Fl'H - 0.176 F2'H s = 0.2054, R2 = 64.3%, p = 0.000 Group ln(K) HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 -13.66 0 0 1 0 rSC>4l = 1.33 - 0.261 (H) P4/240 Bedrock -16.34 0 1 1 0 [SO4I = 0.19 + 0.905 (H) P5/240 Zone 1 -13.99 1 0 1 0 [SO4] = 0.37 + 0.775 (H) P1&W1/241 Zone 1 -13.63 1 0 1 0 rSC>4l = 0.37 + 0.756 (H) Hillside 241 Zone 2 -12.59 0 0 1 0 rSC>4l = 1.26 - 0.955 (H) P1A Zone 1 -11.42 1 0 0 0 [SO4I = 0.37+ 0.633 (H) Seepage -17.45 1 1 1 1 rS04l = 0.967 (H) -1.3 Hillside Edel B Zone 2 -12.09 0 0 0 0 [SO4] = 0.78 - 0.555 (H) Pl/242 Zone 1 -14.05 1 0 0 1 [SO4I = 0.37 + 0.778 (H) P2/242 Zone 2 -13.55 0 0 0 1 [SO4I = 0.36 + 0.214 (H) ) 124 Table 3.2.17: Bicarbonate vs Total Head a: Using Slope: Zones 1,3,4 model is: [HCO3] = -15.8 - 0.083 S'H + 0.678 Slope + 16.9 HZ1 + 70.1 HZ2 - 5.5 Fl - 22.5 F2 - 8.1 HZTH - 28.9 HZ2'H + 4.1 FTH + 21.4 F2'H s = 4.447, R2 = 86.9%, p = 0.000 Zone 2 model is: [HCO3] - - 54.0 - 0.71 H + 3.52 Slope + 30.9 Fl + 21.4 F2 - 7.82 FTH + 0.090 F2'H s = 1.930, R2 = 67.5%, p = 0.000 better fit if Fl is set to 0, intercept becomes 62.4 Group Slope HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 12 0 0 1 0 [HCO3] = 19.1 - 8.53 (H) P4/240 Bedrock 12 0 1 1 0 f H C 0 3 l = 56.9 - 25.8 (H) ** P5/240 Zone 1 49 1 0 1 0 [HCO3I = 28.8 - 8.07(H) P1&W1/241 Zone 1 24 1 0 1 0 [HCO3I = 11.9 - 6.0 (H) Hillside 241 Zone 2 10 0 0 , 1 0 f H C 0 3 l = 12.1 - 8.53 (H) P1A Zone 1 28 1 0 0 0 f H C 0 3 l = 20.0 - 10.4 (H) Seepage 20 1 1 1 1 [HCO3] = 56.8 - 13.16 (H) Hillside Edel B Zone 2 20 0 0 0 0 f H C 0 3 l = 16.4 - 0.71 (H) Pl/242 Zone 1 35 1 0 0 1 [HCO3I = 2.33 + 10.40 (H) P2/242 Zone 2 12 0 0 0 1 [HCO3I = 9.64 - 0.62(H) b: Using Hydraulic Conductivity: Zones 1,3,4 model is: [HCO3] = - 520 - 39.8 ln(K) + 79.7 HZ1 + 12.6 HZ2 - 90.4 Fl - 111 F2 - 0.42 HZTH -15.3 HZ2*H - 6.18 FTH + 3.67 F2'H s = 4.422, R2 = 86.9%, p = 0.000 Zone 2 model is: [HC03] = - 52.2 - 0.71 H - 5.66 ln(K) - 6.52 Fl - 15.0 F2 - 7.36 FTH + 0.091 F2"H H not significant s = 1.950, R2 = 66.8%, p = 0.000 Group ln(K) HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 -13.66 0 0 1 0 [HCC>3l = 18.6 -8.07 (H) P4/240 Bedrock -16.34 0 1 1 0 [HCO3] = 52.5 - 21.48 (H) P5/240 Zonel -13.99 1 0 1 0 [HCO3I = 26.1 - 6.60 (H) P1&W1/241 Zonel -13.63 1 0 1 0 [HCO3I = 11.8 - 6.60 (H) Hillside 241 Zone'2 -12.59 0 0 1 0 [HCO3] = 12.5 - 8.07 (H) P1A Zone 1 -11.42 1 0 0 0 fHC03] = 14.2- 0.42(H) Seepage 47.45 1 1 1 1 fHCC>3] = 65.4 - 18.23 (H) Hillside Edel B Zone 2 -12.09 0 0 0 0 [HCO3] = 16.2 - 0.71 (H) Pl/242 Zonel -14.05 1 0 0 1 rHC03l = 7.9 + 3.25 (H) P2/242 Zone 2 -13.55 0 0 0 1 [HCO3I = 9.49 - 0.62(H) 1 2 5 Table 3.2.18: Calcium vs Total Head a: Using Slope: Zones 1,3,4 model is: [Ca] = 0.0436 S'H + 3.76 HZ1 + 15.8 HZ2 - 1.78 Fl - 4.65 F2 - 3.21 H Z r H - 8.24 HZ2-H + 1.50 F1*H + 5.42 F2'H s = 0.9968 R2 = 89.5% p = 0.000 Zone 2 model is: [Ca] = -10.3 + 0.665 Slope + 5.78 Fl + 3.72 F2 - 0.921 Fl'H + 0.031 F2*H s = 0.7101 R2 = 46.4% p = 0.000 Group Slope HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 12 0 0 1 0 [Ca] = 3.46 - 0.921 (H) P4/240 Bedrock 12 0 1 1 0 [Cal = 14.02 - 6.217 (H) P5/240 Zone 1 49 1 0 1 0 [Ca] = 1.98 + 0.426 (H) P1&W1/241 Zonel 24 1 0 1 0 [Ca] = 1.98 - 0.664 (H) Hillside 241 Zone 2 10 0 0 1 0 [Ca] = 2.13 - 0.921 (H) P1A Zone 1 28 1 0 4 0 0 [Ca] = 3.76 -1.989 (H) Seepage 20 1 1 1 1 [Ca] = 13.13 - 3.658 (H) Hillside Edel B Zone 2 20 0 0 0 0 [Ca] = 3.00 Pl/242 Zonel 35 1 0 0 1 [Cal = 3.736 (H) - 0.89 P2/242 Zone 2 12 0 0 0 1 [Cal = 1-40 + 0.031 (H) b: Using Hydraulic Conductivity: Zones 1,3,4 model is: [Ca] = 9.08 - 9.84 H + 0.780 ln(K) - 0.755 Hln(K) + 3.09 HZ1 + 17.9 HZ2 + 0.20 HZ1"H - 8.95 HZ2-H s = 1.004 R2 = 89.1% p = 0.000 Zone 2 model is: [Ca] = -15.7 -1.55 ln(K) - 0.896 Fl - 3.86 F2 - 2.24 Fl'H + 0.031 F2"H s = 0.6921 R2 = 50.9% p = 0.000 Group lri(K) HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 -13.66 0 0 1 0 [Cal = 4.58 - 2.24 (H) P4/240 Bedrock -16.34 0 1 1 0 [Cal = 14.23 - 6.453 (H) P5/240 Zone 1 -13.99 1 0 1 0 [Cal = 1.26 + 0.922 (H) P1&W1/241 Zone 1 -13.63 1 0 1 0 ICal = 1.54 - 0.651 (H) Hillside 241 Zone 2 -12.59 0 0 1 0 [Cal = 2.92 - 2.24 (H) P1A Zone 1 -11.42 1 0 0 0 [Cal = 3.26 -1.018 (H) Seepage -17.45 1 1 1 1 [Cal = 16.46 - 5.415 (H) Hillside Edel B Zone 2 -12.09 0 0 0 0 [Ca] = 3.04 Pl/242 Zone 1 -14.05 1 0 0 1 [Cal = 1.21 + 0.968 (H) P2/242 Zone 2 -13.55 0 0 0 1 [Ca] = 1.44 + 0.031 (H) 126 Table 3.2.19: Sodium vs Total Head a: Using Slope: Zones 1,3,4 model is: [Na] = 4.70 - 5.62 H + 0.101 Slope - 5.15 HZ1 + 1.35 HZ2 - 0.263 Fl - 0.909 F2 + 5.09 HZl'H + 1.07HZ2-H s = 0.7919 R2 = 47.3% p = 0.000 Zone 2 model is: [Na] = 2.84 -12.3 H + 0.536 S'H - 0.629 Fl - 0.501 F2 + 5.14 Fl'H + 5.16 F2'H s = 0.5478 R2 = 25.1% p = 0.000 Group Slope HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 12 0 0 1 0 [Na] = 2.21 - 0.728 (H) P4/240 Bedrock 12 0 1 1 0 [Na] = 7.00 - 4.550 (H) P5/240 Zone 1 49 1 0 1 0 INal = 4.24 - 0.530 (H) P1&W1/241 Zonel 24 1 0 1 0 [Na] = 1.71 - 0.530 (H) Hillside 241 Zone 2 10 0 0 1 0 [Na] = 2.21 - 1.800 (H) P1A Zonel 28 1 0 0 0 [Na] = 2.38 - 0.530 (H) Seepage 20 1 1 1 1 [Na] = 1.75 + 0.540 (H) Hillside Edel B Zone 2 20 0 0 0 0 [Na] = 2.84 - 1.580 (H) Pl/242 Zonel 35 1 0 0 1 [Na] = 2.18 - 0.530 (H) P2/242 Zone 2 12 0 0 0 1 [Na] = 2.34 - 0.708 (H) b: Using Hydraulic Conductivity: Zones 1,3,4 model is: [Na] = - 83.7 - 5.85 H - 7.07 ln(K) + 5.39 HZ1 - 8.42 HZ2 - 16.3 Fl - 18.8 F2 + 5.30 HZl'H + 1.22HZ2-H s = 0.7903 R2 = 47.5% p = 0.000 Zone 2 model is: [Na] = 2.96 -11.4 H - 0.795 Hln(K) - 0.882 Fl - 0.689 F2 s = 0.5388 R2 = 24.5% p = 0.000 Group ln(K) HZ1 HZ2 Fl F2 Equation P2&3/240 Zone 2 -13.66 0 0 1 0 [Nal = 2.08 - 0.540 (H) P4/240 Bedrock -16.34 0 1 1 0 [Na] = 7.10 - 4.630 (H) P5/240 Zone 1 -13.99 1 0 1 0 TNa] = 4.30 - 0.550 (H) P1&W1/241 Zone 1 -13.63 1 0 - 1 0 [Na] = 1.75 - 0.550 (H) Hillside 241 Zone 2 -12.59 0 0 1 0 [Nal = 2.08 - 1.391 (H) P1A Zonel -11.42 1 0 0 0 [Nal = 2.43 - 0.550 (H) Seepage -17.45 1 1 1 1 [Nal = 1-54 + 0.670 (H) Hillside Edel B Zone 2 -12.09 0 0 0 0 [Na] = 2.96 -1.788 (H) Pl/242 Zonel -14.05 1 0 0 1 [Nal = 2.22 - 0.550 (H) P2/242 Zone 2 -13.55 0 0 0 1 [Na] = 2.27 - 0.628 (H) ( Table 3.2.20: Silica vs Total Head 127 a: Using Slope: Zones 1,3,4 model is: [Si02] = 0.0754 S-H + 15.7 HZl + 21.6 HZ2 - 3.68 Fl - 8.65 F2 - 13.3 HZl'H - 15.3 HZ2*H + 8.41 Fl'H + 13.3 F2-H s = 1.302 R2 = 75.0% p = 0.000 Zone 2 model is: [SiG-2] = - 8.93 + 1.06 Slope - 0.142 S*H + 7.96 Fl + 8.19 F2 s = 0.6037 R2 = 61.8% p = 0.000 Group Slope HZl HZ2 Fl F2 Equation P2&3/240 Zone 2 12 0 • 0 1 0 fSi02l = 11.75 -1.704 (H) P4/240 Bedrock 12 0 1 1 0 [Si02] = 17.92 - 6.810 (H) P5/240 Zone 1 49 1 0 1 0 [Si02] = 11.59 -1.020 (H) P1&W1/241 Zone 1 ' 24 1 0 1 0 [Si02l = 11.59 - 2.270 (H) Hillside 241 Zone 2 10 0 0 1 0 fSi02l = 9.63 - 1.420 (H) P1A Zone 1 28 1 0 0 0 fSi02l = 15.70 -11.180 (H) Seepage 20 1 1 1 1 fSi02] = 24.97 - 5.380(H) Hillside Edel B Zone 2 20 0 0 0 0 rSi02l = 12.27 - 2.840 (H) Pl/242 Zone 1 35 1 0 0 1 [Si02] = 7.05 + 2.640 (H) P2/242 Zone 2 12 0 0 0 1 fSi02] = 11.98 -1.704 (H) b: Using Hydraulic Conductivity: Zones 1,3,4 model is: [Si02] = - 0.301 Hln(K) + 15.7 HZl + 21.1 HZ2 - 4.88 Fl - 9.63 F2 - 14.7 HZl'H - 18.9 HZ2'H + 9.88 Fl'H + 14.4 F2'H s = 1.300 £ = 83.5% p = 0.000 Zone 2 model is: [Si02] = -12.7 - 2.10 ln(K) + 0.304 H'ln(K) - 3.77 Fl - 4.35 F2 + 1.89 Fl'H + 2.83 F2'H s = 0.5898 R2 = 62.5% p = 0.000 Group ln(K) HZl HZ2 Fl F2 Equation P2&3/240 Zone 2 -13.66 0 0 1 0 fSi02] = 12.22 - 2.263 (H) P4/240 Bedrock -16.34 0 1 1 0 [Si02l = 16.22 - 4.102 (H) P5/240 Zone 1 -13.99 1 0 1 0 [SiC>2l = 10.82 - 0.609 (H) P1&W1/241 Zone 1 -13.63 1 0 1 0 [SiC>2] = 10.82 - 0.717 (H) Hillside 241 Zone 2 -12.59 0 0 1 0 [Si02] = 9.97 - 1.937 (H) P1A Zone 1 -11.42 1 0 0 0 [Si02] = 15.70 - 11.263 (H) Seepage -17.45 1 1 1 1 fSiC>2l = 22.29 - 4.068 (H) Hillside Edel B Zone 2 -12.09 0 0 0 0 [Si02] = 12.27 - 2.840(H) Pl/242 Zonel -14.05 1 0 0 1 [SiC>2] = 6.07 + 3.929 (H) P2/242 Zone 2 -13.55 0 0 0 1 [Si02l = 11.98 - 1.704 (H) 128 p o o o . « O 3 TJ LU CO Tt 9 _ ^ £ to o © CO o c o o C D o a g c 0) CD I S o ® CO OQ si CM -z: eo g_ 2 S 3 O BO LL 1 E I I i CO (A | \ J T3 P Q> O X C Op o o d rvi o CD 1 1 ' 1 p 9 CO 1 f t 1 1 t t 1 1 o S E 1 i a . o CO O 03 X + < x«< WJ i < ** +x I I t I I I w x ; + I I I 1 I l-H < I X I I 4* Xt 4 — U • — L _ -II 73 „ , • co (jj o I re o CD d o o co" 0 01 T o a o (l/Biu) uoijejjueouoo s;n|os (l/Biu) uoi;ej)uaouoo B»n|°s 138 3.2.26 along with curves derived from regression of chemical concentration vs total head and hydraulic conductivity for .each group individually. In those graphs, the most significant relationship derived from the individual regression is plotted as a solid line. In most cases, that relationship is linear, although in some cases it is a quadratic and in other cases there is no significant relationship. The relationships derived from the overall regression analysis as presented in Tables 3.2.16-20 are given as dashed lines. The colour of those dashed lines is black so that they will be easily distinguished from the specific fit given as a solid line. In cases where there was no significant specific fit, the dashed line is plotted in the colour that corresponds to the chemical that it represents. In the overall regression analyses, second order models in total head were attempted but in all cases, the second order terms were not significant. Thus all overall regression fits of concentration vs total head are linear. In conducting the overall regression analysis, it was found that the zone 2 (hillside) groups had to be separated from the zones 1,3 and 4 groups to obtain a good fit to the zone 2 data. Land slopes are generally lower among the hillside sites than adjacent to streams and consequently the relationships between chemical concentration and total head have lower slopes. In most cases, the concentration vs head relationship has a negative slope, but when the zone 2 data were lumped with the zones 1,3 and 4 data, the within group variability was hidden by the between group variability, resulting in positive slopes. This problem was corrected by separating the zone 2 data. By lumping together the data from zone 2 and zones 1, 3 and 4 a relationship between chemical concentration and total head is always indicated, whether a significant relationship is indicated by the specific fit or not In some cases (eg PI at 242 Creek, Figure 3.2.25) the scatter plots suggest practical relationships that are not significant in the specific regression analysis because there is high variability among the data points, but the equation indicated by the overall regression is clearly a good fit to the data. (A practical relationship is one that appears to exist by inspection of a scatter plot, but is not statistically significant.) In other cases (eg calcium on P2 at 242 Creek, Figure 3.2.26) the slope of the specific fit is insignificant 1 3 9 because there is no significant or practical relationship between concentration and head, and the overall regression equation also has a slope that is close to zero. The overall regression equations from zone 2 and zones 1, 3 and 4 fit the data reasonably well. R2 values range anywhere from 24.5% to 89.5% with most in the 60-70% range. Many of the R2's are not particularly high because the relationship between concentration and H is not significant for some standpipes. In some cases this can be explained in terms of the head range. For example, at the zone 2 site in Edelweiss Creek the soil is relatively deep and the hydraulic conductivity high so that groundwater is only present for a brief period around peak flow. The regression is therefore carried by the standpipes for which concentration-head relationship is significant, overcoming the within group variability that makes some of the specific regression fits insignificant. On the other hand, lumping of all standpipes within zone 2 and within zones 1, 3 & 4 resulted in linear regressions since most individual standpipes demonstrated a linear relationship between concentration and head. Because second order terms were insignificant in the regression, a linear fit was forced on all standpipes even though in some cases, the individual jfit was non-linear. The regressions using land slope and those using hydraulic conductivity produce similar results for the existing standpipes. However, the land slopes at the groundwater sites do not represent the full range of land slopes in the watersheds, whereas it is felt that the standpipes more closely represent the range of hydraulic conductivities that are likely to occur in the study area. It was later found that extrapolation of the relationships to higher land slopes resulted in unreasonably high concentrations. Thus, the relationships derived using hydraulic conductivity and head are the more logical ones to use for predictive purposes because extrapolation of the relationships is expected to be minimal. This is reasonable since the highest flow conditions that were sampled (May 28-29, 1990) were closest to the highest on record for UPC. The difficulty in using them is that basin wide hydraulic conductivities must be estimated using indirect methods, whereas land slopes can be measured easily from contour maps. \ • . • J 140 3.2.6b: Discussion of groundwater chemistry model. In most cases, chemical concentrations decrease with increasing head. This trend could be explained by dilution; higher heads occur as a result of inputs of new water from rainfall or snowmelt, which has lower concentrations of cations, bicarbonate and silica than the average in groundwater. On the other hand, overall average sulphate concentrations in precipitation are close to the overall concentration in groundwater. This is because sulphate is of atmospheric origin and is concentrated in soils by evapotranspiration (Vitousek, 1977), whereas the other chemicals are produced within the soil either by chemical weathering (cations, silica) (Gibbs^ , 1970), or by biological activity (bicarbonate). Sulphate tends to decrease with increasing head at zone 2 sites, and to show the opposite trend at streamside sites and in bedrock. This can not be explained by the relative concentrations of precipitation and groundwater at those sites. At the hillside sites, sulphate is probably concentrated in groundwater during the growing season by evapotranspiration, and then leached down to streamside sites and to bedrock during spring snowmelt. At the zone 2 (P2&3) site at 240 Creek, there is a relatively large range in head due to deeper soils and moderate hydraulic conductivity, with well defined trends in chemical concentration. This soil profile represents semi-permanent groundwater storage that is capable of generating extended base flows. Piezometer P4 representing bedrock has similarly well denned concentration trends with much higher concentrations due to its greater depth and lower hydraulic conductivity. At the 241 Creek site, soils are shallow with relatively high hydraulic conductivity, with the standpipes sampling water that is closer to the surface. Thus the range in head is low and the concentration trends are also well defined, with lower concentrations than f the 240 Creek site. This profile is saturated over a short period of time around spring snowmelt with the water table quickly rising to the surface and draining rapidly thereafter. This site represents a more transient type of storage. These sites can probably be thought of as similar to subsurface flow partitions as described by Maule and Stein (1990) with P4 representing old 141 subsurface, P2&3 at 240 Creek representing old and new subsurface water and the 241 Creek site representing recent subsurface water] The average concentration of each component reflects the relative residence time in the soil, and the trend with head reflects the relative mixing rate with inputs of new water. The zone 2 site at Dennis Creek also appears to be representative of old and new subsurface water. It has hydraulic conductivity comparable to that at 240 creek but with deeper soil. The range of head is greater than at 240 Creek and trends in chemical concentration are similarly well defined for all chemicals except for calcium. Average concentrations of bicarbonate, silica and sodium are slightly lower than at 240 Creek, probably reflecting lower rates of organic decomposition and mineral weathering due to lower temperatures (a result of higher elevation and denser, lower albedo forest cover). However, concentrations of calcium and sulphate are quite different, possibly a reflection of differences between the Lodgepole pine and Spruce-Fir forests in their use of those nutrients. It can be inferred from meteorological data that evaporation rates are lower at Dennis Creek than at 240 Creek. This may be a factor in the head range and lower chemical concentrations at Dennis Creek (lower evaporation implies less concentration of chemicals), but it is not clear if this is attributable to forest cover or to other factors such as elevation and aspect. Higher concentrations of cations and bicarbonate at Edelweiss Creek than at other sites are consistent with incomplete recovery following logging disturbance that was demonstrated at Hubbard Brook. 3.2.7: Zone 1 Groundwater Chemistry as a Combination of Zones 2 and 4. Visual comparison of the fits produced by the above analysis with the scattered data points demonstrates that in the case of P5/240 Creek, the fit does not seem appropriate. In fact, when total head is at or above 1.5 m, a state of groundwater discharge exists, and below that value, recharge occurs. This is because the soil depth is 1.5 m and head is calculated using the bedrock-soil interface as datum. It is clear that there are different relationships between chemical concentration and total head for these different conditions. When discharge occurs, it is assumed that discharge at that site occurs from both bedrock and soil. This assumtion explains the 142 concentration-head relationship during discharge at P5; some of the water at the completion interval will be bedrock discharge, with a component of soil (ie, zone 2) water that increases as the water table rises. Thus, zone 1 water is actually a combination of water from the hillside and bedrock. This is an essential element in water chemistry modeling because groundwater contributions to streamflow involve ratios between lower soil and bedrock outflows that vary between low and high flow. The variation between these ratios can be determined from a comparison between zone 1 chemistry and its component zone 2 and bedrock contributions. Using silica as a tracer, the relative proportions of zone 2 and bedrock contributions at low and high flow were determined for each zone 1 groundwater site. The method used to make that determination is as follows: 1. The head was determined at some reference low flow that varied from 0.001 to 0.005 m3/s. 2. Head was then determined at the highest streamflow that occurred during the study. This quantity was taken as the maximum of a second order polynomial that was fitted to a graph of groundwater head vs day of the year. 3. The silica concentrations in zones 1, 2 and 4 were derived from the equations given in Table 3.2.20b, using the heads as determined above. 4. The proportion of bedrock contribution to zone 1 was determined by weighting the bedrock and zone 2 silica contributions (Table 3.2.21). The ratio of the hydraulic conductivity of zone 2 to that of bedrock is also given as a reference with the fractional ratio given in brackets. The data show that zone 1 water chemistry can be represented as a combination of groundwater from zone 2 and bedrock in which the relative proportion of these two components changes according to flow. The proportion of bedrock groundwater is greater at high flow than at low flow at the 241 and Dennis Creek sites, and the reverse is true at the 240 and Edelweiss Creek sites. The hydraulic conductivity ratios are in the same range as proportions of bedrock outflow in zone 1 that were calculated from the silica concentrations at 240, Dennis and Edelweiss Creeks but not at the 241 Creek site. This shows that the zone 1 site at 241 Creek is not 143 Table 3.2.21: Zone 1,2 and 4 Head and Silica Concentrations at Low and High Flow Groundwater Hydraulic Conductivity Ratios Proportions of Bedrock Groundwater in Zone 1 Water Relative to Sirica Tracer Creek Groundwater zone H(m)at low Q H (m) at highQ Si0 2 lowQ Si0 2 highQ (fractional ratio) 240 zone 2 0.50 1.1 11.09 9.73 15:1 (0.067) zone 1 1.00 1.7 11.80 9.78 bedrock 0.70 1.1 13.35 11.71 bedrock proportion in zone 1 0.31 0.03 241 zone 2 0.32 0.67 9.35 8.67 48:1 (0.021) zone 1 0.20 0.62 10.67 10.38 bedrock proportion in zone 1 0.33 0.56 Dennis zone 2 0.97 1.45 10.33 9.51 16:1 (0.063) zone 1 0.45 1.03 7.84 10.12. bedrock proportion in zone 1 0.00 0.28 Edelweiss zone 2 0.12 0.70 11.93 10.28 7:1 (0.143) zone 1 0.32 0.65 12.10 8.38 bedrock proportion in zone 1 0.12 0.00 really representative of the true proportion of bedrock groundwater in total groundwater outflow, whereas the other zone 1 sites are representative. At 240 Creek, the proportion of bedrock component in zone 1 groundwater at high flow is as low as 3%, rising to more than 30% at low flow. This implies essentially fixed contributions from bedrock to total groundwater outflow at 240 and Edelweiss Creeks, whereas at Dennis Creek, the proportion of bedrock in zone 1 water increases with flow rate. This demonstrates that groundwater chemistry can be modeled as a combination of zone 2 and bedrock outflows, with a minor contribution from seepage. 3.2.8: Interflow. When it was possible, samples were collected from subsurface seepage as it discharged at the soil surface adjacent to stream channels at 240 Creek and Edelweiss Creek;. These samples were assumed to represent interflow. In all cases the seepage came from what appeared to be soil pipes. This component was highly transient; that is, it only appeared under high flow conditions. Specific sites were identified where these samples could be collected sporadically. The sites were identified as slow and fast seeps depending on the visually observed outflow rate . 144 at the time of the sampling. Chemical concentrations were averaged for the slow and fast seeps (Table 3.2.22). These averaged interflow concentrations were later used in modelling water chemistry by assuming that at low rates of interflow (<0.1m3/s) the interflow is dominated by slow seepage, but at higher interflow outflow rates (up to 0.3+ m /^s), there is increasing dominance by the fast seepage. Table 3.2.22: Average Chemical Concentrations in Slow and Fast Interflow Seepage rate Sample size N03 S04 HC03 Ca Na Si02 Fast 240 5 0.02 0.82 5.26 1.21 1.11 1.64 Slow 240 6 0.02 0.89 10.10 1.76 1.18 2.05 Fast Edel/Dennis 4 0.02 2.80 4.25 1.52 0.94 2.26 Slow Edel/Dennis 5 0.00 0.70 11.44 2.30 1.26 8.89 .145 3.3: STREAMFLOW 3.3.1: Concentration vs Stream Discharge. The water chemistry of 240, 241 and 242 (Dennis) Creeks has already been described in great detail by Golding & Hudson (1991). Highly significant relationships were discovered between the concentrations of bicarbonate, calcium, magnesium, potassium, sodium and silica vs the logarithm of discharge. Similar analyses were performed on water chemistry and measured streamflow from Edelweiss Creek. As discussed in the introduction to this section, bicarbonate, sulphate, calcium, sodium, and silica were selected for the analysis. The concentrations of chemicals in the samples that were collected were related to the instantaneous discharge at the time when the samples were collected, representing a full range of conditions from low flow below the measurable limit in September 1987 to the highest flows that occurred during the sampling period in May 1990. The resulting water chemistry relationships for all four creeks are summarized in Table 3.3.1. These relationships can be thought of as an empirical water chemistry model for streamflow because they have the ability to predict the concentrations of chemicals at the sampling point from the measured stream discharge. The semi-logarithmic relationship of concentration vs. discharge was also used elsewhere by Thorne et al. (1988) in New Hampshire and by Feller and Kimmins (1979) at the U.B.C. research forest near Haney, B.C. The R2 values reported in Table 3.1.1 are higher than those reported in those studies. Thorne et al: (1988) suggest that a high correlation between the concentration of cations vs stream discharge indicates that water chemistry is controlled primarily by mineral weathering. The high correlations at UPC suggests that water chemistry is more strongly influenced by weathering than at those other locations, and that over-winter mineralization within the forest floor is a relatively minor factor. As an example, a graph of bicarbonate vs stream discharge is given in Figure 3.3.1. There is an obvious similarity between 240 and 241 Creeks in terms of HCO3 chemistry. To determine the nature of that similarity, regression comparison analysis (Kozak, 1970) was used. This o - o o o o o o o o o o o o o o 0 0 * 0 0 0 < 2 0 0 0 . 0 0 0 0 0 0 0 0 0 . 0 0 0 o o" o o o 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 S? 6? 6? S§ so in r>. 0\ o C V C N L O co CN r H O / C O C O 00 < 2 rH CO CO O LO s d •>* CO Os s d CN CO N N Sf? 5? cV? H q . « if) o; i n ps CN ,sd 10 "tf Is 00 so LO to O » N so K <# *tf -ft LO SO CO OS Os rH rH «N OS CN rH 00 O O O O O < 2 sO CO OS O SO -fl CO 00 b b c> LO Cv CO 00 o CN rH 8 C O Z E b ^ rH rH b o ' - 1 s i s o to > I ? s g s o •XJ c u s o u u c O tv. —1 CN 00 SO a g to _2 -2 os Os CO CN 0 b • OS II co • - o 00 „ CM 00 O N in CN b 1 ll ~' -s Q T3 cu N j>s . * < o 2 00 3 ° -2 in co CN CN b b • SO SO CN tv, CN . CO o ^ CN LO *# 1 " S H 0 11 -=,0 11 11 ~1 cO 8 ' to a S to s? 8 « so •>* 5.CO CO II II ~> on X ^ c i£ 00 -2 rn a 33to *>.<=>£ Q • 1 LO 1 00 LO sO JTj s d LO 00 I rH O || ll j i L - j r«'—' Q o o o o .. o o o o O O O O O d d d d d CN O O O O sO O O O O O O O O O d d d d d O O O O O O O O O O o p p.p . 0 d d d d d 0 0 0 0 0 O O O O O O O O O O d d d d d a 1 &? &? Cv CO Cv. LO ON rH Csi CO C S N rH Os 00 OS IS i » a? s? co i o C N s q CO. Os cO Sv 6? i*? 5? O OS sO OS rH CO CN 3 5? 6v &S S? S? O O Cv. 10 Os CN C> **• (S CN so LO 00 so LO Os CN CO rH Cv. OS 3 If) SO • * S ifl N N . CN O 00 00 LO LO CN CNCv.Cs LO 00 sO Cv LO rH CN rH Os b b b CO O sO Os sO ** CN Cv. 00 CO LO o CO 00 rH rH d dd d co CN LO <3s rH rH. rH L . b o r H 12 s 8 s cu u s 1 a w w u a 60 to-2 -. O Cv. —< Os CO £ ^ •rH II II "co •31 P s v^ — O' aSts 60 w,2 0 o H SO CO |_j CN ~•* • Os CO CN b b 1 1 1 o LO O ^ 00 LO rH O II . II Ji^—J " « Q U Z cn 60 60 O O • * CO 00 CO o 00 b co" + 1 CN LO II I—I " CO ^ 8 X 60 %>£ 3 ° CN R S ?i 00 CO CN o b ' CO ^> ^ C3< sO rH O II 1 J i ^ " J S Q U 2 cR. Mi2 to M O - ffi rS -2 OS ^ ^ CO CN CN 0 J b b 1 LO 1 - 1 OS ^ ^ . 1 1 o CO 1 -75 ^ ° 11 r - 0 11 11 TJ cn X U Z cn ——I W—i b__ L _ | |__ 01 01 01 3 CN _< 01 Ol CN c C 01 Q -21 a to SO I If jf o + CN OO CO" II CO .60 b o rt Cv CN 1 b o 1 II J ^ - q " • 5 Q U 2 cH . ce 01 • „ 01 "0 W 147 method revealed that for bicarbonate, calcium and silica, the regression lines of concentration vs l°g(Q) f° r 240 and 241 Creeks are colinear. That is, there is no significant difference between either the slope or the intercept. For sodium, those regression lines are parallel, meaning that the slopes are the same, but the intercepts are different For sulphate the analysis is not valid since the regression of sulphate concentration vs log(Q) on 241 Creek is not significant at 90% confidence. Because of this demonstrated similarity, it was decided that 241 Creek streamflow quantity or quality data would not be used in further analysis to avoid redundancy. Because the watersheds have different drainage areas, the water chemistry results are comparable only on the basis of specific discharge. This is particularly true of Edelweiss Creek; with an area of 46 ha, it is an order of magnitude smaller than the other drainages. Therefore, discharge measurements on each of 240, Dennis and Edelweiss Creeks were divided by the drainage area of each creek and expressed as specific discharge in units of litres/ second/hectare. The regressions of concentration vs discharge were rerun using the log of the specific discharges, and are summarized in Table 3.3.1. Graphs of concentration vs specific discharge for each chemical on 240, Dennis and Edelweiss Creeks are given in Figures 3.3.2 to 3.3.6. 3.3.2: Comparison of Concentrations Between Creeks and to Other Studies While it has been noted by Phillips and Stewart (1990) and by Johnson and Reynolds (1977) that cation and bicarbonate concentrations tend to increase with stream size, this is not the case when comparing Edelweiss Creek to Dennis Creek. Vitousek (1977) states that steady state ecosystems (of which Dennis Creek is an example) lose higher concentrations of essential plant nutrients (e.g. sulphate, calcium) than rapidly growing intermediate-aged successional ecosystems (such as the regenerating clear-cut occupying the lower 20% of Edelweiss Creek). Clearly, this does not hold true at UPC, suggesting that other factors are more important than successional stage. Vitousek (1977) and Phillips and Stewart (1990) suggest elevation as a possible factor controlling stream chemistry. Edelweiss Creek has an elevation range intermediate between 240 and Dennis Creeks. Concentrations on Dennis Creek are consistently lower than either 240 or Edelweiss Creek. Concentrations on Edelweiss Creek are 148 § (I/BLU) uo;)Ej;uaauoQ a;ei|d|ns o o o o o o cj c o co CD u c ro o J= O o i Q "fo ~ ji, o M o » a i 0 CD •• Q-* CO CO CO CD 1 HI '— 6 CM II o o o CM o co d O CO 'c c 0) O c O o co ci S3 o cu S CD T 3 149 (l/Bui) U O I ; B J ; U S O U O O uunioieo > c o c CD O CD C O) o >-o j2 CD O CO — C Q 5.a »_ ^ CO rj «2 CD CQ tX . to CO CO • CO CD 1_ 3 _ a ^ j= c o o a co co *5> f i ? CM CM (n • * 09 o CM 1^ CM II CM CO .1 o co O O I 0 ~ 1 .* — a> \j cu CO co li co" O 0 X in co CD 1 CU •o Ui HI (l/Bui) uoiiBJjuaouoo ateuoqjeoig 150 o o o o o o o o ^ o i d o o c b r r o i d (l/Buu) uouejuiaouoQ eoi|!S o r l o o o o co oi ci (l/Biu) uoi je j juaouoo wmpos 151 intermediate between 240 and Dennis Creeks for calcium, bicarbonate and silica, and higher than 240 Creek for sodium and sulphate. There appear to be inverse relationships between average concentrations of bicarbonate, calcium, sodium and silica, and mean elevation at UPC (Figure 3.3.7). Concentrations in the low flow range (i.e., base flow) in all three creeks are similar to concentrations reported by Johnson and Reynolds (1977) in baseflow for watersheds in Vermont and New Hampshire with schist bedrock. In that study, differences in chemical concentrations in baseflow were attributed to differences in bedrock. The chemical differences between the three creeks in the current study are not so great as to be attributable to differences in bedrock type. However, since the soils in Upper Penticton Creek are derived from glacial deposits, the precise origin of those deposits are not known, and there may be differences in the origin of sediments and in the depositional processes from one area to another. The chemical differences are likely due to soil parent material and to .forest cover. This is supported by results of groundwater chemistry analysis discussed in section 3.2. Baron (1983) attributes differences in water chemistry among lithologically similar drainages to differences in vegetation cover, with nutrient export levels being higher in drainages with lesser aerial vegetation cover. Using average crown closure as an indicator of vegetation cover, there also appears to be inverse relationships between vegetation cover and average concentrations of bicarbonate, calcium, sodium and silica at UPC (Figure 3.3.8). 3.3.3: Mass Fluxes of Chemical Species. From the point of view of domestic and industrial water quality, chemical concentration is the most important quantity. However, the mass flux of a given chemical is usually a more useful quantity if one wishes to assess chemical weathering rates or site nutrient loss. For this reason, relationships were also assessed that relate instantaneous mass flux to discharge for the four catchments. The mass flux was calculated from the concentration and the discharge as follows: MFx(g/hr/ha): 1 5 3 3600xCx(mg/ l )xQ(m3/s) drainage-area(ha) where MFx and Cx are the mass flux and the concentration of chemical x. Mass flux was calculated in units of g/hr/ha for computational convenience. The equations for the four creeks that relate mass flux to discharge are given in Table 3.3.2. To compare the mass flux of the five chemicals from,the four drainages, the equations in table 3.3.2 were used to calculate the daily chemical fluxes over the three year modeling period from September 1987 to August 1990. These fluxes were totaled and converted to a mean annual flux from each catchment. These values are summarized in Table 3.3.3. While continuous streamflow records were available for 240, 241 and 242 Creeks, only point measurements were taken on Edelweiss Creek at the time when samples were collected. To assess daily and annual chemical flux rates, the simulated streamflows developed later (see section 4.3) were used in lieu of a continuous observed record. Note that Edelweiss Creek was modeled as a whole, and in upper (mature Forest) and lower (clear-cut) sections. This allowed the chemical budget to be calculated separately for the two distinct forest cover types. A few samples of .streamflow were collected from the creek at the outlet of the forested part of the drainage, but not enough to establish relationships between estimated flow and concentration. Instead, discriminant analysis was used to test for similarity between chemical concentrations from the forested part of Edelweiss Creek and other sites. While sulphate and silica concentrations were found not to be significantly different from those at the weir on Edelweiss Creek, bicarbonate, calcium and sodium were found to be significantly different from concentrations at the weir, and similar to those at Dennis Creek. Thus, it was possible to estimate concentrations and mass flux rates of those three chemicals from the forested section of Edelweiss Creek using the equations that relate chemical concentration to per-hectare discharge on Dennis Creek (Table 3.3.1). The daily mass flux was calculated as the product of the daily discharge and the predicted concentration. The intercept of the equation to predict sulphate Table 3.3.2: Mass flux-discharge relationships, CREEK mass flux Fv(g/hr/ha) vs Q s.e. R2 P 240 Creek Fsa=i4.0(Q)-8.96(Q2) FHCO, = 0.41 + 57.9(Q) - 39.0(Q2) + 16.7(Q3) F C a = 20.2(Q) - 15.36(Q2) + 7.66(Q3) F N a = 0.05 + 5.79(Q) - 3.02(Q2) + 1.68(03) F S i 0 , = 46.6(Q)-12.66(Q2) 0.4500 0.6575 0.2975 0.0927 0.7331 94.3% 99.3% 98.8% 99.0% 99.1% 0.000 0.000 0.000 0.000 0.000 241 Creek Fsa=8.48(Q) FHCO, = 0.68 + 60.0(Q) - 24.29(Q2) FCa=20.9(Q)-8.30(Q2) F N a= 0.09 + 6.09(Q) - 1.30(Q2) F S iO, = 53.7(Q)-18.33(Q2) 0.7117 1.208 0.3546 0.1422 0.8076 88.7% 98.2% 98.7% 98.4% 99.1% 0.000 0.000 0.000 0.000 0.000 Dennis Creek Fsa=6.26(Q) + 22.54(Q3) FHCO, = 39.58(Q) - 45.09(Q2) + 27.36(Q3) Feu = 0.13 + 10.07(Q) ' F N a = 0.06 + 4.48(Q) F S i a = 5l.44(Q)-49.78(Q3) 0.7131 0.6408 0. 3587 0.1244 1.852 97.2% 98.2% • 97.3% 98.2%. 93.3% 0.000 0.000 0.000 0.000 0.000 Edelweiss Cr F S 0 4= 230 (Q) +5515 (Q3) FHCO, = 644(Q) - 9282(Q2) + 48448(Q3) F C a = 174(Q) - 1163(Q2) + 6274(Q3) -F N a = 0.23 + 61.4(Q) F S i a = 510 (Q)-21485 (Q3) 1.250 1.644 0.2692 0.4738 2.429 97.0% . 91.7% 99.2% 90:8% 89.2% 0.000 0.000 0.000 0.000 0.000 Table 3.3.3: Average annual mass flux September 1987-August 1990 (Sep 88 - Aug 90 CREEK N0 3 so 4 HC03 Ca Na Si0 2 240 Creek 0.01 5.55 27.60 8.24 2.98 21.28 241 Creek 0.02 3.87 29.90 8.36 3.39 21.91 Dennis Creek 0.00 3.69 12.72 5.31 2.38 18.87 Edelweiss Creek 0.07 16.38 20.18 10.03 6.84 26.00 Edel-forest 0.00 17.63 15.60 6.33 2.68 22.85 Edel-clr.cut 0.34 12.60 33.98 21.20 19.40 35.49 Table 3.3.4: Averap e Annual Net input(+) or output(-) in kg/ha/yr CREEK N0 3 so 4 HCO3 Ca Na Si0 2 240 Creek 2.63 -0.94 -24.43 -5.94 . -1.08 -17.79 241 Creek 2.62 0.74 -26.73 -6.06 -1.49 -18.42 Dennis Creek 2.64 0.92 -9.55 -3.01 -0.48 -15.38 Edelweiss Creek 2.57 -11.77 : -17.01 -7.73 r4.94 -22.54 Edel-forest 2.64 -13.02 -12.43 -4.03 -0.78 -19.39 Edel-clr.cut 2.30 -7.99 -30.81 -18.90 -17.5 -32.03 concentration was adjusted to account for the differences between sulphate concentration on upper Edelweiss Creek and Dennis Creek. Those estimated three year average mass flux rates are given in Table 3.3.3, and the resultant mass flux rates from the clear-cut portion are also given. The fluxes from the clear-cut were calculated by subtracting the total flux from the forest from the total flux from the whole catchment, and dividing it by the area of the clear-cut within the drainage area (10.7 ha.). The results show greatly elevated fluxes from the clear-cut unit. Table 3;3.4 gives the net mass flux into (positive value) or out of (negative value) each unit in kg/ha/yr. It can be seen that nitrate is accumulated in the watersheds, presumably in the biomass. Sulphate is more qr less in balance except in Edelweiss Creek where there is a net loss. Bicarbonate, calcium and silica are produced within the watersheds, bicarbonate by breakdown of organic matter and calcium and silica by weathering. S°dium is also produced by weathering in the watersheds but at much lower levels, such that in Dennis Creek and upper Edelweiss Creek the effect is masked by the input of sodium from rainfall 3.3.4: Longitudinal Profiles Periodically, samples were collected at upstream sites oh 240, Dennis and Edelweiss Creeks to establish longitudinal stream water chemistry profiles. The sampling sites are indicated on Figures 3.23-4, and included bothmainstem sites and selected tributaries. At sites where tributaries were sampled, the mainstem was also sampled immediately upstream of the tributary junction. This long profile sampling was done on three occasions at 240 and Dennis Creeks, and twice at Edelweiss Creek for the purpose of relating upstream water chemistry regimes to that at the weir under a range of flow conditions from low to medium flow. This profile sampling has already been alluded to ih section 3.3.3 for Edelweiss Creek. A comparison of the profiles for each creek sheds light on the role of soil and forest cover type in generating streamflow chemistry. For this purpose, two parameters were used, total cations and sulphate concentrations in mg/1 that are plotted against the distance upstream of the weir on each creek (Figures 3.3.9-11). • o o o o p (l/6iu) uoflaauaouoo ejBMdfns (l/Bui) uoftB4U90UO3 UOJIBO IBIOI 0 i/> O LO O if> O iO 01 t"~ O W (N CN 1-(|/6UJ) uoijcjiuaDuoo ajGqdins (l/6ui) uoijejjuaDuoo uoijeo |e»oj. » q "o o p (t.,fmi) uoiiej|ua3U03 apifil|iis (t/Bui) uouenuaouoo uoiieo lejox 157 At 240 Creek, the profile indicates a general increase in both cation and sulphate concentration towards the weir. This is consistent with the findings of Philips and Stewart (1990), and is also consistent with the results of the available cation analysis (Figure 3.2.8). Because the soil and forest cover characteristics are relatively uniform, these trends can be explained in terms of increasing soil cover towards the outlet of the watershed, and increased weathering rates of cations and mineralization of sulphate with lower elevation. The influence of the tributaries can also be seen; for example, the mainstem and a tributary were sampled at a point 650 m above the weir. Sulphate concentrations were lower and cation concentrations higher on the tributary than on the mainstem, causing a sharp decrease in sulphate and rise in cations at the next lower sampling site (Figure 3.3.9) on all three sampling dates. The profiles for Dennis Creek show the effect of the different soil and forest cover types; at low flow, there is a general decrease in cation concentrations and an increase in sulphate towards the weir, the opposite of the trend noted for 240 Creek. The upper sampling points are apparently influenced by the Type 11 soil and predominantly pine forest in the upper watershed (Figure 3.2.4). However, at higher flow, the increasing trend is observed for cations. The influence of the tributaries can be seen clearly (Figure 3.3.10). The tributaries enter the mainstem low down in the watershed and are dominated by the type 12 soil/spruce-fir forest cover, and modify the water chemistry of the mainstem significantly. Thus, at Dennis and 240 Creeks, the chemistry of the mainstem channels undergoes gradual changes from the headwaters towards the weir due to the distribution of soil and forest cover types within the drainage area defined by each sampling point, with more abrupt changes occurring due to tributary inflows. Those tributaries can be seen as having more homogeneous characteristics of soil and forest cover types within their drainage areas compared to the watershed area above the weir. Edelweiss Creek is actually a first order tributary of Dennis Creek at a point downstream of the weir. It is logical to assume that the harvesting of the now regenerating clear-cut altered its water chemistry, with an impact similar to that reported at other study areas! The question is, has it recovered from that impact or not? Without pre-treatment data, this question cannot be 158 answered definitively, but only inferred from the available evidence. A comparison of the mainstem chemistry profile on Edelweiss Creek (Figure 3.3.11) with those of Dennis and 240 Creeks suggests that it has not. For instance, the plunge in sulphate concentrations as the creek enters the clear-cut compared with the trends in sulphate concentration on the other creeks is consistent with the drop in sulphate concentrations after logging that has been documented at other locations such as Hubbard Brook. In interpreting these profiles, it is important to note that each successively lower sampling point integrates the inputs of the catchment above that points At Dennis and 240 Creeks, large changes in water chemistry between sampling points were due to inputs from first order tributaries, whereas on Edelweiss Creek, the tributaries that were sampled were ephemeral snowmelt runoff channels. Thus, large changes in the chemistry profile are most likely due to differences in the terrestrial ecosystem. Other lines of evidence to support this conclusion are summarized elsewhere. 3.3.5: Comparison of results with other sites Table 3.3.5 gives the distribution of anion and cation groups according to their percent contribution to charge balance. The average concentrations of all ions were expressed in meq/1. The acidic anions consist of sulphate, chloride, nitrate and phosphate. The organic anions were" assumed to equal the charge balance inequality in meq/1 (Crorian et a!>T978). The distribution of cations is similar to that reported by Johnson and Reynolds (1977) for metambrphic bedrock; The distribution of anions show some differences between sites. Bicarbonate is clearly the dominant anion at 240 and Dennis Creeks, whereas at Edelweiss Creek, acidic anions are slightly dominant. There seems to be a relationship between the dominant anion and cation groups; that is, as dominance of acidic anions increases, dominance of calcium + magnesium decreases. Table 3.3.5: Distribution of Anions and Cations in Streamflow % of negative charge balance due to: % of +ve charge balance due to: Creek organic anions acidic anions bicarbonate calcium + . magnesium sodium + potassium 240 Creek 29.4 18.6 51.9 76.3 23.7 Dennis 24.9 28.6 46.5 70.9 29.1 Edelweiss 14.3 45.5 40.2 68.4 31.6 159 Because there is no significant difference in the chemistry of precipitation from site to site, the differences could be due to differences in forest cover and/or the forest floor (Cronan et al. 1978). The dominance of acidic anions at Edelweiss Creek may be indicative of post-logging disturbance, assuming that it was chemically similar to Dennis Creek prior to logging. It has been noted that Dennis Creek is less well buffered than 240 Creek and is thus more sensitive to logging related changes in stream chemistry. Annual chemical budget information was compiled for 25 research watersheds from temperate, mid-latitude locations in North America, Europe and Australia, including this study (Table 3.3.6). The annual chemical budgets were either taken from the quoted source or calculated. Each source contained information regarding bedrock geology, and many noted the degree of acidification from atmospheric pollution where applicable. To compare those input-output budgets on the basis of geology and level of acidification, the effect of annual precipitation had to be eliminated. Chemical inputs and outputs are directly related to precipitation; that is, the higher the precipitation, the greater the input of chemicals, and the greater the streamflow. Since there is a positive relationship between streamflow and mass flux of chemicals, higher precipitation will also result in greater output of chemicals. Using UPC data, it was found that there was also a relationship between the annual net chemical fluxes and the total annual precipitation for the three water years from September 1987 to August 1990. Therefore, the budget for each chemical in kg/ha/yr was divided by the mean annual precipitation in metres for the period studied, in an effort to account for the effect of precipitation volume on nutrient budgets. The degree of acidification reported for each watershed was assigned a number from 0 to 4 according to the following: 0 not acidified, or well buffered from terrestrial sources 1 potential exists to be affected by acidification, or buffered by atmospheric dust sources 2 moderate acidification, moderately buffered by surficial deposits 3 acidified by atmospheric pollution, and poorly buffered 4 heavily acidified Cluster analysis was used to examine the 25 watersheds for similarities and differences among their standardized nutrient budgets (Figure 3.3.12). Two separate analyses were performed: for ON Tj 5 3 S cu N in ts. ON g > TJ .s ON ts ts. ON t>0 O D s CJN CU C 0 cn 1* 01 TJ g ts. , oo 0> •a g o & t 55 ON QJ ts Cs ON cu -d | Oi ON 60 •5 CA ON 60) . r cn « . CU It ON cu 601 m t s ON rH CU Tj ON rH C s N W B cu cu T ? g -a TJ "•5 > I CB > •6 o OS CU 00 X 160 c. o •J3 •§ .SH 'u ^H a. <*H o cu CB 01 >, S T3 cy a oo 5 11 S3 -3 T3 •S .2 + u s ^ CB t * •5 it 1 2 o CB w T3 tj J CB « ^ ml S § CM H-> 60 .y - I c * OJ II J3 T3 ' 2'< Figure 3.3.12: Cluster Analysis of Research Basins using Cations, Cations + Anions yellow = not acidified, mildly acidified; green = potentially acidified, acidified. Similarity 55 70 85 —I 100 Clustering of Basins Using Cations Only 1 22 2 24 3 16 7 9 4 25 12 5 8 17 19 23 18 2D 6 10 14 21 11 13 15 Not Mldly Aadified or Potential Aadified Aa'dified for Aadification Gneissic or Granitic Sedimentary Similarity Clustering of Basins Using all Parameters 34 -4 56 H 78 100 1 22 2 3 16 4 Not Aadified Gneissic or Granitic Potential 9 12 7 Buffered 18 8 17 19 20 6 14 15 Aadified Sedimentary 162 some of the basins, only cation budgets were reported, so one analysis was run using only cations (Ca, Mg, K and Na), and a second analysis was run using cations plus Cl, N and S (chloride, nitrogen and sulphur). As expected, the watersheds with sedimentary rock are different from those with gneissic or granitic bedrock. Among the basins with gneissic or granitic bedrock, the analysis indicates a separation that can be attributed to the level of acidification. This is more clear when all nutrients are used in the analysis because sulphate is usually the acidifying agent, although not all watersheds can be used in the analysis. There are clear groupings among those basins which are not acidified (240 and Dennis Creeks, Loch Vale), those identified as being at risk from acidification (Haney, Jamieson), those that receive moderate inputs of acid precipitation but are also moderately buffered (Coweeta, Turkey Lakes) and those granitic catchments that are significantly impacted by acidification. The clear-cut section of Edelweiss Creek is grouped with Glendye between the non-acidified group and the "potential" group when all parameters are used, and is included with Rawson Lake and Coweeta when only cations are used for the grouping. This suggests a similarity in effect between a regenerating clear-cut site and mildly acidified sites. It was noted earlier that the literature suggests a parallel between these two effects. On the dendrograms (Figure 3.3.12) the indicated clusters within the gneissic/granitic bedrock type can be seen as showing a trend towards increasing acidification from left to right Since Edelweiss Creek is clearly in a state of partial recovery after logging, it is postulated that immediately after logging, it would have been clustered with more highly N acidified watersheds, and that in future, it will continue to recover until it is similar to catchments that are not impacted by acidification. 3.3.6: Discussion The cation budgets of 240, Dennis and Edelweiss Creeks are typical of gneissic/granitic terrain. The log-normal relationships between the concentrations of chemical species have also been used by other authors, but the high R2 values associated with those relationships indicate that water chemistry at UPC is more strongly controlled by mineral weathering than at other sites. Several lines of evidence indicate that differences in water chemistry between 240 and 163 Dennis Creeks are due to soil/forest cover type, although since soil development is linked to vegetation type, these two factors cannot be separated. That evidence includes the results of the soil survey, the analysis of groundwater chemistry and the interpretation of the longitudinal profiles. When Edelweiss Creek is included, the average concentration of calcium, sodium, silica and bicarbonate can be seen to be inversely related to both mean elevation and canopy closure. The cause of these relationships is likely related to the effectiveness of weathering (Johnson & Reynolds, 1977). That is, weathering rates are higher at lower elevations and where canopy closure is lower, resulting in higher average concentrations of weathering products. Bicarbonate at UPC is likely produced by a combination of weathering and soil respiration, although weathering of feldspar minerals is likely dominant in view of the high R2 between concentration and streamflow. However, it is logical to assume that soil respiration is governed by elevation and canopy closure just as weathering is. In spite of the lack of pre-logging data for Edelweiss Creek, evidence suggests that the regenerating clear-cut portion has not yet recovered from the impact that logging had on water chemistry. There are several lines of evidence to support this conclusion, although none are conclusive on their own. The soil analysis suggests similarity between Edelweiss Creek soils and Dennis Creek soils, but the groundwater and streamflow chemistry analyses demonstrate significant differences between those sites. Those differences are therefore attributable to the difference in forest cover. The longitudinal profiles showed much greater changes in stream chemistry on Edelweiss Creek than on 240 and Dennis Creek over similar distances. While McKnight & Bencala (1990) found significant changes in stream chemistry over short distances that were attributed to instream processes, the changes in sulphate concentrations in the profile were consistent with changes that other studies have shown to occur due to forest harvesting. The analysis of anion dominance (Table 3.3.5) show that sulphate is the dominant anion on Edelweiss Creek, as opposed to bicarbonate on 240 and Dennis Creeks. This is consistent with acidification impact, which the literature shows to have an effect similar to that of forest harvesting. The cluster analysis also corroborates this by grouping the clear-cut section of 164 Edelweiss Creek with mildly acidified catchments according to its nutrient budgets. Taken together, this evidence suggests that Edelweiss Creek has not yet recovered from the impact of forest harvesting on its water chemistry. 165 CHAPTER 4: HYDROGRAPH ANALYSIS AND MODELING RESULTS • s 4.1: Streamflow Events. There is a clear difference between the forest cover types in the rates of runoff production during spring snowmelt. The difference between 240 and Dennis Creeks has already been noted in chapter 1 "Introduction" in discussing the differences between the average annual i, hydrographs. It has been noted that Dennis Creek peaks later than 240 Creek due to snowmelt. The difference is due to a combination of factors. This concept is further illustrated by the graphs r of Figure 4.1, in which mean daily streamflow of Dennis Creek is plotted against that of 240 Creek for the high flow periods of 1989 and 1990. Those graphs show hysteresis loops in which 240 Creek peaks before Dennis Creek. In 1989 there was a double snowmelt peak. Both creeks rose in response to warm temperatures that reached a maximum on 08 May, and again on 02 June. The main peak on 240 Creek occurred in response to the first event, whereas the main peak on Dennis Creek was due to the second event. The maximum temperature associated with the first event was 14.6°C at 240 'Creek and 12.9°C at Dennis Creek. Both temperatures were measured under the forest canopy. The difference cannot be attributed to elevation, because in the clear-cut adjacent to Dennis Creek and at a slightly higher elevation, a maximum temperature of 16.5°C was recorded. In the second event, maximum temperature reached 14.0°C at Dennis Creek and 17.0°C at 240 Creek. However, while this produced the annual peak flow at Dennis Creek, much of the snow at 240 Creek had already melted and consequently the peak was not as high. In 1990, peak flows on both creeks occurred due to rain on snow events in late May and early June. One such peak occurred on 10 June due to 24.2 mm of rain that fell on 09-10 June. The peak flow from this event occurred simultaneously on both creeks. Subsequent to this, 51.2 mm of precipitation fell on 11-12 June. Temperatures were below 0°C both days and reached a maximum of 6.0°C at 240 Creek on 12 June. Thus, the precipitation was entirely snow at Dennis Creek and was probably a mixture of wet snow and rain (mostly snow) at 240 Creek. 686 k (s/cw) >|eejo smuorj 0 167 The subsequent hysteresis loop occurred due to 17 days of melt in which maximum temperatures at 240 Creek on clear days were an average of 3.3°C warmer than at Dennis Creek. The dry adiabatic lapse rate would indicate a temperature difference between the two sites of 1.7°C, thus the actual temperature difference and its impact on melt induced peak flows must be due primarily to the denser spruce-fir forest canopy at Dennis Creek compared to the lodgepole pine canopy at 240 Creek, as well as to the different aspects at Dennis and 240 Creeks. The difference in melt induced flows is much more dramatic when streamflows at Edelweiss Creek are compared with those at 240 and Dennis Creeks. Figure 4.3 shows diurnal melt waves for the period 15-17 April 1990 at Edelweiss (obtained from a series of point measurements), Dennis and 240 Creeks. For comparison, the stream discharges are expressed on a per hectare basis. Snowmelt at Edelweiss Creek was already well under way at that time and producing very distinct diurnal melt waves while the diurnal melt waves were just beginning at the other creeks. This illustrates the effect of the clear-cut at Edelweiss Creek, even with 15-year-old regeneration. Cheng (1989) documented the effect of clear-cutting 30% of a watershed at Gamp Creek on the west side of Okanagan Lake, in which an average 20% increase and 11 day advancement in peak flow was observed relative to the control watershed, Greata Creek. Many other studies have been carried out elsewhere with similar results (Bosch and Hewlett, 1982). These observations demonstrate that the changes in the magnitude and timing of snowmelt induced peak flows persist even with partial regeneration and are likely due to incomplete canopy closure. Assuming that the diurnal melt waves from the upper portion of Edelweiss Creek were the same as those at Dennis Creek on a per hectare basis (due to similar forest cover type, aspect and elevation, which regulate snowmelt), the streamflow generated from the lower portion (i.e., regenerating clear-cut) was calculated as the difference between the total flow and the flow from the upper portion. These flows are plotted along with flows from 240 and Dennis Creeks for the period 04-06 May for comparison (Figure 4.4). The maximum temperature for the April melt event was 12°C compared with 18°C at 240 Creek for the May event. Peak flows from the clear-cut for the April event were 2.7 times the peak flow at 240 Creek in the May event (eu/s/|) aBjeuosiQ despite the fact that the temperature was 6°C lower in April. The flashiness of the response at Edelweiss Creek can be attributed in part to the smaller size of the catchment compared to 240 and Dennis Creeks and the fact that it is first order rather than second, but the advancement of snowmelt induced flow peaks is clearly a result of the difference in forest cover. 4.2: Water Chemistry Modeling: 4.2.1: Background and flow component modeling Hydrograph analysis was carried out to model the chemistry of streamflow based on the chemistry of the components of runoff. This was done to integrate the various components of the study and to provide the basis for physically based computer simulation of streamflow chemistry. Various authors (Pinder and Jones, 1969; Collins and Young, 1979) have used the chemistry of runoff to separate hydrographs into direct and groundwater contributions. These studies have concentrated on analysis of stream water and runoff at different flow regimes to achieve the separation. This analysis represents the reverse approach in which an established computer model was used to separate the components of flow. The detailed knowledge of the chemical behaviour of total streamflow and its various components described earlier was then used to determine the chemical contributions of those components to streamflow. To analyze the relative contributions of the components of flow the UBC Watershed Model version 3.0 (Quick et al. 1995) was used. Water inputs from snowmelt and rainfall are routed through four reservoirs representing fast, medium, slow and very slow runoff. These reservoirs are normally interpreted as representing surface runoff, interflow, upper groundwater and deep groundwater, but this interpretation is based on the relative rates of input to and runoff from each zone. The behaviour of each reservoir is governed by allocation and recession parameters. Initially, the standard automatic calibration method was used to obtain a best fit to the observed flow data. This procedure involves varying key parameters randomly and performing iterations until the maximum value of e! is obtained. Having done this, it was found that this procedure consistently underestimated the flows, particularly the low flows. That is, while the procedure was able to represent the general shape of the hydrograph, total calibrated flows were always less than total 170 observed flow for the calibration period (September 1987 - August 1990). On closer inspection, it was found that the flow components produced by this process did not represent the physical characteristics of the flow components that had been identified by the field study. Specifically, the model represents two groundwater zones, upper and deep groundwater, whereas there are potentially three distinct groundwater zones represented by groundwater sites with different hydraulic conductivity regimes (Table 3.2.1). As an alternative, physical and chemical groundwater information was used to pre-determine the behaviour of the reservoirs rather than relying entirely on fitting the calculated flows to observed flows. Three year average streamflow at 240 Creek and groundwater levels at 240 and 241 Creeks are plotted against adjusted day of the year (as noted above, day of the year was adjusted so that peak flow occurred on May 17 each year, Figure 4.5). Groundwater heads at 240 and 241 Creeks are closely related to streamflow on 240 Creek (Figure 4.6). Because of the similarity between 240 and 241 Creeks, groundwater information collected in either drainage was assumed to be equally representative of both watersheds. It should be noted that the quantity "head" that was used to represent groundwater level is not strictly the same as total head because a common datum was not used. Thus, while the upper soil groundwater site at 241 Creek represents near surface groundwater its head is lower than at the 240 Creek site because the soil is much shallower. Upper soil groundwater may not always be near surface either, but represents relatively rapid subsurface flow that occurs at high flow through preferential flow paths. It has water chemistry characteristics indicative of hydraulic conductivity higher than that of lower groundwater. Upper soil groundwater appears at the start of snowmelt, rises more quickly than the lower soil site at 240 creek, and disappears again shortly after the end of snowmelt, around late June (Figure 4.5). In contrast, the timing and fluctuations of lower soil and bedrock groundwater at 240 Creek are quite similar despite their large differences in hydraulic conductivity (heads in these two types correlate with an of 86%). Both the lower soil and bedrock at 240 Creek are capable of providing extended baseflows, whereas the lower soil is 171 c o IS o E re CO a S CM 2 * 1 I Ji CO a) o ° 5 S CM $ a? It- »-E o (0 o £ CM CO © * E ? p > > • s 0 0) n * o LL (0 pun I I p i l l I pilllll'l o o (s/eui) JO ofZ D 3 B e i 3 A \ / jeaA-e a j i s j u a u i a m s e e w je (ui) p e a n JaiewvpunoJO 172 more responsive than the bedrock, as would be expected. This presented two options for the assignment of these groundwater zones to components in the UBC model. These two options existed because there are four routing zones in the simulator, whereas there are potentially five different outflows that are recognized at 240 Creek; direct runoff as overland flow and channel interception, interflow consisting of rapid non-Darcian subsurface runoff through the forest floor and pipes and macropores, upper soil groundwater, lower soil groundwater and bedrock groundwater. It seems reasonable to assign the bedrock groundwater to the deep zone in the UBC Model, and the soil groundwater represented by lower soil to the upper groundwater zone since their physical and chemical behaviour are different. It would then be necessary to assign the near surface groundwater at 241 Creek to medium runoff since its behaviour is much more flashy than the upper and lower groundwater and only conducts water during the high flow period, although it may not be correct to call it interflow. This would then necessitate lumping the rapid seepage together with direct runoff. The upper and lower groundwater components usually behave differently in the simulator, with deep groundwater supplying long term base flow and upper groundwater supplying short term subsurface flows as suggested by the initial calibration runs. As an alternative approach, the lower soil and bedrock could be lumped together as deep groundwater, the upper soil groundwater could then be assigned to upper groundwater, and the rapid seepage to interflow. Either approach would involve lumping two components together, and the better one would be identified as the one that produced the best fit to observed flows. These two approaches are summarized in the following table; UBC reservoirs UBC interpretation Assignment of Real Flow Components Option 1 Option 2 fast fastflow direct runoff + interflow direct runoff medium interflow upper soil groundwater interflow slow upper groundwater lower soil groundwater upper soil groundwater very slow deep groundwater bedrock groundwater lower soil + bedrock 173 4.2.2: Preliminary base flow analysis on 240 Creek Based on ratios of the measured hydraulic conductivities of the components (Table 3.2.1), upper groundwater should carry an average of 15 times the outflow of deep groundwater, and the near surface groundwater should carry 3.2 times the outflow of upper groundwater at the peak, although these proportions will vary according to the average head in each zone. Furthermore, the recession rates of the zones can be approximately pre-set based on the timing behaviour of each groundwater component in Figure 4.5. Curves were fit to the measured heads using third order polynomials that show lower soil and bedrock groundwater zones peaking about a week before peak streamflow and supplying extended base flows to the stream, whereas the upper soil groundwater peaks slightly after the peak streamflow and is virtually gone by some time in mid July. A preliminary analysis of base flow chemistry was done for 240 Creek using relationships between groundwater head and streamflow as illustrated in Figure 4.6. Table 4.1 gives the equations that were developed to predict head from streamflow at 240, Dennis and Edelweiss Creeks. Thus, groundwater heads were predicted from daily streamflow, and equations developed in section 3.2 (Tables 3.16-20) were used to predict the chemistry of groundwater at different sites and their variability with streamflow. It was assumed that there would be a permanent contribution to base flow from lower soil and bedrock at a ratio of 15:1 (this is the ratio of hydraulic conductivities between P3 and P4 at 240 Creek). Upper soil was allowed a variable contribution to base flow according to a linear function of the head in that zone, assuming that it would not contribute significantly below a head of 0.3 metres (the lower limit of head at which samples were collected), but that at its maximum head of 0.7 metres (equal to the soil depth) it would contribute three times as much water as the lower zone according to the hydraulic conductivity ratio of upper soil over lower soil as discussed above. Note that the seepage site is located in Edelweiss Creek, but head at that site has been shown not to be related to streamflow on Edelweiss Creek (Table 4.1), and only weakly related to flows on 240 and Dennis Creeks. Seepage sites are usually adjacent to streams and make up a small percentage of 174 Table 4.1: Equations to predict head from total streamflow (H in metres, Q in m3/s except where noted) Creek Zone Equation s.e. R2% P 240 zone 2 upper H = 0.68 + 0.0769 ln(Q) 0.041 85.6 0.000 zone 2 lower H = 1.07 + 0.0605 ln(Q) 0.057 67.5 0.000 bedrock H = 1.04 + 0.0650 ln(Q) 0.044 80.2 0.000 seepage H = 1.88 + 0.0092 ln(Q) 0.035 13.9 0.105 Dennis zone 2 H = 1.78 + 0.184 ln(Q) 0.235 63.4 0.000 seepage H = 1.85 + 0.0229 Q/ha (Q in litres/sec/ha) 0.035 14.0 0.104 Edelweiss zone 2 upper H = 1.02 + 0.133 ln(Q) 0.141 55.4 0.000 zone 2 lower H = 0.89 + 0.0890 ln(Q) 0.089 60.3 0.000 seepage H = 1.88 + 0.00448 ln(Q) 0.037 2.7 0.491 each watershed (about 1%). It is assumed that their contribution to flow is proportional to the proportional area that contributes to flow at, any time. Mulholland et al. (1990) suggested that a large proportion of a watershed is involved in generating runoff during peak flow. This would be a reasonable assumption at UPC since peak flows tend to be generated from snowmelt that would presumably be occurring over most of the watershed at any given time. Thus seepage contribution was assumed to vary from 1% at high flow (1.0 m3/s) to 5% at low flow (0.001 m3/s) according to the log of the discharge. This procedure is an obvious oversimplification of the behaviour of groundwater relative to streamflow, but was done to verify the composition of baseflow relative to that of streamflow using the bicarbonate content of the different groundwater zones. The bicarbonate concentration of the simulated baseflow agrees closely with that of streamflow at low flow, and is higher than that of streamflow during high flows, as it should be (Figure 4.7). The procedure has verified that low flows are derived from a combination of outflows from lower soil and bedrock, with a variable (as yet unknown) contribution from upper soil during medium and high flow periods. 4.2.3: Determination of Component Allocation for Hydrograph Modeling Both approaches for allocation of groundwater components described in section 4.2.1 were used to obtain an optimum calibration for 240 Creek. The difference between the two approaches involved variation of the amount of water allowed to enter groundwater storage and 175 o O © CM o .E co oo o cn s i O £5 o m 0) CO o = o a-s < o CL E C O o El 3 O ) CD CD o CM o E co £ co E o (0 • a a> oo cn CD E a. » a. < oo oo cn co .Q E v <* a co • a. < pill 11 1—pill 11 I—pill! 11 I o 6 o o (oes/f;uj) eBjeqosia (l/Bui) uoiiBJiueouoo ejeuoqjeoig 176 Table 4.2: UBC Model Calibration Options for 240 Creek Parameter/attribute Option 1 Option 2 Groundwater zone storage mm 4.5 11.5 Deep zone share of groundwater percolation 0.08 0.59 Fast runoff time constant for rain (days) 0.48 0.32 Fast runoff time constant for snow (days) 0.54 0.40 Medium runoff time constant for rain (days) 1.0 1.0 Medium runoff time constant for snow (days) 8.0 1.0 Upper groundwater time constant (days) 80.0 14.0 Deep groundwater time constant (days) 200.0 57.0 Coefficient of efficiency e! 0.8883 (overall) 0.9570 (1988-89) 0.9060 (1990) the recession constants of the components. Table 4.2 summarizes the differences between the two options in terms of the calibration. In fact, the calibration obtained for option 1 was not as good as that of option 2. For option 1, the peaks were too low. Accurate low flow base flows with a ratio of 15:1 between upper and lower groundwater relied on extending the time constant of upper groundwater to 80 days and setting groundwater storage low. This was necessary since interflow was being interpreted as upper groundwater, but actual groundwater storage was forced so low that flow from upper soil (interpreted as medium outflow) was about 5 times that of upper groundwater. As noted earlier, that ratio should be about 3.2. Early peaks were too low and later peaks too high, because groundwater outflow peaked about two weeks later than indicated in Figure 4.4. In comparison, option 2 achieved a higher overall efficiency with better representation of peak flows and groundwater outflow ratios. Groundwater heads were correlated against outflows for each reservoir for each option to determine which is the better option. The correlation matrices are given in Table 4.3. These correlation matrices clearly indicate that option 2 provides better agreement between the routed outflows and the groundwater heads in the zones that they were set to represent in the calibration. Those specific correlations are highlighted with bold type. The high correlation between upper soil and the natural logarithm of upper groundwater outflow would indicate that it is correct to equate these two components. Deep groundwater is a composite of lower soil and bedrock groundwater, however both those zones are more highly correlated with 177 Table 4.3: Correlation Matrices of Groundwater Heads vs Routed Outflows Option 1 H-bedrock H-lower H-upper H-lower 0.863 H-upper 0.804 0.524 Q-upper 0.454 0.347 0.827 Q-lower 0.264 0.156 0.828 Q-interflow 0.826 0.680 0.823 ln(Q-upper) 0.345 0.318 0.896 ln(Q-lower) 0.246 0.128 0.877 ln(Q-interflow) 0.883 0.898 0.573 Table 4.3: Continued Option 2 H-bedrock H-lower H-upper H/P2 H/P3 H-lower 0.863 H-upper 0.804 0.524 H/P2 0.769 0.621 0.601 H/P3 0.824 0.993 0.385 0.529 Q-upper 0.873 0.776 0.903 0.648 0.745 Q-lower 0.756 0.504 0.802 0.489 0.474 ln(Q-upper) 0.868 0.886 0.938 0.532 0.873 ln(Q-lower) 0.567 0.432 0.893 0.401 0.408 upper than with deep groundwater outflow. The same is true in option 1 except the correlation coefficients are lower. The piezometers that make up the lower soil (P2 and P3 240 Creek) were included separately in the correlation matrix in case they should be assigned to different groundwater zones. It was felt that P2 might belong with lower groundwater and P2 with upper groundwater, but the correlations do not support this theory. Correlations with groundwater outflows are higher when P2 and P3 are combined to represent lower soil groundwater, as was assumed. 4.3: Model Calibration and Water Chemistry Model Development Since it was demonstrated that option 2 was the better one, the same option was used in calibrating Dennis and Edelweiss Creeks. All calibrated hydrographs are for mean daily flows, calibrated to obtain the best fit over the period September 1987 to August 1990. Some key calibration factors for each creek are given in Table 4.4. Note that the coefficient of efficiency could not be calculated for Edelweiss Creek since continuous streamflow measurements were not collected at that creek. Instead, calculated flows were calibrated to point measurements. 178 Table 4.4: UBC Model Calibration Parameters For 240, Dennis and Edelweiss Creeks Parameter/attribute 240 Creek Dennis Edelweiss Groundwater zone storage mm 11.5 9.0 11.0 Deep zone share of groundwater percolation 0.59 0.63 0.40 Fast runoff time constant for rain (days) 0.32 0.28 0.15 Fast runoff time constant for snow (days) 0.40 0.50 0.20 Medium runoff time constant for rain (days) 1.0 1.0 1.0 Medium runoff time constant for snow (days) 1.0 1.0 1.0 Upper groundwater time constant (days) 14.0 13.0 3.0 Deep groundwater time constant (days) 57.0 53.0 13.5 Coefficient of efficiency e! 0.9570 0.9060 0.9323 NA Comparison of the information given in Table 4.4 suggests relationships between some of the parameters and physical and morphological properties of the modeled watersheds. Upper and deep groundwater time constants appear to be related to basin area, while fast runoff time constants seem related to drainage density. Groundwater storage is probably related to soil properties; a function of hydraulic conductivity is suggested, however porosity is hkely to be a factor also. These properties can be estimated from soil samples. Although it is premature to attempt development of these relationships based on three calibrated watersheds, calibration of a larger number of gauged watersheds with a range of sizes and morphological characteristics could lead to development of a system that would allow for calibration of ungauged watersheds. 4.3.1: Development of Relationships Between Groundwater Chemistry and Outflows. -Measured groundwater heads in upper soils were regressed against calculated upper groundwater outflows, and lower soil and bedrock heads against deep groundwater outflows for each creek. As discussed above for 240 Creek, correlation analysis was used preliminary to regressions on Dennis and Edelweiss Creeks to determine which groundwater sites to use to represent upper and deep groundwater outflows, and the form of the regression (i.e., linear vs log-normal). The results of those regression analyses are given in Table 4.5. Equations that relate chemical concentration to groundwater head that were developed in section 3.2 (Tables 3.2.16-20) were then used to predict the chemistry of upper and deep groundwater outflow components. However, it was assumed that the average basin-wide hydraulic conductivity of the 179 Table 4.5: Relationships of Groundwater Head vs Calculated Groundwater Component Outflows Creek Relationship Equation s.e. R2% P 240 upper soil vs upper GW H-upper = 0.854 + 0.103 ln(Qupper) 0.0370 88.0 0.000 lower soil vs deepGW H-lower = 0.858 + 1.48 Q-deep 0.0856 25.4 0.001 bedrock vs deepGW H-bed = 0.789 + 2.45 Q-deep 0.0621 57.2 0.000 Dennis zone 2 soil vs upper GW H-zone2 = 0.689 + 13.3 Q-upper 0.1889 81.9 0.000 zone 2 soil vs deep GW H-zone2 = 0.618 + 13.8 Q-deep 0.1568 87.5 0.000 bedrock vs deep GW same as for 240 Creek Edelweiss upper soil vs upper GW H-upper = 1.15 + 0.134 ln(Qupper) 0.0687 74.1 0.000 lower soil vs deepGW H-lower = 0.185 + 41.0 Q-deep 0.1619 43.8 0.000 bedrock vs deepGW same as for 240 Creek lower soil component of deep zone groundwater would also vary between low and high flow. This is because hydraulic conductivity varies inversely with soil depth, as was demonstrated in section 3.2. Thus the higher water tables that occur at high flow imply higher average hydraulic conductivity. Silica concentration was used as a tracer to make this determination since silica is non-ionic and therefore does not react with the porous medium. Periods of base flow (no fast flow or interflow) were selected for this analysis. Details of the analysis are given in Table 4.6, using 240 Creek as an example. Groundwater heads for lower soil and bedrock were determined from deep groundwater outflow using equations given in Table 4.5. The Si02 concentration in streamflow was determined using the equation given in Table 3.3.1. First, the hydraulic conductivity was determined for periods when streamflow was supplied entirely by deep groundwater by solving for K in the equation given in Table 3.2.20b as follows: [SiQ2] + 16.47-1.89(H l o w e r) 0.304(H l o w e r)-2.1 per Groundwater, 240 Creek 9 I CV) o 9.76 9.59 8.88 9.12 11.23 9.22 8.75 8.49 8.87 ! 8.37 8.25 7.91 7.97 7.23 7.81 7.11 7.07 7.03 7.53 per Groundwater, 240 Creek 0 a tfi §• 8.49 7.96 7.77 7.79 7.22 7.17 7.07 7.29 6.76 6.69 6.60 6.62 per Groundwater, 240 Creek K-lower 1.53E-06 1.68E-06 2.44E-06 2.15E-06 6.92E-07 2.04E-06 2.61E-06 3.00E-06 2.45E-06 3.18E-06 3.39E-06 4.00E-06 3.88E-06 5.55E-06 4.22E-06 5.86E-06 5.98E-06 6.09E-06 4.81E-06 per Groundwater, 240 Creek ln(K) lower -13.388 -13.298 -12.924 -13.050 -14.183 -13.101 -12.856 -12.718 ic Conductivity, Lower and Up K-upper 6.12E-06 7.44E-06 8.18E-06 8.11E-06 9.57E-06 9.66E-06 9.89E-06 8.63E-06 1.11E-05 1.14E-05 1.18E-05 1.15E-05 ic Conductivity, Lower and Up upper -12.003 -11.809 -11.714 -11.722 -11.557 -11.548 -11.524 -11.660 -11.406 -11.381 -11.351 -11.370 ic Conductivity, Lower and Up H-lower 0.865 0.867 0.877 0.873 0.859 0.871 0.880 0.886 0.877 0.891 0.895 0.910 0.907 0.959 0.916 0.970 0.975 0.979 0.933 5 Variability of Hydrau] H-upper 0.000 0.000 0.000 0.000 . 0.000 0.000 0.000 0.143 0.214 0.214 0.214 0.357 0.380 0.407 0.440 0.461 0.474 0.489 0.499 5 Variability of Hydrau] o o 1 "I 9.76 9.59 8.88 9.12 11.23 9.22 8.75 8.49 8.75 , 8.32 8.21 7.78 ) ts' 7.20 7.51 7.03 6.98 6.92 7.18 Table 4.6: Details of Calcnlatin; Deep outflow 0.005 0.006 0.013 0.010 0.001 0.009 0.015 0.019 0.013 0.022 0.025 0.035 0.033 0.068 0.039 0.076 0.079 0.082 0.051 Table 4.6: Details of Calcnlatin; Upper outflow •0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.002 0.002 0.002 0.008 0.010 0.013 0.018 0.022 0.025 0.029 0.032 Table 4.6: Details of Calcnlatin; B | 0.005 | 0.006 | 0.013 | 0.010 | .0.001 | 0.009 0.015 | 0.020 | 0.015 | 0.024 | 0.027 | 0.043 | 0.043 | 0.081 | 0.058 | 0.098 | 0.104 | 0.111 | 0.083 181 where [SiOJ is the silica concentration of streamflow due to deep groundwater. The result is then transformed to K and is related to deep groundwater outflow using a power function. In this example, K can be expressed as ~ . K| o w e r =2.0906X10- 5 Q d e e p 0 4 9 3 4 The concentration of silica in lower groundwater is then calculated using the equation given in Table 3.2.20b, and the result is given in column 12 of Table 4.6. The concentration of silica in upper groundwater is then determined as follows: b,VJ2uPr<>r - - i Q " Supper The hydraulic conductivity of upper groundwater is then determined as [ S i 0 2 U P P J + 16-47 - 1 . 8 9 ( H U P P J 0.304(H u p p e r)-2.1 Note that although upper groundwater was represented by a different groundwater site than lower groundwater, the same equation is used to calculate hydraulic conductivity and silica concentration for both components. The relationship between the hydraulic conductivity and the outflow of upper groundwater is calculated according to the following power function: K u p p e r = 3.9650X10-«Qu p p e r The variation of average basin-wide hydraulic conductivity with groundwater outflow that was calculated as described above is plotted (Figure 4.8). Note that the range of that conductivity is similar to the conductivity measured at piezometers. Using the same method, the apparent basin wide conductivity was calculated using, bicarbonate, calcium, sodium and sulphate, and it was found that in each case, the ionic chemicals yielded much lower hydraulic conductivities than the values calculated from the silica concentration. Since silica is non-reactive with respect to the porous medium, it was assumed that the relationships shown in Figure 4.8 U0B&UU90U03 eojiis UIOJJ (S/LU) y /tyAnonpuoo spuvt-uiseg 183 represent the true values of basin-wide hydraulic conductivity. Therefore, there must be an effect similar to chemical retardation (Freeze & Cherry, 1979) at work that causes reduction in the apparent conductivities due to other chemicals. Chemical retardation is a result of cation and anion exchange capacity of the soil that causes ions in the water to adsorb to soil particles. Because those apparent conductivities are lower than the true values, higher concentrations of those chemicals in base flow are indicated than would be calculated from true conductivity. It is, theorized that chemical retardation causes a decrease in the concentrations of ionic chemicals extracted from groundwater wells and piezometers and that the reverse of this effect must be accounted for when predicting the concentrations of chemicals in base flow. 4.3.2: True Basin-Wide Hydraulic Conductivity and Chemical Retardation Factors The retardation factor for ionic chemicals as discussed above for 240 Creek was calculated as a power function for lower groundwater and as a linear function for upper groundwater (Table 4.7, Figure 4.9). The power function is calculated by taking the log of both sides of the equation, and solving the equation with linear regression. The power function was used for true conductivity variability and for lower groundwater retardation factors because concentrations in base flow are related to the log of stream discharge, and the concentrations in groundwater are related to the log of hydraulic conductivity. The linear relationship for the upper groundwater retardation factors existed because there is a log-normal relationship between upper groundwater head and upper groundwater outflow. The same methods described above were us'ed to calculate true hydraulic conductivity variability and ionic chemical retardation factors for Dennis and Edelweiss Creeks (Table 4.7, Figures 4.10-4.15). R2 scores are not included in the table; for lower groundwater those scores range from 92.4% (sulphate retardation Edelweiss clear-cut) to 100% (calcium retardation Edelweiss clear-cut), and for upper groundwater from 34.4% (calcium retardation 240 Creek) to 94.4% (sulphate retardation Edelweiss clear-cut). For all three creeks, groundwater chemistry is predicted using the equations for zone 2 that are given in Tables 3.2.16-20b, and therefore is not dependent on individual standpipes. On I s U RS UH s 0 •si * 3 1 1 S in 2 o c-. a s o -a 3 u 5 ^ ' o So) in > « > J h 60TJ 5» S v. 60 ir! C C IT. T3 3 & 01 - i - o § ^ I 01 I-H 1 o rH X LO ON LO II 18 00 CN O X ON I N cO a X VO o ON O CN II o CN C u vC "3 60 0 c u 01 a, OH o o CN o o + X rH CO LO I 3 a o rH X o LO vo ON CO II ci 5^ in I o rH X rH LO rH ON II c5 I a 3 '53 s —H SH 185 Table 4.7: Continued Creek Ca retardation Na retardation 240 R l o w e r = 4.5X10^ K 0 5 0 9 3 R u p p e r=2.5X10- 6 - 0.0759Kupper R l o w e r =4.7X10-* K 0 4 2 3 5 RuPPer = 3.263X10-6 + 0.2742K Dennis R = 1.481X10-5K02134 R l o w e r = 1.8397X10"* IC"*0634 u^pper = 7.595X10"" K~° 9 6 9 7 Upper Edelweiss (forest) R = 9.. Q . > W | 2 3 C T3 £ § 3 * "I 2 « JO- a, 0 g o O r g co a •—? ro ** > 1 o » *is F i n i i rr «? UJ S L U o o rrrr LU O O J M E Sc >l ** LU O O 3 <=> -o CO c o O D X L U T3 O . _ 188 co LU o a LU O O o LU s suouej)U90UO3 saioads oiuo| LUOJJ A}iAuonpuoo juajeddv C 3 a> Z S CL fc fl) M B o> o Jo C O fl) ro o UJ S , 2 | I P «*! 5 | l 0> > C E O 3 3 - I O ra £• i l o a "io I O <5 a. s 3 o 5 o CO o o o m °5 3 o s i o 5 d "a o (9 a o o t LU O C J 1 1 - r - 1 1 ' 1 i r "1 "? «? in "? o LU LU LU LU LU • LU o o o o s o O O O O o CO CD oi c uoue-nuaouoo coins way (S/IU)M AjjAaonpuoo epiM-uiseg 189 for snow melt modeling of the upper section. Elevation bands were set to reflect the division in forest cover types such that the lower two bands used canopy conditions that reflect the regenerating clear-cut, and the upper three bands had canopy conditions identical to those of Dennis Creek at similar elevation. An attempt was made to apply the above method to calibrate the base flow chemistry, but it was found that a calibration could not be achieved without yielding unreasonably high hydraulic conductivities. It actually stands to reason that different groundwater chemistry regimes should exist for each section of Edelweiss Creek, since it has been demonstrated that forest cover has a significant effect on groundwater and streamflow chemistry. This implies that base flow chemistry must be modeled separately for the upper and lower sections of the catchment. To achieve this distinction between groundwater outflow chemistry from upper and lower sections of Edelweiss Creek, the upper section was separated out and the UBC model was rerun using the same parameter settings given in Table 4.4, with temperature data collected under the spruce-fir canopy at Dennis Creek. The streamflow chemistry for the upper section was calculated using the equations in Table 3.3.1 for specific discharge on Dennis Creek. Total and component outflows and stream chemistry.for the lower section of the catchment were then obtained by subtracting those quantities for the upper section from the quantities obtained for Edelweiss Creek at the weir. The component outflows and stream chemistry parameters so obtained were then used successfully to calibrate the base flow chemistry components of the clear-cut and forested sections separately using the method described in section 4.3.1. 4.4: Application of Stream Chemistry Model Based on UBC Model: Total streamflow chemistry modeling was based on a literal interpretation of the component outflows of upper and deep groundwater, interflow, snowmelt fastflow and rain fastflow that were generated by the UBC model. The chemistry of upper groundwater was modeled as discussed in 4.3.2. Since the deep groundwater outflow is composed of contributions from lower soil and bedrock, a proportion of the bedrock component was added to the lower soil contribution that was modeled according to 190 the method discussed in 4.3.2. to bring the modeled outflow chemistry in line with observed streamflow chemistry. It was assumed that that proportion would vary with hydrologic conditions. The ratio of lower soil outflow to bedrock outflow in deep groundwater was assumed to vary linearly from low flow to high flow conditions, since the relationship between lower soil and bedrock groundwater heads and deep zone outflow was linear. It was mentioned in section 3.2 that zone 1 streamside groundwater sites were made up of contributions from zone 2 and bedrock groundwater outflows. It was also demonstrated that the bedrock component in zone 1 groundwater is greater at high flow than at low flow and the ratios were presented in Table 3.2.21. This variation in the bedrock component was used to allow the ratio of lower soil to bedrock groundwater in deep groundwater outflow to vary from low flow to high flow. The actual ratio of lower soil to bedrock groundwater at low flow was determined by matching the SiC>2 concentrations in the groundwater components to that of streamflow, but the relative increase in the bedrock component at high flow was treated as a calibration parameter to help fit the modeled streamflow chemistry to observed concentrations. In the case of 240 Creek, it was found that the bedrock component in deep groundwater outflow varied from 4.3% at low flow to 5.8% at high flow (deep zone outflows at 0.001 and 0.113 m /^s respectively), whereas at Dennis Creek a variation in the bedrock component of deep groundwater outflow from 2.5% at low flow to 8.5% at high flow (0.001 to 0.063 m /^s) was found to provide a good fit to observed data. At Edelweiss Creek, the bedrock component was found to be unimportant. Other components that were used in the simulation were rain fastflow, snow fastflow and interflow. Rain fastflow concentrations were assumed to be identical to the average concentrations of rainfall since there was no basis to assume any relationship between rainfall chemistry and hydrologic conditions. Snowmelt fastflow was assumed to be derived from the base of the snowpack, and relationships that were developed that predict the chemistry of the base of the snow pack from daily melt data (Table 3.1.5) were used to simulate snow fastflow chemistry. Interflow was modeled on the chemistry of the slow and fast seepage (Table 3.2.22) by assuming that at low rates of interflow outflow (at or below 0.1m3/s on 240 and Dennis 191 Creeks, 0.004 m3/s on lower Edelweiss Creek and 0.013 m3/s on upper Edelweiss) the interflow chemistry is dominated by the slow seepage, whereas at high rates of outflow (0.3 m3/s at 240 and Dennis Creeks, 0.013 m3/s on lower Edelweiss and 0.038 m3/s on upper Edelweiss Creeks) the interflow is dominated by the fast seepage. Linear interpolation was used to allow the interflow chemistry to vary between these limits. The resultant total streamflow chemistry was derived as a weighted average of the chemistry of the components. That is, for each component, the chemistry variables were multiplied by the component outflows, those products were summed and then divided by the calculated stream discharge to arrive at the simulated chemistry of total stream discharge. In the case of Edelweiss Creek, this procedure was applied to the upper and lower sections separately, and the synthesized water chemistry parameters for the whole catchment were derived as a weighted average of the upper and lower sections. Examples of calibrated streamflow hydrographs and simulated water chemistry graphs are given in Figures 4.16-23 for 240 Creek, 4.24-31 for Dennis Creek and 4.32-39 for Edelweiss Creek. The hydrographs include observed and calculated flows as well as upper and deep groundwater,-and interflow. In Figure 4.39, calibrated streamflow hydrographs for the upper and lower sections are given to compare the flows generated from each section of the catchment. This is done to show that the model is capable of predicting the impact that reduced forest cover has on streamflow, since the calculated flows from the clear-cut are clearly weighted more heavily to the early part of the season, whereas the calculated flows from the forested section are weighted more heavily towards the latter part of the season, similar to Dennis Creek. Also, since the clear-cut only occupies 25 percent the area of the mature forest within the catchment, flows are higher on an area basis from the clear-cut than the forest 4.5: Discussion of Model Calibration While the calibration of the model achieved high efficiency, several possible improvements to the model became apparent during the calibration process. These 192 H- o o o o o (oes/£iu) e6jBL)osia (l/Bui) uoiiBJiueouoo ( 0 8 S / £ U l ) S6JELJ0S|Q TJ 0) •O CM g £ • °* T J O +-(8 u a 3 c 5 J= o < .= O _L .. o a T - o < * i 3 o .5? co 194 (l/Bui) uoijBJijueouoo (|/6iu) UOjlBJJUOOUOO "O O XL > CD o 2 w 0 O 2 TJ *N CO C f- o» CO o *~ OJ CO § . 2 c 0 9 3 © 5 CO O ^ . . O Q. CN © <* « S a .I?co 195 (l/Biu) u o j j B J i u s o u o o e»eud|ns 0) > »- fl) co ir ° o a> " ° 3 0 0 CO (0 T-?Is CO CB O ) ?! = & CM O < N O (0 a> B O) C O (i/Biu) uouejjueouoo (99S/CUJ) 8&iei|9S|a 0 •o ® S o o .52 Ui c • Q C O • o „ Q o > T3 CO 18 C T 5 .2 » ® « 2 . CO o) iS £; 3 3 £ < E 2 o — — W o = o c 2 . - ° m 197 (|/6lU) UOJJBJJU30UO0 i i ° = 8 "D CD o> W CO ** » o 5, E c o w o r o> o a *! o < s i P U - CO 198 (|/BU|) UOJJBJJU30UO0 "D "o 0) C D <0 (0 2 F O c » c Q T -<° to +-"° S 8 d) o ~ E c o .E C D * -w o -.. c »-CO O O. CN O < * E 1 1 U V . Q (i/Biu) uoaejjuoDuoo TJ © © o erv Cr CO w j a "E O c cn nd De 00 cn CO co 4-1 c (0 o o 3 •*-< '•*-« Ui _CB (0 3 3 1— 4-1 < E c fl) O CO u c o Q. CO o < rf" 0) © 3 hat Ui Q. i l Su 199 (|/Biu) uoiiBJjuaouoo TS O > O CO -Q o G) O o co E. cn 5 °> fl> T -CO <* t— 3 TJ C CO TJ fl) I g -3 (0 < E t o CO cu — . . o ^ o c a « o < o © co 3 « c o -ir co S S § S 5 8 U_ c i o d d c i d (s/eiu) eBjeuosiQ 73 fl) © © CO CO J2 (A O © © _ 2 o> § £ *" <0 T » * * T J U J W © M 3 S c T .15 2 w ===. £ S « c < o © o E 3 « 0 1 1 C O CM o CO "TO NJ I x : CM o co CO Q s CO 3 c SH CO •C < 2 0 2 o o o o CO o o o p T o o c\i o o (l/Biu) uouejjuGouoo T J © © CO O TJ C ro T J © © © © O to CO © © o T J * " HI 4-> CO 3 CO 3 O C D CO c o s < •— T_ *-> CO CO w © » | i CO — co < 0) c • o i n o T o o o o ci (l/Biu) uoi^cj^uaouoo (I/Diu) u o i » e j } u a o u o o 203 (s/giu) O6JBIIOS]Q © •a © O (0 © © to .«2 O 2 cn c n r CC L U ^ © C • ro .2 CO 3 ro < Q c a. i l CO © .. o 00 co o< © © 3 ro LL. — 3 CO >. o to co M 3 O) 3 3 C 3 s •c a < o o in T o o T o o 8 CN o o (l/6ui) uouBJjuaouco 204 improvements exist in areas that are likely more important in small than in large watersheds, whereas it has been noted that the simulator was developed primarily for use in much larger watersheds than the UPC study basins. First, it was not possible to obtain a calibration that accurately represented groundwater response during both high and low flow periods. Groundwater levels at 240 Creek peak slightly before the streamflow peaks, implying the groundwater outflow should peak at about the same time (Figure 4.4). This is true of both lower soil and bedrock. This groundwater behaviour could be approximated only at the expense of low flow by setting the deep groundwater time constant at 57 days causing low flows to fall to zero during the winter of 1989 and 1990 (this did not actually occur). To prevent this from occurring, a trial calibration was run in which the deep groundwater time constant was set at 100 days and the upper groundwater time constant set at 30 days. This configuration accurately represented fall and winter base flow, but calculated deep zone groundwater did not peak until the beginning of June in 1989 (measured peak occurred about 11 May), and July - August recession flows were over-estimated. Observations show that groundwater levels also rise much more quickly than the component outflows indicate when those quantities are plotted against adjusted day pf the year at 240 and Dennis Creeks (Figures 4.40 and 4.41). Stottlemeyer & Toczydlowski (1991) and Hendershot et al. (1992) also report that early snowmelt runoff is dominated by groundwater discharge. This suggests that the first month of snowmelt runoff is dominated by the variable source area concept as noted by Prevost et al. (1992) at Lac Laflamme. This effect is more pronounced at 240 Creek than at Dennis Creek. As a result there is a tendency to overestimate early melt season flows because they are weighted too heavily towards direct runoff. To compensate for this, the snow cold content factors were increased to keep these early season flows in line with observed flows, although this does not correct the problem that those flows contain too much direct runoff and too little groundwater. (ui) peeH ja ;EMpunoJ9 (s/£iu) Momno JajBMpuncuQ daaQ o O O O O O LO o u> o o CJ) co -i— r— o o T- ^ d d d d d d d (tu) |ios J S M O i 'PB9H JajBMpunojQ (s/eui) MO |uno euoz daaQ (ui) pesH jajBMpunojo (s/giu) Moi^no Ja»EMpunoj9 207 The solution to these problems probably lies in altering the outflow behaviour of the groundwater reservoirs in the model. The problem noted above does not seem to occur at Edelweiss Creek (Figures 4.42 and 4.43) although the resolution of the groundwater outflows at lower Edelweiss Creek makes it difficult to detect this effect if it exists. One possible explanation for this may lie in the groundwater recession constants; the faster recession required to calibrate Edelweiss Creek allowed the groundwater to rise more quickly, thus reflecting the true rise of groundwater more closely. It was also noted that the delayed groundwater rise was less pronounced at Dennis Creek, where the recession constants were lower than those for 240 Creek. Another problem of groundwater representation occurred in the final peak of 1990. The model caused a rise in upper and deep groundwater outflows in connection with that event, but measurements of groundwater levels indicated that this did not occur. The erroneous rise in groundwater outflows caused the subsequent recession limb of the calculated hydrograph to be too high. Precise temperature data are critical to the calibration of the model. Snowmelt models based on temperature data are often more accurate than radiation based models. This could simply be that temperature is the easiest parameter to measure with high precision. Temperature is in a sense a derivative parameter; increase in ambient temperature occurs due to the residual radiation balance after snowmelt requirements are met, and therefore it integrates several other factors. However, precise temperature is critically important to model snow melt accurately. It was found that small changes in maximum and minimum temperature of the order of 1°C had a large effect on the resultant snow melt hydrograph. In particular, it was essential to have temperature data collected under the forest canopy at Dennis and 240 Creeks. It was noted in Chapter 2 "Methods" that for some periods, temperature data had to be simulated based on data from other sites, particularly at Dennis Creek. Therefore, part of the calibration procedure involved varying estimated temperature data by +1°C, based on accurate calibrations when there was actual measured temperature data. Other effects due to temperature involved certain summer precipitation events, which tend to occur at temperatures close to 1°C. The division of r 208 precipitation into rain and snow is accomplished by assuming that when temperatures are between 0°C and 2°C (or some other value set by the user) precipitation is a mixture of rain and snow. While snow accumulation and melt is driven by temperature under the forest canopy, the occurrence of rain or snow in the summer seems to be related to temperatures outside the canopy. Thus, the model often interpreted precipitation events as snow, when the hydrograph clearly demonstrated that they were rainfall. Therefore, it was sometimes necessary to adjust temperatures upwards to account for this. The model could be improved by allowing for separate inputs of rain and snow where the user has such data, or by allowing the critical temperature to vary as in the SRM runoff model (Martinec & Rango, 1986). Other problems occurred due to orographic precipitation distribution effects. In small high elevation watersheds such as these, orographic effects are relatively unimportant compared to redistribution effects if the precipitation data is representative. However, the model does not simulate snow redistribution effects, but instead represents those effects by manipulating precipitation gradients. Snow that falls on the hillsides is either redistributed down to the valley bottom by wind, or is subjected to a greater rate of winter ablation than valley bottom snow due to the greater exposure of the hillsides. In either case, the result is that the greatest snow accumulation appears in the middle of the watershed, in the valley bottom of the upper mainstem channel. In calibrating the model, this behaviour is represented by setting a negative precipitation gradient above the middle of the watershed, and a positive gradient below, as if the snow fell in that pattern. This caused a problem in calibrating early summer rainfall and rain on snow peaks in 1990 because rainfall is not subject to redistribution. This was solved by calibrating the 1990 hydrograph separately from 1988-89 with a higher rainfall adjustment factor, however the model could be improved by including an algorithm to simulate redistribution. To summarize the above observations, the model produced very high calibration efficiencies but sometimes at the expense of physically correct flow component representation. Primarily, groundwater behaviour could be improved for small watersheds; for UPC watersheds, it was found that lower groundwater should be flashier (that is, more like a Type VI 209 exponential function) and an additional "mega-slow" groundwater component should be added to represent groundwater in bedrock or other media such as till. These features are particularly important in water chemistry, simulation. Also, the precipitation distribution section should be modified to allow for separate inputs of rain and snow, and to simulate snow redistribution. 4.6: Discussion of Water Chemistry Simulation. The example graphs that are given (Figures 4.16-38) show that the model did a good job of simulating water chemistry of streamflow in some cases, and performed less well in others. The graphs representing water chemistry simulation show the total simulated streamflow chemistry and upper and deep groundwater chemistry, along with points that represent observed streamflow chemistry. Some graphs also show the concentrations in snowmelt fastflow. An example of each chemical species that was simulated is given for one spring-summer runoff season on each creek. A method was devised to test the goodness of fit of the chemistry simulations. If the simulated values were within 20% of the observed values, the fit was considered reasonable. This criterion allowed the performance of the model to be evaluated by assessing how often the simulated values were close to the observed values on days when samples were collected (Table 4.8). However, one obvious feature of the graphs is that early season concentrations are too low, and this is likely due to the problem of groundwater behaviour that was discussed above, that calculated deep groundwater outflow rises too slowly and resultant early season flows are weighted too heavily towards direct runoff. It was noted that this effect was most pronounced at 240 Creek and least at Edelweiss Creek, and this feature is reflected in the outcome of the water chemistry modeling. For example, at 240 Creek in April 1990, simulated sodium concentrations are underestimated relative to observed sodium (Figure 4.21), whereas at Edelweiss Creek for the same period the simulated sodium concentrations are much closer to observed values (Figure 4.36). For bicarbonate on 240 Creek in April 1989 a similar underestimation occurs with later season simulation being more accurate (Figure 4.19), whereas for bicarbonate on Edelweiss Creek for the same period, the entire period of simulation is close to observed values, with early 210 Table 4.8: Frequency at which simulated chemical concentrations were within 20% of observed values (expressed as a percent) •• Creek so 4 HC0 3 Ca Na Si02 240 38 72 65 . 78 42 Dennis 80 56 92 67 81 Edelweiss 65 65 78 65 35 season concentrations only slightly underestimated. Silica behaviour on 240 Creek for the same period is similar (Figure 4.22). Because of this problem, the simulated concentrations for April and early May at 240 Creek were not considered in the goodness of fit assessment ih Table 4.8. The 20% criterion is justified because it accounts for errors in the streamflow simulations including incorrect timing of flows and incorrect proportions of the components. Model performance was assessed for the three wafer years of the simulation from September 1987 to August 1990 for 240 and Dennis Creeks, and for the latter two years on Edelweiss Creek. Thus, for 240 Creek the fit is good for sodium, fair for bicarbonate and calcium, and poor for sulphate and silica. For Dennis Creek, the fit is good for sulphate, calcium and silica, fair for sodium and poor for bicarbonate. For Edelweiss Creek, the fit is good for calcium, fair for sulphate, bicarbonate and sodium, and poor for silica. To a large extent, stream water chemistry appears to be driven by the chemistry of groundwater. This is particularly evident in the calcium simulations (Figures 4.20, 28, 35) in which the observed calcium concentrations seem to be more closely related to the curves of simulated groundwater chemistry than to total streamflow chemistry. In hydrology simulation it is convenient to conceptualize discrete flow components such as the four components used by the UBC model, whereas in reality the components are probably not entirely separate. Out of necessity, the water chemistry simulation performed in this study assumes no chemical interaction between the distinct components whose chemistry was described in chapter 3. The largest direct runoff event during the study period occurred on May 28-30 1990. During that event, the dilution effect from snow and rain fastflow that the model predicted for silica and bicarbonate was in close agreement with observed concentrations (Figures 4.27, 30, 37). For 211 sodium during that same event, simulated and observed concentrations were also very close (Figures 4.21, 36). In this case, dilution does not occur during snow fastflow, but instead sodium concentrations rise according to the melt rate as predicted by the equation in Table 3.1.5. The simulation is less successful at predicting calcium concentration during that event (Figure 4.35). The dilution effect that the simulation predicts is not observed to the same extent in the samples. Similarly, in 1989 at Dennis Creek, the simulated sodium concentrations are relatively close to the observed levels for the entire season (Figure 4.29), but for calcium the results are not as close, particularly during April. The above discussion suggests several points, as follows: 1. Because of the influence of groundwater on streamflow chemistry, accurate representation of groundwater behaviour is critical to water chemistry simulation. It has been shown that the groundwater routing section of the UBC model may need some modification to make it more representative of groundwater behaviour of watersheds such as 240 and Dennis Creeks. This may be partially responsible for the tendency of the early season simulations to underestimate observed concentrations. 2. A direct link was established between the UBC model flow components and components that were identified by field studies and yielded very good results for bicarbonate, sodium and silica. This included linking the chemistry of the base of the snow pack to the component of snowmelt fastflow. Thus, it is concluded that either the components can be viewed as independent in their chemical contributions to streamflow, or that the elevated levels of chemicals at the base of the snowpack during rapid melt are due to interaction with the soil. 3. In either case, the above point suggests that elevated levels of calcium at the base of the snowpack do in fact exist, but that the sampling failed to identify them. If this is true, it would explain why the calcium simulations, while not bad, were not as accurate as those for other chemical species. Sulphate was more difficult to simulate than other chemical species. This is most likely because concentration-discharge relationships for sulphate had R2 scores substantially lower 212 than for other species (Table 3.3.1). This could be due to high natural variability of sulphate in streamflow, or to measurement error. Sulphate concentration at 240 Creek also responds to increased concentrations in snowmelt fastflow during runoff (Figure 4.23), whereas at Edelweiss Creek, sulphate was overestimated at high flow (Figure 4.38). The sulphate concentrations at upper Edelweiss Creek were higher than at Dennis Creek for a given specific flow, but since that site was only sampled twice it was assumed that the sulphate vs specific discharge relationship for upper Edelweiss was parallel to that for Dennis Creek. The sulphate concentrations from upper Edelweiss Creek (mature spruce-fir) were actually higher than from the lower clear-cut section. Thus, the over-estimation of sulphate at high flow could be due to too high a concentration from the upper section. In contrast, the concentrations of all other species were 1 higher from the lower section. For example, sodium concentrations from the clear-cut were 3 to 4 times those from the mature forest during the runoff season of 1990 (Figure 4.39). This provides further evidence to show the effect that forest cover manipulation has on water chemistry. 4.7: Model Verification. Model verification was carried out on data for the water year Sep 1990-Aug 1991. This was done to verify the correctness of the calibration of water flow and chemistry modeling. It should be noted that while the UBC model was calibrated by maximizing the coefficient of efficiency, the water chemistry model was not calibrated in the same sense. Original relationships between component chemistry and related hydrological parameters were not altered in the simulation to improve fit, but rather relied on the correctness of the calibration of the UBC model. As noted above, groundwater behaviour could not be accurately simulated without modification of the model, and the problems that this caused for water chemistry were noted. The data used for these verification runs were processed in the same way as the 1987-90 data as described elsewhere. Verification was carried out on 240, Dennis and Edelweiss Creeks. The verification results are given in a series of graphs (Figures 4.43-48). For each creek, the observed and calculated streamflows (Figures 4.43,45,47) are in close agreement with a fit similar 213 to that achieved in the calibration runs. As before, lack of continuous flow data at Edelweiss Creek and the fact that measured streamflows are instantaneous make it difficult to assess the goodness of fit from streamflow data alone. Examples of the streamflow chemistry simulation shown include sodium calcium and silica on 240 Creek, calcium silica and sulphate on Dennis Creek and sodium calcium and sulphate on Edelweiss Creek (Figures 4.44,46,48). As with the streamflow data, the agreement between observed and simulated concentrations is similar to that of the calibration runs. The simulated water chemistry hydrographs agree in general shape and level to the observed data points such that the difference between the two (observed and simulated) data sets could be attributed to natural variability of the sampled flows, with the exception of the early melt season flows. This was probably due to the improper representation of groundwater outflows as noted earlier. ( l /Bui) UOI1EJJU0OUOO (oes/cui) aB je i jos iQ T J 0> © 2 £ o S» to .52 T -•c c ^ O c » ^ Q Ol « E < JS co ~ = O a. E « 8 < CO o c -± to >^ iZ JE CO •dui 1 N O SO o Ca J i co CO co 13 ID O T3 NI T3 N T3 • 5> CD :he e 4 ^ 2 e U) n Is Z O co O CO o 215 (|/BlU) UOIlBJiUOOUOQ 0) © O .«2 c cn c cn o T -Q -~ § I o cn § 5 o o LI r s< I I to ro o 8 - - U— 2 cn (oes/eui) eB-ieqosjQ CO 8 5? 8 8 8 8 ^ c o c o c s i c N i - r ^ - ^ d ( | /BUl) UO!»BJ}U33UO0 (s/cui) aBjeiiosia 217 CHAPTER 5: CONCLUSION 5.1. Summary. The study described above was conducted to describe the water chemistry of streamflow, precipitation and subsurface water and to determine the way in which the chemistry of those components is affected by hydrologic conditions, runoff pathways and forest cover type in three small subalpine catchments. It was demonstrated that the chemistry of those components could be predicted from hydrologic conditions and also forest cover type, and those relationships were then used with relative success to simulate the water chemistry of streamflow using an existing hydrologic watershed simulator. 5.1.1. Snowpack studies. It was demonstrated that snow pack metamorphosis and melt rate affected the chemistry of the snow pack. A process of flushing of chemicals from the snowpack during rapid melt was observed similar to that documented by other authors (e.g. Semkin & Jefferies, 1986) but that process was also accompanied by chemical enrichment of the surface of the snowpack during cold spells in the melt period so that flushing process could be repeated during the next rapid melt period. As a result of the above observations, relationships were derived to predict the chemical concentrations of the base of the snow pack as a function of daily melt In general, concentrations at the base of the snow pack were found to increase with increasing melt according to an exponential curve. These relationships were incorporated into the chemistry simulations to describe the chemistry of snowmelt fastflow, and the modeling results suggested that those relationships reflected chemical interaction between the base of the snow pack and the soil surface during rapid melt. A method was developed that uses a temperature index to predict liquid water content of a melting snowpack. The temperature index is based on current temperature and the previous day's average temperature for upper, middle and lower snowpack layers, and is similar to methods df accounting for snowpack cold content (e.g. USACE, 1956). This suggests that the 218 liquid water content depends partly on a memory of the previous day's melt conditions. The relationships of liquid water content to temperature index are described by exponential curves. As noted in section 1.2, some physically based snowmelt models require estimation of this quantity because of its effect on rates of water transmission through the snowpack. 5.1.2. Groundwater and streamflow chemistry. Four distinct groundwater zones were identified including forested hillside and streamside zones, bedrock and seepage sites. The hillside sites were primarily used for modeling with small contributions from bedrock and seepage. It was shown that the chemistry of streamside groundwater sites is based on varying groundwater contributions from the hillside and bedrock. Ground slope and hydraulic conductivity were both tested in combination with groundwater head and dummy variables to represent soil and forest cover type as predictors of groundwater chemistry. It was found that hydraulic conductivity was the better predictor; it was the more significant variable, and slope was a poorer predictor given the range of land slopes present in the study watersheds. The relationships relating groundwater chemistry to hydraulic conductivity, head and forest cover type at the hillside sites were used in the simulations by first relating both head and hydraulic conductivity to the calibrated groundwater outflow. Upper and lower soil head was related to upper and deep groundwater outflow using linear and log-normal regression. Silica in streamwater at base flow was used as a tracer to develop a surrogate for average basin-wide hydraulic conductivity as a function of groundwater outflow. This analysis predicted that at low flow, the average basin-wide conductivity of upper and deep groundwater components would be similar to hydraulic conductivities measured at the piezometers, and at high flow, about an order of magnitude higher. This is reasonable because at low flow, base flows are derived from groundwater sources close to the stream channels in the lower parts of the watershed where the piezometers are located. At high flow, trunoff is generated due to snowmelt that occurs over a large proportion of the watershed, and consequently baseflow would be derived from groundwater originating not only from the valley bottoms, but also from the upper hillsides 219 where soils are coarser and less developed. Those soils consequently have much higher hydraulic conductivity than the soils on the lower slopes, thus resulting in a higher average basin-wide hydraulic conductivity at high flow by integrating the effect of all soils that supply groundwater to the stream at any given time. Generally, the regenerating clear-cut had the highest concentrations of chemical species in groundwater, the spruce-fir forest the lowest, and the lodgepole pine forest was intermediate. These results are in keeping with the results of the streamflow chemistry investigations, although streamflow chemical concentrations on Edelweiss Creek were intermediate between those of 240 Creek (lodgepole pine) and Dennis Creek (spruce-fir) because of the mixture of mature spruce-fir and immature regeneration types within Edelweiss Creek drainage. The results of the soil survey analysis showed that soils in the clear-cut at Edelweiss Creek are similar to those in Dennis Creek both in terms of texture and available cations. This led to the conclusion that the cause of the relatively high chemical concentrations at Edelweiss Creek is incomplete water quality recovery after logging. This conclusion was supported by other evidence including stream chemistry profiles and statistical comparison of chemical regimes of UPC and other research watersheds. Water chemistry of UPC was found to be typical of catchments with gneissic or granitic lithology, which are known to be susceptible to acidification from atmospheric pollutants. Cluster analysis suggested a similarity of effect between disturbance due to logging and acidification, although the impacts are intense and relatively short lived, whereas the impacts of acidification are more long-term and chronic. High R2 values associated with concentration vs discharge relationships for streamflow indicate that water chemistry is very strongly controlled by mineral weathering. 5.1.3. Hydrology of UPC Clearrcut harvesting, even with about 15 years of regeneration, had a large impact on streamflow involving much higher amplitude diurnal snowmelt waves with higher peaks and 220 earlier melt than at mature forested sites. This effect was likely exacerbated by soil frost, which has been shown to reduce (but not prevent) infiltration. In spite of frozen soil, groundwater levels rose rapidly during early melt such that early streamflows were likely dominated by groundwater outflow. This supposition was supported by the water chemistry simulations. Interflow was observed for brief time intervals around peak flow, as was saturation overland flow. In the above features, the hydrological processes at UPC appear to be similar to those at Lac Laflamme (e.g. Roberge & Plamondon, 1986). Both also contain coniferous forest cover. 5.1.4. Hydrologic and chemical simulation. Wherever possible, links were established between observed flow components and calibrated UBC model components. Within the hillside groundwater sites, there was reason to assign some sites to upper soil and lower soil groundwater for use in the chemical simulations. Assignment of the flow components identified in the field studies to the components that were calculated by the UBC model was successful at simulating water chemistry. It was found that streamflow water chemistry is largely driven by groundwater chemistry. The differences in groundwater chemistry between the upper (mature forest) and lower (regenerating clear-cut) sections of Edelweiss Creek were used successfully to model the streamflow chemistry of the entire catchment, and to demonstrate the differences between the two sections in terms of the generation of chemicals. These results support the above discussed conclusion that forest harvesting has a large potential to impact streamflow chemistry by causing a significant increase in the concentrations of chemical species with the exception of sulphate, which decreases. Deficiencies in the way that the model handles groundwater outflow were identified. The inaccuracies in the handling of the groundwater component explain the tendency of the chemistry simulator to underestimate early melt season concentrations. However, the relative success at modeling the differences in water chemistry between the upper and lower sections of Edelweiss Creek despite the lack of continuous streamflow data, suggest that the model shows 221 great promise as a modeling framework to predict the impacts of forest manipulation on water flow and quality, given the proper development. 5.2. Recommendations. The above conclusions lead to several recommendations, as follows. 1. The groundwater outflow section of the UBC model should be modified to better represent the observed behaviour of groundwater in the study area. Other identified problems that were discussed in section 4.5 should be attended to as well. 2. The groundwater and snowpack studies that were carried out at 240, Dennis and Edelweiss Creeks should be continued. It would be particularly useful to establish more groundwater sites in areas that are slated for harvesting in 241 and Dennis Creeks before the harvesting occurs, and to monitor those sites before and after treatment. This monitoring could be further enhanced by establishing weirs on first order tributaries within those watersheds to allow for monitoring of the water chemistry response of those tributaries to harvesting, assuming that the level of harvest within those tributary drainages is high. Monitoring of existing groundwater sites should resume, and some sites should be equipped with pressure transducers to provide continuous measurement of groundwater levels. This will assist in the further quantification of forestry related changes in water chemistry and model development. Other such studies will be set up in different regions to quantify those changes under different hydrologic environments. The weir on Edelweiss Creek should be rebuilt Another weir should be constructed at the outlet of the upper section, and both weirs should be equipped with continuous stage monitoring. This would serve the purpose of providing a different order of scale for model development as a comparison to the three larger creeks, so that the groundwater section can be fully developed. The two weirs discussed above would also allow for improved monitoring of water chemistry from the upper and lower sections of the catchment, and over time would lead to modeling of water chemistry recovery after logging. 3. A system to determine the origin of chemically enriched water at the base of the snowpack during rapid melt should be implemented. It is unknown whether those high chemical 222 concentrations are derived within the snowpack, or are a result of interaction between the melt water and the soil/forest floor. This question can be answered by selecting a suitable study plot and installing collecting snowmelt and forest floor lysimeters. The snowmelt lysimeter would isolate the snow that it collects from the forest floor; the matching forest floor lysimeter would collect runoff that occurs through the forest floor, identified as direct runoff or interflow depending on the level at which the lysimeter was installed. Both lysimeters should be capable of measuring the outflow rate and collecting samples of the outflow to improve water chemistry vs outflow prediction. 4. Because canopy temperatures and representative precipitation are identified as critical factors, it is necessary to monitor air temperature under the forest canopy at Dennis Creek, and to install a snow and rain monitoring site at the mid watershed level within the drainage area. Another such site should be installed at high elevation within 240 or 241 Creek. 5. UBC model calibration should be undertaken on a wide range of other gauged watersheds with varying morphologies and drainage areas. If this is done, it is likely that a method will be developed to determine model calibration parameters based on basin morphology. This is an essential step in developing the model as a tool to predict forestry related impacts in watersheds that are not gauged. 61 To improve UBC model performance it should be converted into a 32 bit Windows program using a combination of 32 bit FORTRAN and C++ compilers. For long calibration runs, the model would run a lot faster. The water chemistry simulator that was developed as a set of spreadsheet formulae should also be programmed into the model. 223 REFERENCES Anderson, E.A. 1968. Development and testing of snow pack energy balance equations. Water Res. Res. 4(1): 19-37. Anderson, E.A. 1973. National Weather Service River Forecast System - snow accumulation and ablation model. NOAA Tech. Memo. NWS HYDRO-17. Anderson, E.A. 1976. A point energy and mass balance model of snow cover. NOAA Tech. Rep. NWS HYDRO-19, U.S. Department of Commerce, Silver Spring, Md. Barbaris, B. & E.A Betterton, 1994. Snow chemistry and physics of the Mogollon Rim in Arizona. Proceedings of the Western Snow Conference, Santa Fe, N.M. 1994 Baron, J. 1983. Comparitive water chemistry of four lakes in Rocky Mountain National Park. Water Res. Bull. 19(6): 897-902. Baron, J., M.A. Arthur, A.S. Denning, M.A. Harris, M.A. Mast, D.M. McKnight, P. McLaughlin, B.D. Rosenlund, S.A. Spaulding & P.M. Walthall. 1992. Biogeochemistry of a Subalpine Ecosystem, Loch Vale Watershed. Springer-Verlag, New York. Barry, R., M. Prevost, J. Stein & A.P. Plamondon. 1990. Simulation of snowmelt runoff pathways on the Lac Laflamme watershed. J. Hydrol. 113:103-121. Bergstrom, S. & G. Lindstrom. 1987. Experience from integrated simulations of runoff, water quality and residence time. Int. Symp. on Acidification and Water Pathways. Bolkesjo, Norway. Bernier, P. Y. 1982. VSAS2: A revised source area simulator for small forested basins, unpubl. PhD thesis, U. of Georgia. Binkley, D. & T.C. Brown. 1993. Forest practices as nonpoint sources of pollution in North America. Water Res. Bull. 29(5): 729-740. Bloschl, G. and R. Kirnbauer. 1991. Point snowmelt models with different degrees of complexity - internal processes. J. Hydrol. 129:127-147. Bloschl, G., R. Kirnbauer & D. Gutknecht 1991a. A spatially distributed snowmelt model for application in alpine terrain. Snow, Hydrology and Forests in High Alpine Areas. IAHS Publ. 205: 51-60. Bloschl, G., R. Kirnbauer & D. Gutknecht. 1991b. Distributed snowmelt simulations in an alpine catchment 1. Model evaluation on the basis of snow cover patterns. Water Res. Res. 27(12): 3171-3179. Bloschl, G., R. Kirnbauer & D. Gutknecht. 1991c. Distributed snowmelt simulations in an alpine catchment 2. Parameter study and model predictions. Water Res. Res. 27(12): 3181-3188. Bonell, M. 1993. Progress in the understanding of runoff generation dynamics in forests. J. Hydrol. 150(2/4): 217-275. Bosch, J.M. and J.D Hewlett, 1982. A review of catchment experiments to determine the effect of vegetation changes on water yield and evapotranspiration. J. Hydrol. 55: 3-23. 224 Brakensiek, D.L., H.B. Osborne & W.J. Rawls. 1979. Field Manual for Research in Agricultural Hydrology. USDA Handbook # 224. Cheng, J.D., T.A. Black and R.P. Willington. 1975. The generation of stormflows from small forested watersheds in the Coast Mountains of southwestern British Columbia. Proc Can. Hydrol. Symp., August 11-14, Winnipeg. 542-551. Cheng, J.D. 1982. Upper Penticton Creek Experimental Watershed Study. Kamloops Region, B.C. Ministry of Forests and,Lands. , Cheng, J.D. 1988. Subsurface stormflows in the highly permeable forested watersheds of southwestern British Columbia. J. Contam. Hydrol. 3:171-191. Cheng, J.D. 1989. Streamflow changes after clear-cut logging of a Pine Beetle-infested watershed in southern British Columbia, Canada. Water Res. Res. 25 (3): 449-456. Colbeck, CS. 1981. A simulation of the enrichment of atmospheric pollutants in snow cover runoff. Water Res. Res. 17(5): 1383-1388. Colbeck, S.C. and G. Davidson. 1973. Water percolation through homogeneous snow. The Role of Snow and Ice in Hydrology. IAHS Publ. 107: 242-257. Colbeck, S.C. 1974. Water flow through snow overlying an impermeable boundary. Water Res. Res. 10(1): 119-123. , " ' Collins, D.N. and G.J. Young. 1979. Hydrochemical separation of components of discharge in alpine catchments. Western Snow Conf. Proc. 47, Reno Nevada: 1-9. Cornwell, J.C. 1992. Cation export from Alaskan arctic watersheds. Hydrobiologia 240:15-22. Cronan, C.S., W.A. Reiners, R.C Reynolds, Jr. & G.E. Lang. 1978. Forest floor leaching: contributions from mineral, organic, and carbonic acids in New Hampshire subalpine forests. Science 200: 309-311. Cundy, T.W., K.N. Brooks & D. Sveum. 1980. Snowmelt modeling of upper Missouri River Watersheds with SSARR. Western Snow Conf. Proc. 48, Laramie Wyo: 671-680. Davis, J.C. 1973. Statistics and Data Analysis in Geology. John Wiley & Sons, Toronto. Denning, A.S., J. Baron, M.A. Mast & M. Arthur. 1992. Hydrological pathways and chemical composition of runoff during snowmelt in Loch Vale Watershed, Rocky Mountain National Park, Colorado, U.S.A. Water Air Soil Pollut. 59:107-123. de Vries, J. & T.L. Chow. 1978. Hydrologic behaviour of a forested mountain soil in coastal British Columbia. Water Res. Res. 14(15): 935-942. Drake, J.J. and D.C. Ford. 1974. Hydrochemistry of the Athabasca and North Saskatchewan Rivers in the Rocky Mountains of Canada. Water Res. Res. 10(6): 1192-1198 Draper, N.R. & H. Smith. 1981. Applied Regression Analysis. John Wiley & Sons, Toronto. Dunne, T. & R.D. Black. 1970. Partial area contributions to storm runoff in a small New England watershed. Water Res. Res. 6(5): 1296-1311. 225 Dunne, T., A.G. Price & S.C. Colbeck. 1976. The generation of runoff from subarctic snowpacks. Water Res. Res. 12(4): 677-685. Dunne, T. 1978. Field studies of hillslope processes. In M.J. Kirkby, Ed., Hillslope Hydrology. Wiley & Sons. New York. 227-293. Dunne, T. and L. Leopold. 1978. Water in Environmental Planning. W.H. Freeman and Company, San Francisco. English, M.C, D.S. Jeffries, N.W. Foster, R.G. Semkin & P.W. Hazlett 1986. A premilinary assessment of the chemical and hydrological interaction of acidic snowmelt water with the terrestrial portion of a Canadian Shield Catchment. Water Air Soil Pollut. 31 (1/2): 27-34. Espeby, B. 1990. Tracing the origin of natural waters in a glacial till slope during snowmelt. J-Hydrol. 118(1/4): 107-127. Freeze, R.A. 1972. The role of subsurface flow in generating surface runoff. 2. Upstream source areas. Water Res. Res. 8:1271-1283. Freeze, RA. and J.C. Cherry. 1979. Groundwater. Prentice-Hall, Englewood Cliffs, N.J. U.S.A. Feller, M.C. & J.P. Kimmins. 1979. Chemical characteristics of small streams near Haney, British Columbia. Water Res. Res. 15(2): 247-258. Feller, M.C. 1986. Analysis and Interpretation of Stream and Precipitation Chemistry at the Arnews Plot at the U.B.C. Research Forest D.S.S. Canada Report Frederiksen, R.L. 1972. Nutrient budget of a Doulgas-fir forest on an experimental watershed in western Oregon, in: Proceedings: Research on Coniferous Forest Ecosystems - A Symposium: 115-131. Gibbs, R.J. 1970. Mechanisms controlling world water chemistry. Science 170:1088-1090. Gjessing, E.T., A. Henriksen, M. Johanessen & R.F. Wright. 1976. Effects of acid precipitation on fresh water chemistry. Impact of Acid Precipitation on Forest and Freshwater Ecosystems in Norway. Fr6176: 65-85. SNSF Project, Oslo. Golding, D.L. 1982. Snow accumulation patterns in openings and adjacent forest. Proc. of Canadian Hydrol. Symp. 82, Fredericton N.B. Assoc Committee on Hydrol, NRC Canada: 91-112. Golding, D.L. & R.O. Hudson. 1988. Upper Penticton Creek Water Quality Study; year-end report March 1988. Faculty of Forestry, U.B.C. Golding, D.L. & R.O. Hudson. 1989. Upper Penticton Creek Water Quality Study; year-end report March 1989. Faculty of Forestry, U.B.C. Golding, D.L. & R.O. Hudson. 1990. Upper Penticton Creek Water Quality Study; year-end report May 1990. Faculty of Forestry, U.B.C. Golding, D.L. & R.O. Hudson. 1991. Upper Penticton Creek Water Quality Study; year-end report April 1991. Faculty of Forestry, U.B.C. 226 Goodison, B. 1978. Accuracy of Canadian Snow Gage Measurements. J. Applied Meteorology 17: 1542-1548. Granger, R.J., D.S. Chanasyk, D.H. Male and D.I. Norum. 1977. Thermal regime of a prairie snowcover. Soil Sci. Soc. Am. J. 41: 839-842. Haltiner, J.P. and J.D. Salas. 1988. Short-term forecasting of snowmelt runoff using ARMAX models. Water Res. Bull. 24(5): 1083-1089. Harker, A. 1964. Petrology for Students. 8th edition. Cambridge University Press. Hazlett, P.W.; English, M.C.; Foster, N.W. 1992. Ion enrichment of snowmelt water by processes within a podzolic soil. J. Environ. Qual. 21(1): 102-109. Hendershot, W.H., L. Mendes, H. Lalonde, F. Courchesne & S. Savoie. 1992. Soil and stream water chemistry during spring snowmelt Nordic Hydrol. 23:13-26. Henderson, G.S., W.T. Swank, J.B. Waide and CC. Grier. 1978. Nutrient budgets of Appalachian and Cascade region watersheds: a comparison. Forest Sci. 24(3): 385-397. Hetherington, E.D. 1982. A first look at logging effects on the hydrologic regime of Carnation Creek experimental watershed. In: Proc. Carnation Creek Workshop, a 10-year review. Pacific Biological Station, Nanaimo B.C. Hetherington, E.D. 1995. Subsurface water flow rates over bedrock on steep slopes in the Carnation Creek experimental watershed. Mountain Hydrology; Peaks and Valleys in Research and Applications: Proc. of Conf., Vancouver B.C., ed Guy, B.T. and J. Barnard. CSHS/CWRA, Cambridge Ontario. Hetherington, E.D., B. Walsh and M. Leytham. 1995. Calibration of HSPF hydrologic simulation model using Carnation Creek Experimental Watershed data. Mountain Hydrology; Peaks and Valleys in Research and Applications. Proc. of Conf. May 1995 Vancouver B.C. CSHS/CWRA Cambridge Ont Hewlett, J.D. and A.R. Hibbert. 1967. Factors affecting the response of small watersheds to precipitation in humid areas. Int. Symp. on For. Hydrol., Pergammon, Oxford: 275-290. Holland, S.S. 1976. Landforms of British Columbia. B.C. Dept. of Mines and Petroleum Resources Bulletin 48. Hudson, R.0.1991. Upper Penticton Creek Watershed Experiment; Experimental Treatment Plan. Robert Hudson, Forest Hydrologist, Vancouver B.C. Hvorslev, M.J. 1951. Time lag and soil permeability in groundwater observations. U.S.A.C.E. Waterways Exp. Sta. Bull. 36, Vicksburg Miss. Jacks, G. & T. Paces. 1987. Chemical changes in acid runoff along its pathway through granitic minicatchments. Int Symp. on Acidification and Water Pathways. Bolkesjo, Norway. Jefferies, D.S., J.R.M. Kelso & I.K. Morrison. 1988. Physical chemical and biological characteristics of the Turkey Lakes watershed, central Ontario, Canada. Can. J. Fish. Aquat. Sci. 45, Suppl: 3 Johanson, R.C., J.C. Imhoff, J.L. Kittle, Jr. & A.S. Donigian. 1984. Hydrological Simulation Program - FORTRAN (HSPF): Users Manual for Release 8.0. U.S. EPA, Athens, Ga. 227 Johanessen, M. & A. Henriksen. 1978. Chemistry of snow meltwater: changes in concentration during melting. Water Res. Res. 14(4): 615-619. Johnson, A.H. and R.C. Reynolds Jr. 1977. Chemical character of headwater streams in Vermont and New Hampshire. Water Res. Res. 13(2): 469-473. Johnsson, H. & L.C. Lundin. 1991. Surface runoff and soil water percolation as affected by snow and soil frost J. Hydrol. 122 (1/4):141-159. Kane, D.L. & J. Stein. 1983. Water movement into seasonally frozen soils. Water Res. Res. 19(6): 1547-1557. Kelso, J.R.M., CK. Minns, J.H. Lipsit & D.S. Jefferies. 1986. Headwater lake chemistry during the spring freshet in north-central Ontario. Water Air Soil Pollut 29: 245-259. Kim, S., J.R. Busch and M. Monlau. 1986. Snowmelt runoff simulation using a degree-day and nonlinear multiple recession model. Proc. West. Snow. Conf. Ft. Collins, Apr 1986:134-145. Kirkby, M.J. 1988. Hillslope runoff processes and models. J. Hydrol 100: 315-339. Kozak, A. 1970. A simple method to test parallelism and coincidence for curvilinear, multiple linear and multiple curvilinear regressions. Prepared for the 3rd Conference of the Advisory Group of Forest Statisticians, Section 25,1.U.F.R.O., Jouy-en-Joses, France, September 7-11,1970. Krajina, V.J. 1969. Ecology of forest trees in British Columbia. Ecol. Western N. Amer. 2:1-46. Lawson, D.W. 1968. Groundwater flow systems in the crystalline rocks of the Okanagan Highland, British Columbia. Can. J. Earth Sci. 5: 813-824. Leaf, CF. 1966. Free water content of snowpack in subalpine areas. Proc. West Snow Conf. Seattle WA 1966:17-24. Likens, G.E., F.H. Bormann, N.M. Johnson, D.W. Fisher & R.S. Pierce. 1970. Effects of forest cutting and herbicide treatment on nutrient budgets in the Hubbard Brook watershed-ecosystem. Ecological Monographs 40: 23-47. Likens, G. E., F.H. Bormann, R.S. Pierce, J.S. Eaton & N.M. Johnson. 1977. Biogeochemistry of a forested ecosystem. Springer, New York. Likens, G.E., F.H. Bormann R.S. Pierce & W.A. Reiners. 1978. Recovery of a deforested ecosystem. Science 199: 492-496. Linsley, R.K. and J.B. Franzini. 1979. Water Resources Engineering. McGraw-Hill Inc., New York Male, D.H. and D.M. Gray. 1981. Snowcover ablation and runoff. In Handbook of Snow, ed. D.M. Gray and D.H. Male. Pergammon Press, Willowdale Ont. Martin, C.W., D.S. Noel & CA. Federer. 1984. Effects of forest clearcutting in New England on stream chemistry. J. Environ. Qual. 13: 204-210. Martin, C.W., D.S. Noel & CA. Federer. 1985. Clearcutting and the Biogeochemistry of Streamwater in New England. J. For. 83(11): 686-689. 228 Martinec, J. 1960. The degree-day factor for snowmelt runoff forecasting. IAHS Publ. 51: 468-477. Martinec, J. and A. Rango. 1981. Aerial distribution of snow water equivalent evaluated by snow cover monitoring. Water Res. Res. 17(5): 1480-1488. \ Martinec, J., A. Rango and E. Major. 1983. The Snowmelt-Runoff Model (SRM) User's Manual. NASA Ref. Publ. 1100 Washington D.C. Maule, CP. and J. Stein. 1990. Hydrologic flow path definition and partitioning of spring meltwater. Water Res. Res. 26(12) pp 2959-2970. McKnight, D.M. & K.E. Bencala. 1990. The chemistry of iron, aluminum, and dissolved organic material in three acidic, metal-enriched, mountain streams, as controlled by watershed and in-stream processes. Water Res. Res. 26(12): 3087-3100. Mosley, M.P. 1979. Streamflow generation in a forested watershed, New Zealand. Water Res. Res. 15(4): 795-806. Mulholland, P.J., G.V. Wilson & P.M. Jardine. 1990. Hydrogeochemical response of a forested watershed to storms: effects of preferential flow along shallow and deep pathways. Water Res. Res. 26(12): 3021-3036. Nicolson, J.A., D. Craig & N.W. Foster. 1987. Precipitation, surface and subsurface water chemistry in a tolerant hardwood forest basin. Int Symp. on Acidification and Water Pathways. Bolkesjo, Norway. Nodvin, S.C, CT. Driscoll & G.E. Likens. 1988. Soil processes and sulfate loss at the Hubbard Brook Experimental Forest. Biogeochemistry. 5(2): 185-199. Paces, T. 1986. Weathering rates of gneiss and depletion of exchangeable cations in soils under environmental acidification. J. Geol. Soc. 143: 673-677. Parkinson, D.L. 1985. U-Pb Geochronometry and Regional Geology of the Southern Okanagan Valley, B.C.: The Western Boundary of a Metamorphic Core Complex. M.Sc. thesis, Univ. of B.C., Vancouver, B.C. { Perla, R.L, R.E. Davis, J. Dozier & E.R. LaChapelle. 1985. Dilution method for measuring liquid water in snow: Field tests in a subalpine snowpack. Snow Properties Measurement Workshop, Lake Louise, Alberta, 1-3 April, 1985: 71-78. Perla, R.I. & M. Martinelli Jr. 1976. Avalanche Handbook. Agric. Handbook 489, USDA Forest Service. Phillips, R.A. and K.M. Stewart. 1990. Longitudinal and seasonal variations in a northern Appalachian stream. Water Res. Bull. 26(3): 489-498. Pilgrim, D.H., D.D. Huff & T.D. Steele. 1978. A field evaluation of surface and subsurface runoff, II: Runoff Processes. J. Hydrol. 38: 319-342. Pinder, G.F. and J.F. Jones. 1969. Determination of the groundwater component of peak discharge from the chemistry of total runoff. Water Res. Res. 5(2): 438-445. Prevost, M., R. Barry, J. Stein & A.P. Plamondon. 1990. Snowmelt runoff modeling in a balsam fir forest with a variable source area simulator (VSAS2). Water Res. Res. 26(5): .1067-1077. 229 Prevost, M, R. Barry, J. Stein & A.P. Plamondon. 1991. Snowmelt modeling in a balsam fir forest: comparison between an energy balance model and other simplified models. Can. J. For. Res. 21 (1):1-10. Price, A.G. and T. Dunne. 1976. Energy balance computations of snowmelt in a subarctic area. Water Res. Res. 12(4): 686-694. Quick, M.C, A. Pipes, D. Nixon, E. Yu, A. Loukas, R. Millar, H. Assaf and B. Start. 1995. U.B.C. Watershed Model Manual, Version 4.0. Mountain Hydrology Group, Department of Civil Engineering, University of British Columbia, Vancouver, B.C. Rango, A. and J. Martinec. 1981. Accuracy of snowmelt runoff prediction. Nordic Hydrol. 12(4/5): 265-274. ' Rango, A. 1989. Evolution of a research-oriented snowmelt-runoff simulation model into an operational forecasting tool. Proc. West Snow Conf., Fort Collins, April 1989: 45-51. Rango, A. 1992. Worldwide testing of the Snowmelt-Runoff Model with applications for predicting the effects of climate change. Nordic Hydrol. 23:155-172. Rascher, CM., CT. Driscoll & N.E. Peters. 1987. Concentration and flux of solutes from snow and forest floor during snowmelt in the west-central Adirondack region of New York. Biogeochemistry 3:209-224. Reid, D.C, D.A. MacLeod & M.S. Cressen. Factors affecting the chemistry of precipitation and river water in an upland catchment. J. Hydrol. 50:129-145. Roberge, J. & A.P. Plamondon. 1987. Snowmelt runoff pathways in a boreal forest hillslope, the role of pipethroughflow. J. Hydrol. 95 (1/2): 39-54. Sauter, K.A. & J.J. McDonnell. 1992. Prediction of snowmelt rates at a forested alpine site in northern Utah. Proc. West. Snow Conf., Jackson Wyoming. Schindler, D.W., R.W. Newbury, K.G. Beaty & P. Campbell. 1976. Natural water and chemical budgets for a small precambrian lake basin in central Canada. J. Fish Res. Bd. Canada 33:2526-2543. Scrivener, J.C. 1974. Water, water chemistry and hydrochemical balance of dissolved ions in Carnation Creek watershed Vancouver Island, July 1971-May 1974. Tech. Rep. 564, Environment. Canada Nanaimo B.C. Semkin, R.G. & D.S. Jefferies. 1986. Storage and release of major ionic contaminants from the snowpack in the Turkey Lakes Watershed. Water Air Soil Pollut. 31: 215-221. Sit, V. and M. Poulin-Costello. 1994. Catalog of Curves for Curve Fitting. Research Branch, Ministry of Forests, Victoria B.C. Sklash, M.G. & R.N. Farvolden. 1979. The role of groundwater in storm runoff. J. Hydrol. 43: 45-65. Smith, M. 1984. An Ecological Classification of Three Sub-basins in the Penticton Creek Watershed in the Engelmann Spruce Subalpine Zone. B.Sc. thesis, Faculty of Forestry, U.B.C. \ 230 Storr, D. and D. Golding. 1974. Preliminary water balance evaluation of an intensive snow survey in a mountainous watershed. In: Advanced Concepts and Techniques in the Study of Snow and Ice Resources. Dec 2 r 2 6 1 9 7 3 Monterrey Ca.; National Academy of Sci., Washington D.C.: 294-303. Stottlemeyer, R. & C. Troendle. 1987. Trends in streamwater chemistry and input output balances, Fraser Experimental Forest, Colorado. USDA Forest Service Res. Pap. RM-275. Stottlemeyer, R. & D. Toczydlowski, 1991 . Stream chemistry and hydrologic pathways during snowmelt in a small watershed adjacent Lake Superior. Biogeochemistry 1 3 : 1 7 7 - 1 9 7 . ' Sugawara, M.I., I. Watanabe, E. Ozaki & Katsuyama. 1984. Tank Model With Snow Component National Research Center for Disaster Prevention, #69. Japan. Technicon Industrial Systems. 1971 . AutoAnalyzer II, Industrial Methods. Technicon Instruments Corporation, Tarrytown N.Y. Thorne, J.F., J.E. Anderson & K.M. Horiuchi. 1988. Cation cycling in a base-poor and base-rich northern hardwood forest ecosystem. J. Environ. Qual. 17:95-101. U. S. Army Corps of Engineers. 1956. Snow hydrology: summary report of the snow investigations. U. S. Army Corps of Engineers, North Pacific Division, Portland, OR. U. S. Army Corps of Engineers. 1976. Development and Application of the SSARR Model: summaries of technical reports. U. S. Army Corps of Engineers, North Pacific Division, Portland, OR. Vitousek, P.M. 1977. The regulation of element concentrations in mountain streams in the northeastern United States. Ecological Monographs 47: 65-87. Vitousek, P.M., J.R. Gosz, CC. Grier, K.M. Melillo & W.A. Reiners. 1982. A comparitive analysis of potential nitrification and nitrate mobility in forest ecosystems. Ecological Monographs 52(2): 155-177. Wakahama, G. 1968. The metamorphism of wet snow. IAHS Publ. 79: 370-379. Weyman, D.R. 1973. Measurements of the downslope movement of water in a soil. J. Hydrol. 20 : 267-288. Whipkey, R.Z. 1965. Subsurface stormflow from forested slopes. Bull, of I.A.S.H. 10(2): 74-85. White, D C 1977. Chemical Variation in a Small Mountain Watershed. M.Sc. theses, Utah State University. World Meteorological Organization. 1986. Intercomparison of Models of Snowmelt Runoff. Operational Hydrology Report No.23. WMO No. 646. Geneva, Switzerland. Zeman, L.J. 1973 . Chemistry of Troposheric Fallout and Streamflow in a small Mountainous Watershed Near Vancouver, British Columbia. Ph.D. thesis, U.B.C.