Measurements of depth, and the average velocity at each vertical, were entered into a spreadsheet template designed to calculate discharge by the mid section method. (TABLE: sample discharge calculation) The mid section method computes the sum of discharges for all segments between two successive verticals, and is currently used by the U.S. Geological Survey (Rantz and others, 1982). The computation of total cross section discharge Q can be represented as:
where a represents the area of an individual segment (m^2), and v is the mean velocity through that segment (m s^-1).
The spreadsheet computation allows the relative importance of each vertical to be quickly seen in tabular form, as a percentage of the total discharge. This is useful, because the discharge computed for any one segment should not be greater than 10 percent of the total for the cross section (Rantz and others, 1982). During the 1991 field season, the spacing between verticals was reduced from that used in 1990 to address this problem, and done so again in 1992. Whereas the largest 1990 segments averaged 22 percent, few segments represented greater than 15 percent in 1991, or eight to nine percent in 1992.
The relationship between total cross-sectional area and discharge, for all measurements at the gaging station site, was remarkably consistent over all three years. The outliers to the plotted points in all cases were early season measurements, when the channel floor and/or walls was ice or snow. Ice and snow reduced the hydraulic roughness of the channel, causing these points to plot above the others. Although the consistency of three years measurements should not be unexpected, given that the same cross section was used all three years, the plot gives confidence that the smaller number of verticals used in 1990 seems not to have substantially decreased overall accuracy.
To allow accurate relationships between stage and discharge to be developed without extrapolation, it is desirable that discharge measurements are made over the full range of gage heights measured during the season. Histograms constructed for each year indicated that measurements were skewed toward lower discharges, but spanned the full range of gage heights.
Discharge measurement was also required throughout the field season to account for changes in stream channel geometry, and consequently, the stage-discharge relationship. These changes occurred throughout the period, due to early season channel constraint by snow and ice, as well as aggradation and degradation of the channel. Decisions to measure discharge were subjectively made in the field each day, based on perceived changes in the channel and the need to make measurements over the full range of discharge. The distribution of discharge measurement through each field period was intentionally skewed toward the early season, when the channel changed most rapidly. Measurement continued throughout the season, however, allowing rating curve adjustments to reflect more subtle changes in the channel.
The objective of sampling was to provide a representative measure of suspended sediment concentration at the station, in both a spatial and temporal sense. As in any sampling design, choices had to be made between the desired accuracy of this representation, and the actual amount of sampling possible. (IMAGE: suspended sediment sampling)
Spatially, the sampling strategy was required to represent the sediment concentration through the stream cross section, which cannot be assumed to be homogeneous at all discharges. The DH-48 sampler obtains a sediment concentration equal to the average discharge-weighted concentration of the approaching flow, when moved up and down at uniform rates (Edwards and Glysson, 1988), and in so doing accounts for vertical concentration variability. The number of such verticals sampled, and the selection of their location, determined the degree to which lateral variability in sediment concentration was resolved.
When more than one vertical in the cross section was sampled (e.g., most of the 1992 season), the locations were determined by the equal-discharge-increment (EDI) method. This method requires partitioning streamflow into equal discharge increments, based on the distribution of flow across the channel; sampling is then done at the centroid of each increment (Edwards and Glysson, 1988). To illustrate, the calculations shown in the sample discharge calculation TABLE indicate that equal discharge occurs to the right and left of station 5.1; if two EDI samples were to be taken, the right centroid would be centered at station 4.25 (25 percent), while the left centroid would be centered at station 5.9 (75 percent).
The distribution of streamflow across the channel varied with discharge, as the wetted perimeter and velocity field changed. As a result, optimal sampling accuracy by the EDI method required that discharge be measured, and the centroids of each increment calculated, immediately prior to each sampling time. This procedure was impractical at Taconite Inlet, and sampling locations were determined from calculations over a range of discharges. Fortunately, the flow distribution at the Taconite Inlet sampling site was quite stable warranting this approximation.
The spatial considerations discussed above apply to the sampling of SSC at one point in time. The frequency of sampling determines the degree to which variability of SSC through time is represented, and determinations of sediment flux must account for this variability as well. The frequency of suspended sediment sampling, in a remote location with limited personnel resources, must be a subjective determination based on the observed variability of SSC. At Taconite Inlet visual observations were capable of detecting small changes in SSC, and indicated a need to sample (i.e., changes in SSC of 5-10 mg L-1, determined ex post facto, were recorded in field books as noticeable). In general, the variability of SSC was considerably greater early in each season than after the snowmelt discharge peak, which the sampling frequency reflects.
During the first part of the 1990 season, one sample was taken at each sampling time, from a midstream vertical. During the second half of the season, except at low flow, two verticals 1 m apart were generally sampled, near midstream. Overall, 158 samples were taken from the stream, at 90 different sampling times.
Suspended sediment concentrations were low during the 1991 field season, relative to those observed in 1990. During periods when the suspended sediment concentration appeared very low (i.e. clear water), only one vertical was sampled per time period. However, when the concentration increased, the number of verticals sampled increased to two or three. Overall, 248 samples were taken, which represent 166 different sampling times.
The strategy for the 1992 field season was to increase both the number of verticals sampled at each time, and the frequency of sampling. Therefore, three verticals were sampled at each sampling time, each representing one third of the discharge (using the EDI method). Sample collection was attempted at a sufficiently high frequency that linear interpolation could be used to define the SSC between sampling times, for the purpose of producing an hourly time series. Sampling was done most frequently during times of the diurnal cycle when discharge was changing rapidly, and early in the season (half-hourly to hourly). The minimum sampling interval was 8 hours, during the most stable periods of discharge and SSC. When SSC was uniformly very low (at concentrations approaching the errors in measurement, approximately 5 mg L^-1), only one vertical was sampled each time. A total of 868 stream samples were taken, representing 333 different sampling times.
Samples of streamwater and suspended sediment were processed in the field for transport and later analysis. Vacuum filtration was used to concentrate the suspended sediments, which were then air dried and packaged. The filtrate volume from each sample was measured, then either discarded or collected for isotopic analysis. The section below details the equipment and methods used to obtain dry samples of the suspended sediments. Field methods were required to balance filtration speed against minimum particle size retention, and subject to constraints on the type and amount of equipment brought into the field.
The vacuum filtration equipment was simple and well designed, and the system was highly successful. Samples were poured into a 1 L capacity Gelman stainless steel parabola filter funnel (model 4230), which securely incorporates a horizontal 47 mm filter on a stainless steel wire mesh. The funnel was fixed on a glass vacuum flask by a rubber stopper, and the flask received the filtrate. Vacuum pressure was applied by a variety of electrical and manual devices, the most successful of which was a hand pump marketed by Nalge Co. as model 6131-0020 (Neward Enterprises, Mit Y Vac model 1G57). The fluid sample volume was measured prior to filtration in 1990 in the calibrated sampling bottle, and measured as filtrate volume in 1991 and 1992 using a 500 mL graduated cylinder. Dried filters, with sediments, were packaged for transport in 3 * 5" polyethylene bags.
The type and size retention characteristics of filters selected to process suspended sediment samples depends on many factors. The choice is particularly difficult without knowledge of the concentration and size distribution range to be measured. Balancing the advantages and disadvantages of different filter characteristics is required, and compromises are inevitable. During the three field seasons, several different filter types were used.
Without prior knowledge of sediment transport at Taconite Inlet, the primary filter planned for the 1990 season was one that retained all particles larger than 0.1 µ m. These turned out to be too fine, in that filtration times per sample were on the order of hours. The alternative that year was to hand-cut 47 mm discs from 11 cm circles of Whatman 541 filters, which greatly reduced filtration time, yet allowed passage of sediment finer than medium-coarse silt (20 to 25µ m). In 1991, two types of filters were used, with comparable size retention characteristics (0.45µ m andf 1.5µ m). The fiberglass filters were very fast, yet retention of sediments within their three dimensional structure precluded sediment recovery later. In 1992 the filters were consistently the same through the field season (0.45µ m). Whatman WCN membrane filters were reasonably fast for the concentrations measured (i.e. less than five minutes were generally required). Subsequent sediment recovery is possible using WCN filters.
The basic field processing procedure is implicit in the discussion above, yet a few additional details are noteworthy. In 1990 and 1991, samples were transferred from the sampling bottle to 1 L storage bottles prior to filtration. Sufficient sampling bottles were acquired in 1992 to avoid this extra step. In all years, bottles containing the samples were rinsed following transfer to the filtration funnel, except on occasions of very low SSC, using either filtrate or water from the sample itself. The primary concern, or difficulty, was removal of coarse material from the bottles (e.g. 0.5 to 4 mm).
Following filtration, filters and sediments were air dried overnight on a drying rack. During 1990, filters were oven dried at 100°F before and after filtration, because weighing took place in camp, on an electronic balance. All dried, filtered samples were ultimately placed in labeled plastic bags for transport and storage until further processing.