Methods
Sites used
Three sites were selected in the upper part of the Styx River to monitor river flows, velocities, depths, turbidity, and suspended solids (Figure 1). Where possible, all sites were situated on relatively straight reaches, although the river was too sinuous at Site C to enable a long, straight reach to be used. Site locations were located using a handheld GPS receiver, and marked by sticks pushed into the soil on each side of the river. Site A represents conditions present in the upper part of the spawning range used by trout in 1990 and again in 2000. Site B represents the reach that was the primary spawning area in 1990, but now supports only a few non-viable redds. Site C represents an unsuitable trout spawning reach. This site was chosen in part because water depths further downstream were too deep to use chest waders in high flow conditions, and also because it was below a discharge source from a nearby urban development. No records suggest that any significant spawning activity has occurred previously in this reach (1 redd in 1990, 0 redds 2000). Accordingly, data for this site should provide an indication of unfavourable sedimentation conditions. However, it is possible that this site once provided suitable habitat for spawning trout and that sedimentation has subsequently buried the underlying gravel substrate.
Site monitoring
All three sites were monitored on seven occasions between the 27th of July and the 1st of September, 2000.
River width at each site was measured using a tape measure stretched between two sticks on opposite sides of the river. Along each meter of the stream width, water depth was measured using a telescoping depth staff with graduated markings every centimeter. Water velocity was measured midway between each depth-reading, using a Global Flow probe (FP101/FP201) which features a digital, true-averaging function. Velocities were measured by moving the probe smoothly up and down in a vertical profile until the average velocity reading stabilised (30-40 seconds). Total stream flow was calculated using the area-velocity integration method.
Water samples were taken from each third of the river channel and turbidity measured on-site using an Orbeco-Hellige portable turbidity meter (model 966). A 2L water sample was also taken (c. one third from each third of the stream channel) to determine the concentration of suspended solids. Known volumes of water samples were filtered through previously weighed, dehydrated Whatman Glass Fibre filters (GF/C) and oven-dried at 110 °C for a minimum of 2 hours before re-weighing. The difference in mass (g) was divided by the volume of sample used (L) and multiplied by 1000 to convert to mgL-1. The filters with dried sediment were subsequently fired in a kiln at 550 °C for 24 hours and weighed again. The mass difference represented the amount of volatile, or organic, material present in the sediment.
River flow data
Daily mean flow data for the Styx River at Radcliffe Road, between the 1st of January, 1995, and the 13th of November, 1999, were obtained from Environment Canterbury. Some gaps existed in the data but 1730 daily readings were used to construct an annual frequency-flow hydrograph for the Styx River. Analogous flow data was produced for the three monitored sites by assuming a linear relationship exists between flows at the three sites, and flow at Radcliffe Road. This relationship was based on the mean of the ratios obtained by measuring flow at each of the sites, and dividing by the flow recorded at Radcliffe Road on the same day.
Sediment deposition rates
As sediment deposition in trout redds is likely to be a function of both water velocity and suspended solid concentrations, an experiment to measure sediment accumulation in clean gravels at different velocities was set up 20m upstream from Site B (Plate 1). This location was chosen for two reasons. Firstly, because it is in the spawning reach abandoned (for the most part) since the 1990 redd survey. Secondly, because a non-viable (anoxic) trout redd was present at the site. This had recently been excavated, enabling the sediment traps to be embedded in the substrate more easily.
Instream baffles and channels were used to locally alter water velocities to enable all traps to be placed within a small area (<3 m2). Manipulating velocities and keeping all traps in close proximity to each other was done to minimise the potential effects of confounding variables such as localised variability in stream sediment concentrations. Initially, the position of the velocity treatments was decided randomly using random number tables. However, because of a severe flood event (on the 18th of August) that washed away instream apparatus, the experiment had to be re-done. Because some stream baffles were damaged or lost in the flood, the fastest velocity treatment was relocated to where natural water flow was already at the desired velocity. The relocated treatment was less than 1m from the original location.
It was assumed that minor amounts of sediment would be deposited at high water velocities. Accordingly, the velocity treatments (0, 0.2, 0.4, 0.6, & 0.8 m.s-1) were targeted towards the middle and lower range of velocities experienced at the three sites under normal or low flows. Unfortunately, because of the first experiment washout, and the length of time taken to process the samples, a direct investigation of accumulation rates at higher velocities could not be attempted.
Three replicate sediment traps were used at each velocity treatment. These traps consisted of 2L plastic ice cream containers (0.2 x 0.2 x 0.1m) filled with clean, twice-washed river gravels. Use of gravel-filled, solid-walled containers has been previously used to examine sediment accumulation rates affecting salmonid redds in the USA (Lisle, 1989; Lisle & Eads, 1991; Rhodes & Purser, 1996; Rhodes et al., 1999). However, records of this method to examine sediment accumulation in spawning gravels in New Zealand could not be found.
Plate 1. Site of the experiment to measure velocity-dependent, sediment deposition rates.
Gravels used in the traps were obtained from exposed gravel bars on the floodplain of the nearby Waimakariri River. Small stones (<6mm) and sediment were removed by washing the gravel in a brass-wire sieve at the river. Washed gravels were taken back to the laboratory and re-washed with clean tap water before use in the traps. Before gravels from sediment traps were reused, they were cleaned and washed again using tap water (after the first experiment was washed-out).
Biofilm-free, dry gravels were used because it would not have been possible to ensure that the amount, and extent, of biofilms on river-sourced stones would be the same between traps. Biofilms can contribute to the accrual of silt on the surface of stones because fine sediment adheres to the sticky surface (Jowett & Biggs, 1997). Unfortunately, dry gravel was not available from the Styx River catchment, and instream gravels tended to be covered in films of periphyton. Gravel from the Waimakariri floodplain was used because it was similar in size to gravel in the excavated redd, and because it is from the same geological formation.
With lids in place, sediment traps were placed into shallow trenches in the river bed deep enough that the surface of the traps was level with the surrounding substrate. Where substrate irregularities left any part of the trap surface elevated, spare gravel was used to fill in the surface depression around the trap. This was done to prevent differences in flow type (i.e. laminar versus turbulent flow) between traps that might confound the results. Water velocities and depth at each trap were recorded using the same methods as for the site monitoring exercise, and piece-meal water samples were taken from inside the experimental area to make up a volume of 2L. Suspended solids concentrations were obtained via the same method used for the site monitoring exercise. Finally, the lids of the sediment traps were carefully removed, beginning with the upstream treatments so that any sediment disturbed from the riverbed would not be able to enter the downstream traps during this process.
After 24 hours the lids were replaced, beginning with the downstream treatments, and water velocity and depth were measured once again at each trap. Another 2L water sample was collected to measure suspended solid concentrations at the conclusion of the experiment. The mean water velocity for each trap, and the mean suspended solid concentration, over the 24 hour period was used for all subsequent analysis.
The sediment traps were returned to the laboratory and the gravels carefully washed in the brass-wire sieve, so that accumulated sediments were collected in a clean, plastic bucket. Some of the heavier sediments, and a little water, were left in the plastic container to save extra filtering time. Wash water was filtered through 120 mm diameter, Whatman glass-fibre (GF/C) filters. Used-filters were placed back inside the icecream container with the heavier sediments that remained after careful removal of the gravel, and oven dried at 110 °C for 48 hours. The dried sample was then weighed, and the amount of sediment in each sample was calculated by deducting the mean weight of 3 identical icecream containers, and also the mean weight of individual glass-fibre filters (multiplied by the number of filters in each sample).
Dried sediments from each trap were thoroughly mixed and 10g sub-samples taken and fired in a kiln at 550 °C for 24 hours. The mass difference after firing was used to calculate the organic (volatile) percentage in each sediment sample to determine the likelihood that biochemical oxygen demand in sediment may differ between sites.
Statistical analysis
Site-level differences for each parameter were analysed with a single-factor analysis of variance (ANOVA). Regression analysis was used to determine relationships between suspended solids and flow, and an ANOVA was used to determine how significant the R2 value for the relationship was. Prior to analysis, data was transformed using the natural logarithm because variation tended to increase with increasing flows. All analyses were performed with the statistical package within Microsoft Excel 7.0.
Sediment deposition modeling
Data obtained during site monitoring was used to generate flow-based relationships for velocity and depth for each meter width of the stream at the three sites. These were used in combination with the annual range of flows, to determine flow-specific velocities and depths for each part of the stream channel. Similarly, flow-specific suspended solid concentrations were predicted for each site using exponential relationships generated by the regression analysis (Figure 2). Because it seemed unlikely that very high concentrations of suspended solids (in excess of 500 mg.L-1) were plausible (given that the worst discharge observed was only 300 mg.L-1), the exponential relationship was modified by the expression 500/(500+Predicted SS) which created a sigmoid curve that capped sediment loading at 500 mg.L-1 at each site.
The number of days assumed to be necessary for egg incubation was chosen arbitrarily from within the range 70-115 days provided by Maret et al. (1993). The 76 day incubation period was used because it provides a good scenario for trout recruitment, and consequently should provide a conservative bias towards sedimentation predictions.
To find the frequency and magnitude of flows at each site during the incubation period, the following steps were taken. The number of days for each flow increment in the annual hydrograph (at Radcliffe Road) was divided by the total number of days in the year. This proportion was multiplied by the incubation period (76 days), and the result rounded to the nearest day. The resultant flows were then corrected to allow for the increase in water flow between each site and Radcliffe Road.
For each meter width of each site, daily sediment accumulation for each of the expected flows was calculated by using the velocity-dependant deposition rates measured in the experiment, and the combination of predicted depths and velocities (to derive daily volume of water in each meter width) and the suspended solid concentrations at each site.
Incubation simulations
A range of exponential declines in sediment accumulation rates was used to model the effect of armouring during the incubation period. Twenty one simulated incubations periods were run for each armouring rate trialled, and the mean value from the 21 runs was used as the final estimate of sedimentation at each site.
The order in which expected flows would occur during a hypothetical incubation period was randomly generated using a spreadsheet. Sedimentation rates were modified by the relevant armouring factor for that day, and the sum of the daily sediment accumulations during the incubation period was calculated. Multiple runs were deemed necessary to avoid atypical flow regimes biasing the results.
The entire process was repeated using organic deposited solids instead of total deposited solids. This was possible by using the relationship between organic sediment percentage and velocity, obtained during the experiment, and factoring this into the daily sediment deposition estimates.