Discussion
Limitations of the model
Any exercise in modeling the real world is based on the assumption that physical processes can be described using mathematical equations. Consequently, when interpreting the results of mathematical models it is necessary to be aware of the limitations of the model, and any assumptions made.
Data Issues
The raw data used in this study allowed good, flow-based relationships to be determined for suspended solid concentrations, velocity, depth, and turbidity. However, there was insufficient data to determine whether between-site differences were statistically real (Table I). This is probably because the samples were obtained across a wide range of flow conditions. For example, flow was always higher downstream than upstream and this is clearly a real difference. However, data from this study showed that while flow increased downstream (Table I) there was no statistical difference between sites at the 95% confidence level. This was because the difference in flow between the three sites, on any given day, was small compared to the variation in river flows between sampling dates. To statistically detect the real difference in flow between the three sites either more samples would need to be taken, or else stratified sampling (ensuring similar river flows on each sampling date) could be done.
Likewise, the lack of a statistical difference between overall suspended solid concentrations at the three sites, or the response of suspended solids to flow at the three sites, probably doesn’t reflect real physical differences between the sites. The inability to detect differences statistically may have been resolved by increasing the number of samples but logistical constraints on time precluded this. It should also be noted that Hicks & Duncan (1993) found that suspended solid concentrations in the middle reaches of the Styx River partly depended on whether the river flow was increasing or decreasing. This factor was not incorporated into the present study and, potentially, represents a source of variation that makes it hard to detect real differences. Different relationships between suspended solids and flow were used at each site because the observed differences between sites were considered to be real. While this cannot be supported on a statistical basis using the data obtained in this study, large differences in adjacent land-use, riparian conditions, and discharges exist between the three sites. Since there are strong linkages between land-use, riparian conditions, discharges, and suspended solids in affected waterways (Ryan, 1991), it is very likely that real differences exist at the three sites (pers. obs).
Deposition rates in the manipulative experiment should be regarded with some caution. The lack of a true control for the effects of weed build-up was a major flaw in the experimental design. In this regard, the flood that washed out the first experiment may have been a blessing in disguise. Because two of the baffles were lost, the traps in the highest water velocity treatment were relocated to an area where ambient velocities were within the desired range. The change in position was relatively minor (<1m) and consequently this treatment differed from the rest only by the lack of weed-trapping baffles. By projecting trends in the data from the other treatments (to predict the results for weed-affected traps at the fastest treatment), it was possible to correct for the effects of weed accumulation (Fig. 5). However, it should be noted that this is not as reliable as using real controls to determine the extent of weed effects on sediment accumulation in the affected traps. For instance, the amount of sediment filtered, and the rate of weed accumulation on the baffles may be a function of water velocity. This simple approach to estimating the effects of weed build-up was preferred because it is the least complex, and also because there was no information available on which to base more complex solutions. Similar plumes of dislodged sediment were observed when removing weed from baffles at all treatments and, based on this subjective impression, it seems unlikely that weed-effects would be overly responsive to changes in velocity.
The sensitivity of the model (to potential error in accumulation rates at lower velocities) is likely to be minimal for sites A and B because most velocities were within the vicinity of the fastest treatment, or higher. However, the model may overestimate sedimentation at Site C by a factor of (up to) 3.9 in the worst case scenario, because velocities at this site tend to be lower than at Sites A & B. This scenario would result from the unlikely situation where weed-effects on the treatments below 0.8 ms-1 were nil.
The lack of data on the rate at which armouring effects reduce sedimentation rates is the major obstacle to predicting the amount of sediment accumulated at each site. Unfortunately, no attempt was made to measure this process during the field-component of this study and consequently the specific aim to estimate total sediment inputs cannot be made with confidence.
Deriving an estimate for armouring effects from scientific literature was, unfortunately, not directly possible. Lisle (1989) reported that sediment accrual in redds was almost as high midway through incubation, as it was at the end of the incubation period. This is probably due to gravel spaces clogging and preventing further sediment intrusion into redds. Consequently, sediment deposition rates in recently constructed redds are initially high and then decline through time. The most likely scenario to mathematically describe this pattern is an exponential decline similar to that shown in figure 10. To match Lisel’s observations, and assuming incubation periods and armouring rates are similar, an exponential decline of e-0.1 would result in a deposition rates declining to c. 5% of the initial rate, midway through the incubation period.
However, this estimate is entirely theoretical and further work to determine actual changes in deposition rates through time is necessary to provide a better prediction.
Assumptions Made
The model assumes that sedimentation rates are directly proportional to the amount of sediment in the water column and also directly proportional to the depth of the water column (Hicks and Griffiths, 1992). The range of depths generated by the model varied only over a small range, and consequently this assumption is likely to have only a small effect on the predictions generated. Suspended solid concentrations varied over a much larger range than depths. Consequently, the model is likely to be more sensitive to the assumption that sedimentation rates are directly proportional to the suspended load.
Flows used in the model were determined using the annual probability of flows in the frequency-flow hydrograph for the Styx River (Fig. 4). Seasonal factors that might influence flow probability (e.g., winter floods) during the incubation period were not included.
Conditions that existed during data collection were assumed to be representative of conditions at the sites throughout the incubation period. Also, the relationship between deposition rates and water velocity at the three sites was assumed to be the same as that measured at the experimental site. Deposition rates occurring at particular velocities might differ between the sites if the sediment was very different at the three sites. However, the lack of obvious differences in organic content (Fig. 3) between the three sites supports the assumption that velocity-dependent deposition rates are comparable between the sites.
Steady state conditions are assumed to exist within each of the hypothetical days during incubation. As a result, the model is not sensitive to changes that occur over a short time scale (<24 hours). Similarly, the model makes no attempt to predict sedimentation patterns over a small spatial scale (<1m2). Consequently, any tendency for redds to distribute sediment spatially because of small-scale physical differences are not accounted for. This level of detail would be required for more deterministic work that attempts to model dissolved oxygen levels, or the distribution of sediment within redds.
Factors Included
The computer model specifically incorporates the effects of suspended solid concentrations, water velocity, water depth, and velocity-dependent deposition rates. It also incorporates possible changes in deposition rates during the incubation period of trout eggs. It was not considered necessary to control the particle sizes of sediment or gravel because the data was derived using sediment particles and gravels that are present in the river. Consequently, the model is only applicable to the Styx River. If significant changes in the size distribution, or nature, of the suspended solids or gravels were to occur, then deposition rates would have to be reevaluated.
Model Predictions
The model developed in this study is designed to predict sediment accumulation within recently cleaned gravels, at each of the three study sites. These predictions exclude sediment inputs from surrounding substrates, and this can be important in some situations. For example Lisle & Eads (1991) reported that bed-load migration accounted for three quarters of the sediment in mountain-stream redds. However, Alonso et al. (1996) reviewed sedimentation processes and concluded that bed-load transport of sediment is only important for high-gradient, high-energy rivers with mobile beds. Low-gradient rivers, such as the Styx, are more likely to be characterised by immobile beds and the bed-load migration of sediment in these rivers is considered to be negligible. However, bedload migration may become significant during particularly high flows that are capable of mobilising the gravel framework of the riverbed
The model does not directly predict the viability of trout redds at the three sites, but instead predicts sedimentation regimes and infers redd viability from them. Sedimentation has important implications for dissolved oxygen levels within redds, but no attempt to model interstitial dissolved oxygen was attempted in this study. A more detailed examination of sediment intrusion into redds, and dissolved oxygen levels, is possible utilising the SIDO computer model (Havis et al., 1993; Alonso et al., 1996). Also, clean gravels can be roughly correlated with recently constructed redds, but there may be differences in sedimentation regimes due to flow characteristics associated with the placement and morphology of redds.
Relative accumulation rates
The model developed in this study predicted that sediment accumulation is lowest at the upstream site, somewhat higher at the middle site (c. 32% more sediment), and a great deal higher at the downstream site (c. 350% more sediment; Fig. 8). This matches the pattern that would be expected if sedimentation rates are responsible for the reduction in trout spawning range. If the tolerance threshold to sediment was close to being exceeded at the upstream site, then 32% more sediment in redds at the middle site could be biologically significant and prevent spawning success.
Organic sediment loading in gravels at Sites A and B was not predicted to be markedly different to each other, but organic sediment in gravels at Site C are predicted to be much higher (Fig. 9). This suggests that redds would be smothered through lack of water penetration into the redd (because of armouring) rather than large differences in biological oxygen demand.
Total accumulation of sediment
While the relative proportions of sediment accumulation at each site are the same regardless of armouring effects, the actual amount of sediment accumulating in gravels at the three sites cannot be predicted without determining how quickly armouring reduces deposition rates.
Using the exponential rate of decline in deposition rates (e-0.1) derived from the observations in Lisle (1989), then the total amount of sediment accumulated at the three sites is predicted to be 38 kg.m-2, 50 kg.m-2, and 171 kg.m-2, respectively, during a 76 day incubation period. However, these values only represent a ‘best guess’, and more data on deposition rates through time is needed.
It is difficult to put these figures into context without knowing how much sediment makes redds non-viable. Lisle (1989) refers to a paper by Reiser & Bjornn (1979) which states that sediment loadings of 40 kg.m-2 begin to have serious adverse effects on egg survival. Unfortunately Reiser & Bjornn’s paper was unavailable to determine the basis for this figure, but no other sources could be found that relate sediment accumulation directly to survival of salmonid eggs and larvae. Assuming that this figure is representative for the viability threshold for brown trout redds, then the results of this study suggest that redds in the upper reach are marginally within tolerance levels, but approaching levels at which redd viability may be threatened. Redds in the middle reach, by contrast, are predicted to exceed the sediment threshold and would be unlikely to produce viable trout larvae. The downstream reach is clearly not suitable for trout spawning (Fig. 8).
Spawning reach shift
Most trout redds in 1990 were in the middle reach (Site B) with some in the upper reach (Site A). It is unlikely that trout would spawn en masse in areas that are unsuitable for spawning, and this suggests that sedimentation rates at the middle site were ‘tolerable’ at that time. However, it is clear that the availability of suitable spawning gravels was already becoming a limiting factor because some trout were found to be reabsorbing eggs instead of spawning (Eldon & Taylor, 1990). In addition, whatever the actual sedimentation rates were in 1990, trout preferred spawning in the middle reach than in the upper reach. This situation suggests that conditions in the upper reach were probably not as good as conditions at the middle reach at that time.
In the present spawning season, only half the total number of trout redds were found compared to the 1990 spawning season. The majority of these redds were located in the upstream reach with very few located in the middle reach (Mark Taylor, unpublished data). Obviously something has changed over the last decade that has caused the trout to shift their main spawning grounds upstream. Either habitat in the middle reach has exceeded some critical threshold, habitat in the upper reach has improved dramatically, or a combination of both of these scenarios has occurred.
During the intervening 10 years, substantial improvements to riparian management in the upstream reach have been implemented through the establishment of the Styx Mill Conservation Reserve. Riparian vegetation has been planted and provides a measure of fish cover, as well as providing more effective riparian sediment buffering. The full benefits of riparian planting, however, will probably not be likely to be felt until the vegetation is better established. Even so, it is clearly feasible that habitat improvements in the reach have increased the attractiveness of this reach to spawning trout. However, this scenario alone does not explain why the middle reach has been almost completely abandoned or why trout are reabsorbing eggs rather than spawning.
The other possibility, that a critical tolerance threshold in the middle reach has been exceeded since 1990, is supported by the interstitial dissolved oxygen concentrations found within redds in this reach. Redds examined in the middle reach are suffering oxygen limitation and suffering massive egg mortality through suffocation (Greg Burrell, unpublished data). It is not likely that this would have been the case in 1990 because most trout preferred spawning in the middle reach instead of further upstream.
If tolerance levels to sedimentation rates in the middle reaches were being approached in 1990, and have subsequently been exceeded, then it would be expected that sedimentation rates at the middle reach would be higher than at the upstream reach, but not enormously so. A very large difference in sedimentation rates would either suggest that sedimentation rates are changing rapidly, or else that the model is deficient.
In light of this reasoning, an increase of approximately 30% in sedimentation rate between the two reaches seems to be quite reasonable. This is particularly true if sedimentation rates at the upstream site are within a range of 10-25% of the tolerance threshold. The bottom-most site in this study is clearly subjected to enormously higher sedimentation rates, and is unlikely to have been a suitable spawning reach during the last few decades.
The Styx River in decline?
There is a growing consensus amongst citizen groups such as Guardians of the Styx, and fisheries workers (e.g. Eldon & Taylor, 1990) that the smothering of gravel substrates has become a problem in the Styx; even though it is hard to empirically prove this. Riverbed agradation is occurring in the middle and lower reaches of the Styx (Hicks & Duncan, 1993), but no work on agradation near areas of gravel substrate has been done. However, it is likely that bed agradation in the middle reaches has reduced the hydraulic gradient of the river, and consequently decreased water velocities immediately upstream of the agrading reach. Reduced velocities result in increased deposition rates, and may relocate agrading zones slightly further upstream. Over a long period of time this may result in soft-bottomed substrates extending upstream and smothering previously gravel-bottomed reaches. The process will continue until a new equilibrium has been reached between the amount of sediment entering the water, and the ability of the river to remove the sediment. Because this process would occur over a long period of time, it would be difficult to measure on a time scale that is usual for most field investigations. If this is the case, then the process of upstream smothering of gravels may go undetected except for the observations and perceptions of people who have had a long term association with the river.
It is difficult to quantify gradual, longitudinal changes in a river without detailed descriptions of the river over a long period of time. Descriptive data from the early part of the century, of sufficient detail to compare with present-day conditions, is unfortunately not available. However, there is some evidence that stream health has been declining since the settlement of Christchurch. Taylor et al. (2000) showed that stream health (represented by UCI scores) was declining between 1979 and 1989. Even though invertebrate diversity in the Styx remains relatively high, there has been a decline in taxonomic richness from the early part of the century (Parrot, 1929; Eldon & Taylor, 1990; Taylor et al., 2000), accompanied by a reduction in the distribution and abundance of some important species (such as freshwater crayfish, freshwater shrimps, etc.) The recent upstream shift in trout spawning areas also suggests that something has adversely affected downstream parts of their previous spawning range.
When taken as a whole, the anecdotal and scientific evidence suggests that gradual sedimentation of the river bed may have been occurring since the settlement of Christchurch. The increased sedimentation has probably resulted from changing land-uses in the catchment.
The recent decline in trout abundance in the Styx River may have resulted from the upper limit of soft-bottomed substrates reaching the main spawning areas. Reductions in hydraulic gradient, and increasing sediment inputs, may have led to the abandonment of the main spawning reach between 1990 and 2000. It is likely that if sediment inputs to the river are not reduced significantly, and soon, then the trout population will continue to decline. If sediment inputs continue to increase, and soft-bottomed substrates continue to extend further upstream, then the trout population may become locally extinct.