Title Elizabeth Thornton
Chapter Five
Discussion
5.1 Introduction
The main objective of this project was to identify whether channel morphology scales up with increasing stream scale within a network in a similar pattern as the way in which channel morphology is seen to change in the downstream direction along a single river. In addition, the implications of coarse bed material for channel morphology were investigated to determine the way in which sediment is influencing the channels’ ability to adjust to alterations in discharge. These objectives were achieved through the investigation ofseveral scale dependent river morphology variables. This chapter discusses the results of the research undertaken and provides a comparison of the findings with previous findings from other studies.
5.2Downstream (upscale) spatial variability
5.2.1 KangarooValley Network
The study was undertaken in a tectonically stable region with a temperate climate. Despite the temperate climate, the region is prone to variable rainfall, a good example of which was the lack of storm events during the study period which, resulted in low flows. The valley is characterised by steep valley sides and cliff lines prone to landslides, geologies vary between the north and south boundaries of the valley. The topography of the valley suggests a highly flashy hydrograph, which would mean a large flow during a storm event. Although the sites are fully alluvial, they are close to bedrock and clifflines resulting in the presence of coarse bed material. The soils at the sites within the KangarooValley are either Sandy Loams or Loamy Sands.
The different scales of the study reaches were established from the catchment areas for each site, as well as the stream lengths. Catchment area doubles from DGC to SC1 and then againfrom SC1 to SC2. However, between SC2 and KR1 there is an increase of approximately six times the catchment area. Catchment area then approximately doubles between KR1 and KR2. The stream network analysis showed that despite an increase of approximately double the catchment size SC1 and SC2 are of the same stream order. This indicates a clear difference in scale between these two sites despite the same stream order classification and that perhaps stream order classification is an unreliable mechanism for delineating scale.
5.2.2 Up-scaled network hydraulic relationships
Hydraulic geometry relationships were computed for w, d and vat lowflow, bankfull and dimensionless bankfull discharges. In this study,the derived hydraulic geometry exponents suggest a significant correlation exists for all relationships computed, with the exception of bankfull v against bankull Q. As this study represents the first effort to describe the way in which hydraulic geometry relations behave for a stream network it is important to consider how they compare with the established relations for downstream variations in channel morphology.
Since the concept of hydraulic geometry was first proposed by Leopold and Maddock (1953) an interesting regularity of exponents with increasing discharge downstream has been observed, where the typical exponents are 0.5, 0.4 and 0.1 for b (width), f (depth) and m (velocity), respectively (Andrews, 1984). Where the exponents of b, f and m varyfrom these typical values, they usually retain the condition of width increasing faster than depth, which increases faster than velocity in the downstream direction(Fergurson, 1986; Leopold and Maddock, 1953; Church, 1992; Knighton, 1998). Andrews (1984) compared Colorado gravel bed rivers to British gravel bed rivers of similar nature and demonstrated no significant differences in the exponents of the hydraulic geometry equations. This reinforces the notion that the exponents for hydraulic geometry are commonly independent of locality.
The results for bankfull Q in the KangarooRiver system indicate that the exponent for width is only slightly larger than that for depth, suggesting that these two variables increase at similar rates across the network under the conditions of bankfull flow. When compared to the exponents provided by Leopold and Maddock (1953) (Table 5.1) this river system increases its width slower and its depth faster than “typical” rivers. It was found that when vbf was related to Qbf the trend was not significant. The lack of significance suggests that at bankfull flows velocity values exist irrespective of scale and are products of the local (reach) conditions.
As lowflow discharge is not considered a channel forming discharge, relationships from other river systems have not been derived, and thus there is no means for comparison of the results reported here. However, it is interesting to observe that under low flow Q, the exponents of the hydraulic geometry relationships do not follow the usual pattern of b>f>m. Rather, for this flow condition, downstream width increases the fastest followed by velocity and depth and the exponent for velocity is relatively high in comparison to that suggest byLeopold and Maddock (1953). This suggests that at low flow conditions, flow velocity responds more dramatically to network variations in discharge than it does under bankfull conditions.
To allow for comparisons between the behaviour of the KangarooRivernetwork and downstream behaviour of rivers studied by Parker in Canada, the UK and the USA (Pitlick and Cress, 2002) (Table 5.1) Q*, B*, H* and swere computed. The exponents of the hydraulic geometry equations for theKangarooValley network show some similarity but also a marked difference when compared to downstream trends as presented by Parker. The exponent b for the KangarooValley network is similar to that presented by Parker for Northern Hemisphere gravel bed rivers, suggesting that these channels adjust their width in similar manner when responding to variations in discharge. Despite the similarity for changes in downstream w, an interesting contrast is demonstrated between the two sets of equations. The exponent f in the relationships computed for gravel bed rivers by Parker is much higher than f computed for the KangarooValley river network. This implies that channels of this study adjust their depth to alterations in the discharge regime at a much lower rate than is observed in “typical” gravel bed rivers. However, b>f for relationships with Q* and thus the ratio of adjustment is consistent with patterns of adjustment for “typical” rivers as described above. The relationship between s and Q* was not significant. Thus, unlike rivers studied by Parker, the bedslope of the channels within the Kangaroo Valley network is controlled by a process other than Q*, which is occurring independent of scale.
Table 5. 1: Downstream hydraulic geometry relationships for previously studied rivers.
Past research / Equation / Coefficient / ExponentLeopold and Maddock (1953) / w=aQb / na / 0.5
d=cQf / na / 0.4
v=kQm / na / 0.1
Parker (In Pitlick and Cress, 2002) / B*=aQ*b / 4.9 / 0.46
H*=cQ*f / 0.37 / 0.41
s=kQ*m / 0.098 / -0.34
Downstream hydraulic geometry relationships were computed using catchment area as a surrogate for discharge. Given the uncertainties involved in determining the bankfull channel and estimating bankfull discharge, as discussed below, the possibility of substituting catchment area for bankfull discharge in hydraulic geometry relationships may allow for a more reliable scaling variable. It has been suggested that in homogeneous geographical contexts that this would be the case (Petit and Pauquet, ). However, despite similar exponents for the d and w relationships when comparing power relations based on catchment area and those on bankfull discharges (Table 4.# p in results somewhere) the exponents for both vlf and vbf are abnormal. Thevery high exponent for the vlfrelationshipwith catchment area and the negative exponent for the relationship between vbf and catchment area suggest that this is not a suitable surrogate for Q, at least not for this system.
5.2.3 Channel Shape
A trend analysis was performed using a Mann-Kendall test to identify upscaling trends between sites for cross-sectional form (F) and absolute asymmetry (A*). No statistically significant trend was observed between sites for For absolute A*, indicating that there is no trend in which these morphologic features scale up within the network.
Interestingly, when F was plotted against Q, there was not an easily defined increase. This suggests that F is constant in this network irrespective of scale. The manner in which these values present themselves (Figure #. Page #) shows how F remains constant between the four smaller reaches, DGC, SC1, SC2, and KR1, and then increases for KR2. this suggests that between the scales of KR1 and KR2 a threshold is crossed where by discharge becomes sufficiently large to mobilise sediment and control the morphology of the channel. However, for the four smaller reaches, the scaling variable Q does not control F suggesting that other processes are the dominant channel forming variables at these scales.
The low A* values exhibited by all of the channels in this study indicate a high level of symmetry for the cross-sectional shape of the reaches. The lack of a discernable difference between reaches is surprising given the complex nature of the cross-sections (See Appendix I for individual channel cross-section diagrams). In addition, the banks at each site show differences in shape, within and between sites, and it follows that as a result the banks would erode at different rates according to their shape, thus leading to a more asymmetrical channel. However, this is not the case for this river network. The fact that no significant trend exists across the network for A* indicates that asymmetry is not a function of differences in scale, rather that it exists independent of scale.
5.2.4Sediment
Sediment particle size information was used to determine whether the traditional pattern of downstream fining exists in this system. A clear trend between reaches for sediment particle size is not evident in the data collected in this study. This is possibly due to the small sample size, although field observations suggest some visual evidence for sediment fining as the scale of channel reached increases. There are three possible explanations for the weak trend in downstream (upscale) fining observed in the KangarooRiver network. In a previous study undertaken by Pitlick and Cress (2002), weak downstream fining relationships were attributed to a high rate of coarse material supply from tributaries, adjacent hillslopes and terraces which was sufficient to replenish material worn by abrasion (Pitlick and Cress, 2002). In addition, the variable source material ie sandstones for KR1 and KR2 and Gerringong volcanics for SC1, SC2 and DGC, may impact upon the size of materials as the resistance of rocks to abrasion is dependent on its geology. Although there has not been a demonstrated trend of downstream fining, it has certainly been shown that as one progresses down the system, the presence of local sorting mechanisms become more apparent as pools are seen to contain increasingly finer material with respect to their adjacent riffles as one moves through the system.
A second possible explanation for the lack of a increased fining as channel size increase the close proximity of the study sites. It is quite possible that a stronger trend would have been identified if sites were spaced at greater distances apart. Finally, because the sites are closely spaced the material in the KangarooRiver channels is probably sourced from adjacent floodplain deposits with similar sediment characteristics.
5.2.5 Synthesis
One hypothesis in this study was that the downstream relationships for channel morphology in cobble bed streams differ from those in gravel bed streams due to the controlling nature of the larger sediment on the rivers ability to self adjust to varying discharges. Despite the fully alluvial (i.e., unconstrained by bedrock) nature of the study sites, it is expected that the large sediment in the system is controlling bedform and cross sectional morphologies. The alternative hypothesis is that the streams in this network are free to adjust to changing discharges, which is typical of truly alluvial gravel bed streams.
As the discussion of spatial variability of sediment properties within the network, above, portrays the channels of this network have relatively sandy banks and coarse bed material. The coarse bed material is likely to produce a high level of resistance to flows. This is possibly why the system exhibits such a slow response to discharge with respect to adjusting channel depth. Highlighting the importance of the coarse sediment in this system to aspects of channel form adjustment. Thus, there are certainly parameters of channel morphology in this network that appear to be influenced by large sediment.
The analyses of downstream hydraulic relationships computed in this study for the KangarooValley stream network above indicate that channels in the KangarooValley are able to adjust their width to accommodate changes in discharge faster than they are able to change their depth. This pattern is typical of downstream hydraulic relationships for individual rivers (Fergurson, 1986; Leopold and Maddock, 1953; Church, 1992; Knighton, 1998). Thus, the most interesting finding of this portion of the study is that there is a similarity between downstream and network hydraulic geometry relations. This indicates some self-similarity between river network branches, in that irrespective of which tributary is investigated with respect to the main channel, a similar result is derived. This finding may not be universally applicable in river networks with highly complex geologies, vegetation covers and landuses where peculiarities in individual network links may complicate these relations.
5.3 Spatial variability- Local
5.3.1 Bedform morphology
Reach averaged bedform lengths and elevations were related to the scale variables of Qbf, catchment area, reach averaged d and reach averaged w to determine whether bedform morphology scales up with the network. The p values indicate that there is a stronger relationship between reach averaged bedform length and scale variables than reach averaged bedform elevation and scale variables (Table 4.5). Indeed, mean bedform length showed a statistically significant relationship at the 0.10 level with all of the scale parameters where as mean bedform elevation was statistically related to only reach averaged w at the 0.10 significance level. However, mean bedform length and elevation are not constant proportions of reach averaged bankfull width and depth (Table 4.6). In addition, the spacing of bedforms are not constant proportions of bankfull width and depth, suggesting that, using these longitudinal profiles as a reference, it is evident that at the reach scale pool and riffle spacing in the KangarooRiver system is irregular. Based on these outcomes it is possible to state that the spacing of pools and riffles in this system does not exhibit the dependency on scale that has been characterised for other rivers.
Bedform elevations and lengths were both significantly related to the scale variables, at least to some extent. This result implies that for the KangarooRiver system bedform length and elevation exhibit some degree of scale dependency, especially bedform length. However, bedform elevation and length do not increase proportionally with reach averaged bankfull width and depth. This means that it is not possible to estimate bedform length or height based on bankfull channel width or depth.
5.3.2Channel shape
The results of this study have demonstrated a significant differenceF between pools and riffles at all sites except KR2 where it was difficult to survey cross-sections in the deep pools. Mean F values for the pools and riffles indicate that pools tend to be deeper and narrower than riffles for all sites, regardless of scale. This result is consistent with previous research on pools and riffles (Richards, 1976;Knighton, 1998). The pattern within reaches here suggests that F is locally variable and controlled by the bedforms of the stream.
Given the results of local variability for F, it is not surprising that when pools and riffles were grouped for all sites there was a significant difference between the F values of the two populations. This indicates that in this system there is a clear difference between F for these two bedform types.
A comparison of A*for pools and riffles indicated that, in the KangarooRiver system, pools and riffles were not statistically different. Past research has suggested that pools tend to be more asymmetric than riffles, often due to the presence of lateral bars (Knighton, 1998). The streams studied in this research appear to be an exception to this, as the majority of pools do not have a lateral bar (Figure).In particular, the large pools from the three downstream sites occupy the entire channel in the lateral plane. Milne (1983) identified that A* values for pools and riffles tend to be greater in reaches located on meander bends than in straight reaches. Thus, as high A* is linked to meander bends, low A* values in this network probably result from the low sinuosities of the chosen study reaches.
5.3.3 Velocity
Under low flow conditions, mean velocity was higher across riffles than pools at all study sites. This reflects the fact that, at low flow, water surface slope is relatively high over riffles and gentle across pools (Knighton, 1998). As discharge increases, however, slope (and hence velocity) differences between pools and riffles decline (Robert, 1997). Thus, the greatest difference of velocities between pools and riffles exists at low flow, which explains the large variation of lowflow measurements, between bedform features, that was observed in this study. When vbf values are plotted, the reverse is shown, as at bankfull level velocities are higher across pools than riffles. Thus, the reaches in this network behave in a manner that supports the theory of velocity reversal.