Landscape Connectivity in the Upper Mzimvubu River Catchment: An Assessment of Anthropogenic Influences on Sediment Connectivity
Abstract
Connectivity has been altered in many landscapes world-wide and impacts the routing of water and sediment. Diminished ecological infrastructure can result in a reduced ability to retain water and soil that will impede ecosystem health and sustainable use of land and water resources. In the northern Eastern Cape Province of South Africa, the headwaters of the Mzimvubu River, land use change and land degradation over the past 50–100 years has initiated increased sediment connectivity through features such as gullies and incised river channels. The water resources of this catchment are largely undeveloped, but future development will be threatened by high suspended sediment loads. The Vuvu catchment, a headwater tributary of the larger Mzimvubu River system, is used as a case study to assess how hillslope–channel and channel–valley fill sediment connectivity has changed over the past 100 years. From detailed mapping of gullies, roads and livestock tracks, it was concluded that downslope connectivity added 22% and across-slope drainage 159% to the drainage network. As a result of increased drainage efficiency, the main channel was incised and straightened, in turn reducing channel–valley fill connectivity. A conceptual model of changes to sediment connectivity is presented with recommendations for future restoration. Copyright © 2017 John Wiley & Sons, Ltd.
Introduction
Landscape connectivity, which encapsulates sediment and hydrological connectivity, has been altered in many landscapes world-wide, impacting the routing of water and sediment through the landscape. The transformation of both urban and rural land surfaces through anthropogenic features such as roadways, paving, other impervious surfaces, drains, agricultural fields, storm water gutters, reservoirs and weirs (Lenat & Crawford, 1994; Aber et al., 2010; Peñuela et al., 2016; Cavalli et al., 2013; Masselink et al., 2016) has altered the natural landscape processes involved in water and sediment transport, often leading to enhanced hillslope–channel connectivity. These changes to landscape function tend to diminish ecological infrastructure and related ecosystem services (Foley et al., 2005). Weakened ecological infrastructure can result in a reduced ability to retain water and soil on hillslopes and will impede ecosystem health and sustainable use of land and water resources (Foley et al., 2005).
Landscape connectivity has been used as a framework in various geomorphological studies over the past 20 years (e.g. Brierley & Fryirs, 1998; Brierley & Murn, 1997; Cammeraat, 2002; Hooke, 2003; Cavalli et al., 2013; Bracken et al., 2015) as it integrates structure and function over various temporal and spatial scales (Brierley et al., 2006). Landscape connectivity can be defined as the movement of matter through a system of interconnected units (Hooke, 2003). Both natural and anthropogenic features can either enhance or decrease connectivity and thus affect water and sediment movement (Fryirs et al., 2007). Understanding the structure and functioning of landscape connectivity is crucial to the effective management of a catchment (Lexartza-Artza & Wainwright, 2009).
Where: landscape connectivity is hindered by natural barriers, buffers or blankets (for example alluvial fans, floodplains, resistant base level, low gradient slopes, floodplain sediments and armoured channels), matter is stored until landscape connectivity is re-established (Fryirs et al., 2007; Cavalli et al., 2013; Fryirs, 2013). Only a small proportion of sediment that is eroded in an event will make it to the bottom of the slope or catchment (Walling, 1983). Sediment movement is often symbolized as a ‘jerky conveyor belt’ (Ferguson, 1981), moving sediment down a slope or through a catchment in a series of overland flow events rather than one continuous event. This is due to sediment deposition down the slope and is a function of the degree of landscape connectivity within and between the various landscape units. Anthropogenic interventions that decrease connectivity include the construction of dams, levees and elevated roads. While these have socio-economic and biophysical benefits (for example water provision, economic growth), they also have drawbacks such as decreased fish migration and downstream impacts on channel morphology (Nakamura & Tockner, 2004; Sandercock & Hooke, 2011).
Increased connectivity (boosters such as development of gully networks, straightened river channels, sunken lanes, roads, bare soil) leads to more efficient transport and export of materials, such as water, sediment and nutrients (Harvey, 1992; Brierley & Murn, 1997; Brierley & Fryirs, 1998; Harvey, 2002; Brierley et al., 2006; Boardman, 2013; Wemple, 2013; Pechenick et al., 2014; Peñuela et al., 2016; Masselink et al., 2016). The natural capital of upstream areas is degraded, and downstream areas are affected by the increased delivery of flood water and eroded materials, which can have severe biophysical and socioeconomic effects such as flooding, eutrophication and sedimentation (Brierley & Fryirs, 1998; Sandercock & Hooke, 2011). Increased connectivity can lead to channel incision that causes a disconnection between the river and its flood plain, thus leading to a reduction in lateral connectivity between the river and the ecologically important riparian zone (Kondolf, 2006). This can also result from barriers to longitudinal connectivity that reduce flood magnitude, such as an upstream dam. Increased connectivity can lead to reduced resilience of the system to absorb shocks and stresses (Poeppl et al., 2016).
In South Africa, ecological infrastructure is often compromised by increases in landscape connectivity due to hillslope erosion and valley floor disturbance (Rowntree, 2012). Extreme cases have been reported where buffers have been destroyed, such as wetlands that were purposefully drained to expand agricultural activities (e.g. Cowden et al., 2014). More general cases of ploughing on steeper slopes and the subsequent development of features such as sheet erosion, rills and gullies are examples where hillslope–channel connectivity has been increased (Kakembo & Rowntree, 2003). Downstream response of hillslope erosion was investigated by Dollar & Rowntree (1995) who linked channel instability and straightening of the Bell River, Eastern Cape Province, to widespread gully erosion in the catchment. Increased landscape connectivity in the form of sheet, rill and gully erosion is prevalent throughout the Mzimvubu catchment, northern Eastern Cape Province and former Transkei homeland. Increased sediment loads are evident in the estuary as the once navigable lower Mzimvubu River at Port St. Johns is currently a shallow silted up river mouth, barring large vessels from entering the river. The Mzimvubu catchment is currently undammed, but proposals to dam the Tsitsa tributary have been put forward. Future sedimentation of the reservoirs is a serious concern.
Several special issues on connectivity beg for better quantification of connectivity and its interactions between components (e.g. Parsons et al., 2015; Wohl et al., 2017). The aim of this work is to quantify anthropogenic changes in landscape connectivity, especially the links between hillslope–channel lateral connectivity, longitudinal connectivity and lateral connectivity between the channel and valley floor, over the past 50–100 years in the Vuvu catchment, a representative headwater tributary of the Thina River and larger Mzimvubu system. Three objectives were set: map historic gully development; map present-day hillslope–channel connectivity; determine the downstream response in terms of historic channel form changes and modified lateral and longitudinal connectivity. The research findings were used to synthesize connectivity change in the Vuvu catchment, make recommendations for future rehabilitation and develop a conceptual model that illustrates the links and feedback between the different components of connectivity.
Study Area
The Vuvu catchment (65 km2) is situated along the Drakensberg Escarpment in the Eastern Cape Province of South Africa. The catchment is characterized by high relief (Figure 1), steep slopes, high density dendritic drainage network and limited sediment accommodation space on the valley floor. Rainfall varies from 707 to 928 mm per year, increasing with elevation up to 2,100 m a s l (Mucina et al., 2006), whereafter rainfall amount and intensity decrease (Nel et al., 2010). Precipitation peaks in summer, in the form of high intensity thunderstorms (Nel, 2008). Temperatures range between −10·5 and 31·4 °C (Mucina et al., 2006). Soils in the upper catchment are derived from basalts of the Drakensberg Formation, whereas soils of the lower catchment are mainly derived from erodible mudstones of the Elliot Formation. Due to the steep topography, climate and erodible nature of some parent materials, headward erosion towards the escarpment has been dominant over geologic time, resulting in narrow steep valleys. Small localized valley fills exist where the river gradient is low; colluvial fills drape the lower slopes. A wandering channel consisting mostly of boulders and cobbles with sand deposits on higher benches and banks is characteristic of the valley fill. The channel shifts during infrequent large magnitude events. Grassland is the dominant vegetation (Mucina et al., 2006).
The Vuvu catchment is communal land, and, prior to 1994, it was governed as the independent Transkei homeland. Subsistence farming has been practised over the past ~200 years by Sesotho and IsiXhosa people who settled the area. Livelihoods depended mostly on grazing throughout the catchment and cropping in the lower catchment until social grants were introduced in 1956. The value of these increased significantly after 1994, decreasing reliance on land-based livelihoods. Many agricultural fields were abandoned, and livestock grazing became the main land use. Continuous grazing and annual burning are often seen as the cause of land degradation. Sheet, rill and gully erosion are currently common landscape features, especially in the lower parts of the catchment on the more erodible soils. The Department of Environmental Affairs of South Africa has initiated slope rehabilitation in an attempt to reduce the catchment sediment yield and protect downstream water supply structures. Research findings presented in this paper will help to inform rehabilitation efforts in the Vuvu and in similar catchments in the area.
Methods
Three different approaches were used to collect data in relation to the three research objectives. Historic and recent aerial images were used to assess changes in connectivity due to the initiation and expansion of gullies over the past ~50 years. Recent high-resolution aerial images were used to map present day hillslope–channel connectivity. Topographic surveys and depth probing of deposited fine sediment were used to investigate changes to longitudinal and channel–valley fill connectivity along the main Vuvu River.
Mapping Gully Evolution
The evolution of gullies in the Vuvu catchment was mapped from historical aerial photos for 1956 (scale 1:30,000), 1975 (scale 1:50,000) and 2009 (georeferenced ortho photos, 0·5-m resolution). Photos from 1956 and 1975 were used because they were of reasonable definition (despite the smaller scale of the 1975 photographs) and allowed for a relatively equally spaced temporal succession of photos. The 1956 and 1975 photos were georeferenced using stable landscape features such as rocky outcrops, large boulders and houses. Due to the small scale and variation in image quality across the different years, only the larger gullies (2009 length > 50 m) could be accurately digitized (Martinez-Cacasasnovas, 2003). Nine gullies were identified as being suitable for analysis (Figure 1). The aerial photos for the three dates were used to map the change in gully area over time. A logarithmic trend line was fitted to a time series of the data in order to predict the probable initiation date of the features (Harvey, 1992; Huber, 2013).
Mapping Hillslope–Channel Connectivity
Colour aerial images from 2009 (0·5-m resolution; acquired from Chief Directorate: National Geo-spatial Information, South Africa) were used in ArcInfo 10·0 to digitize water and sediment pathways as line features at a scale of 1:2,000 for the entire Vuvu catchment. The different types of pathway are detailed below. Several field visits conducted prior to the mapping (between 2010 and 2012) aided the process of photo interpretation. Geo-tagged land-based photos taken during field trips also helped with aerial image interpretation and provided the basis for ground-truthing the captured features.
Pathway features increasing hillslope–channel connectivity that were identified in the Vuvu catchment included downslope (natural drainage lines, connected gullies, disconnected gullies) and cross-slope (roads and livestock tracks) linkages. Natural drainage lines ranged from small, steep, grassy topographic lows to the well-developed main trunk channel. Gullies were considered large (>2 m or 3 to 4 pixels wide), incised, linear erosional features with steep sidewalls that concentrate flow in a down-slope direction. Gullies were classified as connected or disconnected, based on whether or not the feature was directly linked to the major drainage network (Brierley et al., 2006; Le Roux & Sumner, 2012). Gullies that had more than 20 m of vegetated slope below the outflow of the feature were classified as disconnected. From field evidence, it seemed that a vegetated strip of more than 20 m was an effective sediment buffer as there were signs of sediment build-up on these vegetated areas below the gully outflow.
Dirt roads were easily identifiable as they were larger features (>2 m wide) of a constant width, followed the contour and linked homes to the main dirt road that traverses the lower catchment. Livestock tracks were observed in the field to be narrow (<2 m wide), mostly shallow, linear features, mainly following the contours (aligned across slopes) or ridgelines. Only sections that were clearly visible were digitized. Where livestock tracks were close together or parallel (<5 m apart), only the most prominent track was digitized to prevent duplication, as water would functionally be routed in the same direction. Livestock tracks that occurred within a 5-m radius of a gully were removed from the analysis (Clip tool) to prevent further functional duplication.
Assessment of Down Valley Response—Channel Incision and Straightening
Recent valley fill and channel dynamics were assessed using historical aerial images, surveyed transects and coring. Aerial images for 1956, 1966 and 1975 were georeferenced in relation to georectified aerial images for 2009. The active channel along the Vuvu valley fill was digitized for each set of images. Sections where the river had abandoned a channel or had been straightened were identified. It was assumed that, when a channel is abandoned as a result of river straightening, the abandoned channel bed, consisting of boulders and cobbles, would remain stable and fine sediment would slowly accumulate above it, thus preserving the former channel bed elevation. Sites were located where the position of an abandoned channel could be identified from the aerial photographs. Sites where channel straightening had occurred were selected, thus limiting the chances of cobble bar formation along the outside of these reaches (cobbles expected to be deposited on the inside of bends). Transects across the channel and modern flood bench were surveyed (using a differential GPS). A gouge corer was used to systematically core down to a cobble layer at 2-m intervals along the transect, recording distance along the transect and depth below the surface. The coring data were combined with the surveyed profile to determine the buried profile of the abandoned channel in relation to the current channel.
Because only two suitable sites could be identified along the Vuvu River, the neighbouring Phiri-e-Ntso River with similar hillslope–channel connectivity increases, and catchment area of 78 km2 was included in the assessment. An additional four sites were identified and surveyed along the Phiri-e-Ntso River.
The difference in elevation between the active and abandoned channel was calculated based on the average of the five lowest points along each channel profile (active and abandoned). The difference in abandoned and active bed elevation was calculated for all six sites to determine the extent of incision. The timing of incision was estimated from aerial photograph evidence.
Results
History of Gully Development
A time sequence of aerial images is shown in Figure 2, which indicates the development of one of the gully features (B5; see Figure 1). Results for the time sequence analysis show that gully features in the Vuvu catchment are still expanding in size (Figure 3). The majority of the smaller gullies started post-1950s as was evident from the aerial photographs. Back extrapolation (Figure 3) indicated that the larger features were initiated from the 1920s to the 1950s. Similar results were found for gullies in the adjacent Phiri-e-Ntso catchment (Huber, 2013). It can therefore be concluded that the majority of gullies in the area are of recent origin, post-19th century human settlement. These gullies make a significant contribution to present-day hillslope–channel connectivity which has therefore increased over the last century.
Hillslope–Channel Connectivity
Gullies and livestock tracks added significantly to the drainage density (76 m ha−1) (Table 1). Overall, gullies increased the downslope drainage density by more than 20% above natural. Connected gullies contribute ca. 10% to the drainage density and were effective conduits as these features were pathways flowing down-slope and were directly linked to the drainage network. The 10% contributed by disconnected gullies have less sediment and water transport potential as these features were not directly connected to the drainage network.
| Unit | Catchment | Elliot | Clarens | Drakensberg | |
|---|---|---|---|---|---|
| Area | ha | 5,329 | 1,355 | 529 | 3,444 |
| % of total | 25 | 10 | 65 | ||
| Natural drainage | Length (km) | 404 | 114 | 39 | 251 |
| Density (m ha−1) | 76 | 84 | 74 | 73 | |
| Connected gullies | Length (km) | 42 | 15 | 3 | 24 |
| Density (m ha−1) | 8 | 11 | 6 | 7 | |
| % of natural drainage | 11 | 13 | 9 | 10 | |
| Disconnected gullies | Length (km) | 43 | 15 | 3 | 25 |
| Density (m ha−1) | 8 | 11 | 5 | 7 | |
| % of natural drainage | 11 | 14 | 7 | 10 | |
| Roads | Length (km) | 29 | 29 | 0 | 0 |
| Density (m ha−1) | 1 | 22 | 0 | 0 | |
| % of natural drainage | 1 | 26 | 0 | 0 | |
| Livestock tracks | Length (km) | 637 | 222 | 57 | 357 |
| Density (m ha−1) | 120 | 164 | 109 | 104 | |
| % of natural drainage | 158 | 195 | 147 | 142 |
The Elliot Formation mudstones had the highest natural drainage density of 84 m ha−1 compared to 73–74 m ha−1 for the other formations. Connected and disconnected gullies both followed this trend with the highest densities for gullies on the Elliot Formation.
Roads and livestock tracks, both cross-slope features, added ca. 159% to the drainage density of the entire catchment. Overall, roads contributed far less (<1%) compared to livestock tracks (158%) but were a significant connectivity component within the Elliot formation where they increased the drainage density by 26%. Livestock track densities were ca. 50% higher for the Elliot Formation (195%) than the Clarens (147%) or Drakensberg (142%) Formations. The total increase in cross-slope drainage was more than 220% for the Elliot Formation.
Roads and livestock tracks tend to follow the contour to some extent (gently sloping) or are aligned along ridges, but frequently cross down-slope orientated drainage features, allowing them to accumulate and route hillslope runoff towards the drainage network during overland flow events.
Overall, it is apparent that the relative increase in drainage density due to erosion features, roads and tracks was greatest for the Elliot Formation and least for the Clarens Formation. The absolute increase was greatest for the Drakensberg Formation, reflecting the larger area.
Down Valley Response—Channel Straightening and Incision
River migration could be tracked successfully from the historical images, which allowed field measurements to be made of river incision for the observed period (Figure 4). In Figure 4, it can be seen that the current channel of VT7 for 2012 is ca. 50 cm deeper than that for 1956 and for VT14, 900 m higher up the Vuvu River, the bed elevation difference for was 1·4 m (Table 2; Figure 4). The Phiri-e-Ntso River also experienced a lowering of the current bed elevation over the period since 1956 (Table 2) with incision of 0·28 to 0·89 m being measured. Although this observation was based on limited evidence, the parallel incision along both rivers would suggest that synchronous incision took place post-1956. This incision is likely to be caused by changes in hillslope–channel connectivity rather than the upstream migration of changes in base-level. Base levels along both rivers are fixed by resistant dolerite dykes that act as local base levels and should prevent upstream propagation of incision.
| River | Cross section | Difference in bed level (m) | Average (m) |
|---|---|---|---|
| Vuvu | VT7 | 0·50 | 0·95 |
| VT14 | 1·40 | ||
| Phiri-e-Ntso | P1 | 0·65 | 0·54 |
| P2 | 0·28 | ||
| P4 | 0·89 | ||
| P7 | 0·33 |
Both rivers experienced river straightening after 1956 (Table 3). Channel straightening for the Vuvu River is evident from Figure 4. Over the period 1956 to 2009, the sinuosity of the Vuvu decreased from 1·21 to 1·13 and of the Phiri-e-Ntso from 1·48 to 1·32 (Table 3). The channel also has widened along sections (VT14 and X) as can be seen in Figure 4. The decreased sinuosity is linked to incision and river steepening that will ultimately increase the transport efficiency of the channel.
| River | Valley fill length (m) | Year | Length (m) | Sinuosity |
|---|---|---|---|---|
| Vuvu | 1977 | 1956 | 2,391 | 1·21 |
| 2009 | 2,241 | 1·13 | ||
| Phiri-e-Ntso | 905 | 1956 | 1,340 | 1·48 |
| 2009 | 1,194 | 1·32 |
Discussion
The detailed mapping of gully development and hillslope–channel connections made it possible to determine the age of the gullies and the recent changes to hillslope–channel connectivity. Mapping of the channel straightening and incision enabled changes in hillslope–channel connectivity to be linked to longitudinal and lateral connectivity between the channel and valley fill. The causes of connectivity change and the links between the different components [hillslope–channel, longitudinal and lateral (channel–valley fill)] are considered below.
Landscape Setting of the Vuvu Catchment
Landscape degradation associated with anthropogenically induced changes to sediment connectivity in the Vuvu catchment, and future restoration options are presented in Figure 5. This degradation should be seen within the context of high natural connectivity due to steep relief, high rainfall (for South Africa) and a well-developed natural drainage network that contributes to efficient hillslope–channel connectivity (Figure 5A). The Vuvu catchment is in a phase of geological incision as headward erosion cuts into the Drakensberg Escarpment (Partridge & Maud, 1987). The gentler colluvial slopes of this stepped landscape comprise small buffers within the sequence of steeper slopes. A narrow valley fill in the lower catchment stores both coarse-grained and fine-grained sediment. Holocene cut and fill cycles are evident on the hillslopes and valley fill as sediment not exported from the catchment is redistributed and stored in new locations, such as alluvial fans (van der Waal, 2015). Soil derived from the erodible mudstones in the lower part of the catchment contribute to naturally high suspended sediment concentrations in the Vuvu River (van der Waal et al., 2015; Rowntree et al., 2016). This high energy system is thus prone to efficient sediment transfer, but human induced pressures have further increased the sediment transfer potential.
Changes to Hillslope–Channel Connectivity
Livestock grazing and trampling, frequent burning, the development, and the subsequent abandonment of agricultural fields and the introduction of housing areas and associated road networks have led to general disturbance, reduction in vegetation cover and successive increases in runoff and erosion in the catchment (Figure 5B). Gullies, roads and livestock tracks increased the down-slope and cross-slope routing of water, limiting water infiltration and concentrating flows, thus increasing hillslope–channel connectivity (Figure 5C). Increased runoff volumes have resulted in increased energy available to erode soils and transport sediment. The increase in hillslope–channel connectivity would have increased both fine-grained and coarse-grained sediment transport. However, the generally fine nature of the sediment eroded from the Elliot mudstones and the high energy system have favoured transport in suspension, thus limiting both sediment deposition and the effectiveness of existing buffer features on the hillslopes and the valley fill.
Gullies are highly visible erosion features throughout the catchment and have contributed to a 22% increase in drainage density (Table 1). Connected and disconnected gullies contributed equally to the measured increase in hillslope–channel connectivity, but connected gullies were the most important pathway identified as they are directly linked to stream channels, making the timing of their flow contribution synchronous with that of the natural drainage network (cf. Croke et al., 2005; Peñuela et al., 2016). Disconnected gullies are assumed to be hydrologically and sedimentologically disconnected from the drainage network during low intensity flow events due to dispersive flow through vegetation buffers (cf. Croke et al., 2005), but can contribute significantly once they become connected during high magnitude events (Fryirs et al., 2007a, 2007b). Both connected and disconnected gullies are for the most part headward extensions of the natural drainage network on slopes that would otherwise be vegetated under un-impacted conditions. The removal of vegetation and subsequent entrenchment of the natural drainage features have altered hydrological and sedimentological behaviour, allowing water and sediment to be efficiently transported down the catchment (Brierley & Murn, 1997; Cammeraat, 2002; Vanacker et al., 2005; Rommens et al., 2006; Grenfell & Ellery, 2009; Pechenick et al., 2014; Masselink et al., 2016).
Although other authors have shown that gullies in South Africa may predate recent human occupancy (cf. Temme et al., 2008), this study has shown the Vuvu gullies to be more recent features, probably resulting from recent human settlement (Figure 3). Large-gully formation seems to have started in the early 1900s in the Vuvu and is in agreement with similar results found in the neighbouring Phiri-e-Ntso catchment (Huber, 2013). The smaller gullies were mostly initiated post-1950. Historical reports (Bundy 1987) indicate that the timing of gully initiation is coincident with changes in livestock practices and numbers. In the early 20th century, measures were enforced to prevent East Coast fever in cattle spreading throughout the Transkei region (Bundy, 1987). In 1906, cattle dip tanks were introduced to limit the potential pest carrying ticks, and, by 1911, it became law to dip cattle every week. Restrictions on cattle export from the area hampered livestock trade which, in turn, led to an increased density of cattle and altered patterns of transhumance pasture usage. Local farmers complained about the general exhaustion of the cattle walking to and from the dip tanks on such a frequent basis, as it decreased milk production and the ploughing power of their animals (Bundy, 1987). It is surmised that this increase in numbers, coupled with the weekly trip from the pastures to the dip tanks, led to general vegetation degradation and an increase in the number of livestock tracks throughout the catchment. This would have led to increased runoff and hydrological connectivity that could contribute to gully formation and channel incision.
The second period of gully initiation can be linked to increased cattle numbers in the 1950s. According to local elders (born in the 1940s) talked to during field work, livestock did extremely well at this time due to well-timed and sufficient rainfall, putting an added wave of pressure on the already stressed landscape.
Livestock tracks not only contribute to gully erosion by diverting runoff downslope along well-defined pathways, they also create their own form of cross-slope connectivity, forming where livestock frequently traverse the catchment. Livestock tracks were shown to have increased the drainage density by 158% (Table 1). Roads link the various homesteads to the main road that traverses the catchment and links the various scattered villages. Their impact is much smaller, increasing drainage density by less than 1% overall, but by 26% in the settled area underlain by the Elliot Formation. The cross-slope features (not strictly horizontal) were not regarded as efficient a conduit as the downslope gullies, but they do concentrate and canalize overland and upper subsurface flow (Pechenick et al., 2014). The cross-slope features crossed downslope drainage features on a regular basis, potentially discharging concentrated flow into the drainage network (cf. Hoffman & Todd, 2000; Croke et al., 2005; Wemple, 2013; Huchzermeyer, 2014; Pechenick et al., 2014). Huchzermeyer (2014) found that livestock tracks and roads altered the size and shape of catchment areas feeding into gullies and added significantly to both the catchment area of gullies and the hillslope-gully connectivity. These findings are supported by Pechenick et al. (2014) who detected an influence of road network density and landscape position on downstream channel condition as a result of increased hillslope–channel connectivity.
Drainage efficiency was thus intensified by anthropogenic influence for all slopes, increasing hillslope–channel connectivity, even on gentle slopes that could potentially act as buffers. The measured increase is an order of magnitude higher than the 10% calculated by Croke et al. (2005) for an extensive forest road network in New South Wales and highlights the extremity of the connectivity increase in the Vuvu catchment.
Sediment tracing research by Van der Waal et al. (2015) suggests that these changes in landscape connectivity have not changed the provenance of sediment over the past 100 years; the relative contribution of sediment from igneous (upper catchment) and sedimentary (lower catchment) sources has remained constant. This suggests that the spatial distribution of connectivity has not changed significantly over the same period.
Changes to Channel–Valley Fill Connectivity
Increased hillslope–channel connectivity is surmised to have impacted on processes on the valley floor (Figure 5D). Historical aerial photo analysis combined with cross-sectional surveys was used to assess the incision that has taken place in the Vuvu and Phiri-e-Ntso Rivers since 1956. Incision occurred over the wider area of both river systems, supporting the assertion that input–output relationships have been altered for the wider area, rather than incision being related to a localized extreme event (Brunsden & Thornes, 1979). Given the synchronicity with increased gully erosion, we postulate that the increase in flood discharge and associated stream power was a response to increased landscape connectivity throughout the catchment. Booth (1990) related similar levels of incision (0·2 m) in the Soos Creek, western Washington, USA, to increases in discharge as a result of increased runoff efficiency.
Increased runoff from the hillslopes is thus invoked to explain the observed channel incision, whereas increased erosion and sediment delivery explain the sediment deposition over the abandoned channels to form the modern flood benches. Together, these two actions have reduced channel–valley floor connectivity. Van der Waal et al. (2015), using evidence from flood bench stratigraphy, estimated that sediment accumulation rates have increased over the past 50 years, probably linked to higher sediment concentrations during floods (possibly with greater magnitude) in response to increased hillslope–channel connectivity. These findings support the conclusion of Rowntree and Dollar (1996) that catchment and channel processes are inseparable, and that disequilibrium in the catchment will be transferred to the channel.
This recent incision must be seen in the context of long-term cut and fill phases. Van der Waal (2015), using OSL dating, identified at least two periods of downcutting and terrace formation in the Vuvu during the Holocene, c. 5000 bp and c. 2000 bp. These were probably in response to regional changes in climate, as has also been postulated by Bookhagen et al. (2006) for the Himalaya and Temme et al. (2008) for headwater streams in the Kwa-Zulu Natal Drakensberg. There is no evidence, however, to suggest that the recent incision is a response to climate change so hillslope connectivity resulting from human settlement is likely in this case to be the main driver.
The valley floor has limited storage space due to its confined nature. Van der Waal et al. (2015) showed that narrow flood benches are stores of mostly fine-grained sediment with coarse-grained sediment or bedrock forming the base of the flood benches. They presented the following hypothesis of landscape change. Increases in runoff and hillslope–channel connectivity resulted in larger flood magnitudes and increased sediment concentrations. This increased flow energy would have enhanced the sediment entrainment and transport capacity and resulted in channel incision, straightening and widening. This in turn reduced the channel–valley fill connectivity, but due to increased sediment concentrations, sediment deposition on flood benches increased during the few events that do inundate the flood benches. The overall result is increased down-valley or longitudinal connectivity due to the small storage potential of the narrow flood benches. This reduction in channel–flood bench connectivity and greater flow energy regimes reduced the potential to store sediment in the valley fill, enhancing an already high sediment yield from the Vuvu catchment. This shows that the degraded nature of the catchment reduced the buffering capacity of the landscape, enhancing sediment production and sediment transfer, leading to high sediment yields. These higher yields are in line with reports by Hey (1957) who stated that rivers draining the escarpment ran ‘gin’ clear even after large floods in the early 1900s, but became turbid by the 1950s.
Restoration Recommendations
Catchment restoration should be based on a strategy that increases disconnectivity at the hillslope scale, buffering the landscape and increasing sediment retention (Figure 5E). Restoration of the catchment's vegetation cover will limit rain drop-related soil erosion, the volume and energy available of runoff to erode and transport sediment, while targeting pathways will lead to reduced hillslope–channel connectivity. Buffers in and around gullies, livestock tracks and roads can further facilitate sediment deposition and water infiltration, reducing the energy of storm runoff. Channel incision should slow down, with a reduction in net sediment export from the valley fill. The river could continue to rework stored valley sediment until a new stable phase is developed as flood benches are rebuilt to the new channel level, supporting renewed channel–valley fill connectivity that facilitates greater sediment storage. It is envisaged that restoration will reduce sediment contributions from hillslopes within a few years, but valley fill degradation could continue for another 10–100+ years while a new sediment storage phase is developed on the valley fill. The outcome of the catchment-wide restoration is, however, mere speculation and warrants further research and monitoring.
A Conceptual Model of Connectivity Linkages
Research in the Vuvu catchment has demonstrated that anthropogenically induced changes to hillslope and valley floor sediment connectivity are linked, with increased hillslope–channel connectivity in turn causing channel straightening and incision, increased longitudinal connectivity but reduced channel–valley floor connectivity, further exacerbated by increased overbank deposition of sediment during high magnitude events. These relationships are captured in Figure 6 which relates the connectivity/disconnectivity concepts of Fryirs et al. (2007a) to sediment relationships at different landscape scales. At the hillslope scale, disconnectivity is sustained through a good vegetation cover, but connectivity is increased by degradation process that reduces vegetation cover and causes rill and gully erosion (Harvey, 2002). Hillslope–channel connectivity is naturally high in landscapes such as that for the Vuvu with steep slopes, a high drainage density and valley confinement but is buffered by river terraces, floodplains and alluvial fans. Alluvial fans and terraces are the main buffers in the Vuvu catchment. Longitudinal connectivity operates at the channel network scale. Increased stream power arising from more effective hillslope–channel connectivity leads to channel straightening, effectively increasing longitudinal connectivity, whereas valley constrictions, bedrock barriers and, locally, coarse sediment bars, provide barriers to sediment movement. Although not present in the Vuvu, dams and weirs are among the most drastic artificial barriers reducing longitudinal connectivity, trapping sediment and reducing flood peaks. Lateral and vertical connectivity are effective (and affected) at the channel reach scale. Increased longitudinal connectivity translates into channel incision but can also lead to heightened flood levels due to a more flashy runoff response (Peñuela et al., 2016). Increased hillslope connectivity brings in more sediment to be deposited as blankets on flood prone areas, further reducing lateral connectivity. Vertical connectivity in the channel can also be reduced by the deposition of sediment blankets, although this was not observed in the Vuvu due to a high sediment transport capacity over the channel bed.
Conclusions
The method of digitizing pathways that contribute to the drainage efficiency helped to unravel how landscape connectivity has changed at the slope and catchment scale over the last 50–100 years. Gully development was successfully tracked using historical images and could be used to calculate likely initiation dates. Valley floor mapping and channel surveys helped untangle changes to the channel depth and channel form as a response to altered landscape connectivity in the catchment. This detailed mapping approach provided insights into changes in connectivity of various features and will allow researchers worldwide to make links between changes to structural connectivity and related processes from the hillslope to catchment scale.
To summarize the findings of this terrain and aerial photo analysis, it is evident that the Vuvu catchment is a steep headwater catchment with a high drainage density. It is naturally well connected, but due to anthropogenic influences such as continuous grazing, ploughing of fields and subsequent abandonment leading to gullying, erosion and landscape connectivity have been increased over the past 100 years.
Detailed quantification of human induced changes to landscape connectivity shows pathways linking slopes to the channels have increased both in the cross-slope (roads and livestock tracks) and down-slope (gully) directions. This increase in connectivity results in sediment that has been mobilized off the slopes having a higher potential of being transported to the main river, even during lower intensity rainfall events. The potential to store sediment in the catchment has thus been reduced due to anthropogenic activity. From historic photo analysis, it was predicted that the major gullies were initiated between 1910 and 1970. It is thus likely that other erosion features would have developed during the same time. Gullies and livestock tracks were the main contributors to increased hillslope–channel connectivity. The greater impact of erosion and connectivity changes on the Elliot Formation is likely to be a function of the erodible nature of the soils and ongoing anthropogenic activity in the lower catchment.
Field observations and detailed mapping support the assertion that channel incision and straightening of the Vuvu River are likely to be related to increases in hillslope–channel connectivity that increase peak discharge, sediment supply from hillslopes and sediment entrainment capacity in the channel. The effect of the channel incision and straightening would be increased sediment transfer through the system, changing the function of the valley fill from a sediment buffer to a booster.
Catchment management efforts should aim to improve vegetation basal cover to reduce soil erosion and overland flow. Pathways should be turned into sinks where possible and existing sinks should be protected in order to prevent these becoming sources and pathways.
The Vuvu catchment is likely to remain a non-equilibrium system where sediment export dominates over sediment deposition. This is mainly due to its steep topography, limited accommodation space and erodible soils. The recent anthropogenic influence has initiated an overall accelerated cut phase on the hillslopes and valley fill, but with the appropriate restoration strategies, a catchment-wide fill phase can be initiated with nett reductions in sediment yield.
Acknowledgements
We would like to thank the funders, the Natural Resources Management programme of Department Environmental Affairs and the Water Research Commission of South Africa, for making this research possible.
