ELSEVIER ARTICLE INFO Available online 6 November 2009 1. Introduction Understanding mechanisms of sediment generation, storage and flux to the ocean is a key goal of the MARGINS Source to Sink Programme. Estimating the flux of fluvial sediments discharged to the ocean from small, steep, catchments located in tectonically active margins has proven to be difficult (Milliman and Syvitski, 1992; Walling and Fang, 2003). Equally challenging is to understand factors that influence different erosion processes operating in steepland catchments and which largely dictate temporal and spatial variability in the supply of sediment to fluvial systems and ultimately to the ocean. Gully erosion can contribute a significant proportion of sediment to fluvial systems in steepland catchments (De Rose et al., 1998; Hicks et al., 2000; Kasai et al., 2001; Marden et al., 2005; Page et al., 2008). Sediment supplied from gullies tends to be more persistent than that delivered by influx from shallow landsliding in these environments because established gullies are activated by small, frequent rainstorms, whilst landsliding is activated during less frequent, high magnitude events (Hicks et al., 2000). At the head of the catchment sediment cascade, gully erosion may be enhanced by mass movement processes and the two mechanisms of sediment delivery may combine. Whilst smaller gullies tend to be linear features in zones of topographic convergence in otherwise unchannelled, zero-order basins (De Rose et al., 1998), larger features develop an amphitheatre-like form in conjunction with mass movements. These mechanisms give rise to fluvio-mass movement gully complexes (sensu Betts et al., 2003), where sediment is generated both by fluvial incision and mass movement, with the latter being dominant in terms of sediment volumes produced (De Rose et al., 1998; Betts et al., 2003). In these systems, mass movements tend to comprise debris flows and deep seated and shallow sliding, which may be (re) activated by gullying within the complex as part of intrinsic feedback mechanisms (e.g. slope undercutting), as well as high magnitude rainstorms. In the steepland East Coast Region of New Zealand's North Island, more than half of the sediment load of the Waipaoa and Waiapu Rivers is derived from these complexes (Marden et al., 2005; Page et al., 2008). The suspended sediment discharge of the Waipaoa River is contingent upon gully erosion as a whole, which is the dominant process of sediment delivery to this river (Gomez et al., 2003a). Annual average suspended sedimentyields are among the highest recorded in the world (Walling and Webb, 1996; Hicks et al., 2000). The current sediment yield of the Waipaoa is 15 million tonnes per annum, with a mean specific yield of 67501 krrrV-1 (Hicks et al., 2000). De Rose et al. (1998) suggest that-3% of this figure is derived from a single gully-mass movement complex known locally as Tarndale Slip. The significance of the Tarndale gully complex to the Waipaoa sediment cascade has increased as other gully complexes have been shut down by reforestation since the 1960s (Gomez et al., 2003b; Marden et al., 2005). The significance of gully erosion and contribution from the Tarndale gully complex in the Te Weraroa Stream (a principal headwater tributary of the Waipaoa) can be further appreciated in recognising that whilst gully erosion in this sub-catchment affected - 6% 46 of catchment area at its peak, in the period 1970-1988 62% of all sediment in this catchment was generated from the Tarndale complex (Gomez et al., 2003b). The Tarndale gully complex is buffered from the channel system by a small (-11 ha) depositional fan. These fans are typical of large gully complexes in the region (De Rose et al., 1998; Betts et al., 2003) and sediment is supplied directly to them by both fluvial and slope processes operating in the contributing gully complex. The cutting or filling of these fans may respectively amplify or modulate sediment supplied to the stream system from contributing gully complexes. The extent of buffering is therefore temporally variable and requires assessment to better understand sediment delivery processes in this upper component of the sediment cascade. Page et al. (2001) identified the need to improve understanding of the drivers of sediment production, transport and deposition, their interaction and spatial controls on sediment fluxes. This paper therefore seeks to extend that understanding at a key location within the upper Waipaoa sediment cascade by addressing the temporal variability of stream channel buffering via assessment of cutting and filling of the Tarndale Fan over decadal and seasonal timescales. In terms of addressing source to sink, this component of the sediment cascade is highly significant, given the contribution of gully erosion to the sediment yielded to the ocean from the Waipaoa catchment (cf. Gomez et al., 2003a). This implies that a large proportion of sediment deposited on the shelf is sourced by gully complexes such as Tarndale. This paper focuses on quantifying those mechanisms and volumes of sediment involved over a 33 month period from December 2004 to August 2007. We do so by examining in detail the cutting and filling of the Tarndale Fan in the headwaters of the Waipaoa catchment as it responds to sediment delivered from a major gully complex. This extends our understanding of sediment production, storage and transfer in a significant gully-fan system in the Waipaoa by increasing the temporal and spatial resolution of investigation. To date research has tended to be decadal or at best annual. Here a seasonal approach enhances the temporal resolution of the processes responsible for sediment generation, storage and transfer. Furthermore, in adopting a high-resolution survey approach we are able to provide a higher spatial resolution of fan aggradation and incision than has been allowed to date using cross-section surveys or photogrammetry. As such, this paper will provide new insights into fan behaviour and the role of these fans in sediment transfer within the Waipaoa catchment as a whole. 2. Study site 2.1. Location The Tarndale gully complex and fan are located in Te Weraroa Stream, an ungauged headwater tributary in the 2200 km2 Waipaoa River basin (Fig. 1). The catchment surrounding the Tarndale gully Fig. 1. Location map of Tarndale gully and fan in the context of the upper Waipaoa catchment: (a) LiDAR derived digital elevation model of the gully-fan complex (LiDAR data1 Gisborne District Council, courtesy Dave Peacock), (b) upper Waipaoa catchment showing major lithologies and gullies (after Marden et al, 2005). 47 complex (Fig. 1) at its highest elevation is 580 m. Valley side slopes are c. 800 m long, with an average slope angle of 25°. The fan extends from the base of the headwall (elevation 420 m) to its junction with Te Weraroa Stream (elevation 320 m), a distance of -1 km. At its upstream end the fan system comprises four tributaries each draining a different part of the -20 ha (De Rose et al. 1998), amphitheatre-shaped, gully complex headwall and converging to form a single feature (Fig. 1). 2.2. Geology The gully-fan complex at Tarndale is underlain by variably indurated, sheared and crushed, well bedded, light and dark grey siliceous mudstone alternating with thin sandstone, and poorly bedded, pale grey calcareous mudstone of Late Cretaceous to Palaeocene age (Black, 1980; Mazengarb et al., 1991). In the vicinity of Tarndale extensive crushing associated with major faults and an extant slump (Black, 1980; Mazengarb et al., 1991), together with argillites that are especially susceptible to acid sulphate weathering (Pearce et al., 1981), have predisposed lithologies of the Whangai Formation (Fig. 1) to mechanical disintegration under the influence of water (Marden et al., 2005). These conditions generate a highly erodible substrate with potential for high rates of sediment delivery from the gully complex to the fan. Furthermore, rapid mechanical disintegration produces an abundance of fine grained material; the D50 of surface material on the upper fan is 1.4 mm (Gomez et al., 2001) and 60% is finer than 2 mm (Phillips, 1988). 2.3. Climate Climate is humid temperate, but the area is subjected to periodic high intensity cyclonic storms. There is a 29% chance that an extreme rainfall event will occur every decade (Kelliher et al., 1995). Since 1900 there have been 33 extreme rainfall events, when the discharge of the Waipaoa River (at Kanakanaia, cf. Fig. 1) exceeded 1500 m3s~1, and there is concommitant evidence of widespread and accelerated gully erosion (Cowie, 1957; Phillips et al., 1990; Kelliher et al., 1995). The largest recorded cyclonic storm (Cyclone Bola) occurred in March 1988, and generated 500 to 700 mm of rain in a 5-day period. Rainfall figures from rain gauge sites near the study site (Fig. 1) show that August has the highest monthly rainfall with a maximum of 284 mm (Gage and Black 1979). At 200 m average annual rainfall is -1339 mm, increasing to 2500 mm at elevations of 800 m (Pearce et al.,1981). Monthly flood occurrence, precipitation and rain days all show a pronounced late-winter maximum. Flooding is also associated with prolonged periods of wet weather and not simply with exceptional rainfall intensities. 2.4. Gully-fan morphology Along the longitudinal profile of the Tarndale gully-fan system, three morphologically distinct zones can be distinguished. Bergstrom (1982) suggests distinct breaks in valley morphology generally mark boundaries between headwater, transitional, and lower braided zones. Drainage from the headwall scarp is confined to steep and narrow rills which progressively widen and deepen in a downslope direction. At the base of the headwall scarp, slope decreases abruptly at the nexus with the fan (Fig. 2). Excessive sediment accumulation in the reach of the central fan immediately downstream of its junction with the lowermost tributary (upper mid-fan) results in over-steepening and widening. Between this point and the junction of the fan with Te Weraroa Stream channels are braided. For much of its length the fan consists of an active zone, swept by channels, incised within 'older' aggradational terrace/fan alluvium. The latter spans the total width of valley floor and represents the maximum elevation of valley-floor infilling attained in this gully-fan system. This stable and currently elevated surface exists today as discontinuous remnants flanking the channel along much of its length (Fig. 3). The current fan morphology dates to the 1960s. In May 1960 sediment emanating from the Tarndale gully-mass movement complex completely destroyed fascines emplaced to mitigate sediment supply to the Te Weraroa Stream (Marden et al., 2005) and buried existing benchmarks established to measure changes in bed level across the valley floor. With the destruction of the fascines it is likely a phase of channel incision began with the development of a nick-point at this site, which propagated upstream to the base of the gully headwall. This process created a narrower active fan incised below the level of the former fan surface with remnants of the latter being temporarily preserved along either side of the valley. More recently, and as a result of a major storm event in 1980 (Cyclone Bernie), sediment deposition across the width of the fan and subsequent channel incision culminated in the formation of a new and extensive terrace flanking much of the 1 km length of Tarndale Fan. This elevated surface is here referred to as the 'inactive fan' (Fig. 3). In 1983 a new network of benchmarks on this elevated surface was established and channel cross-section measurements commenced. These data contribute to the 20 year record of fan behaviour reported in this paper. Early planform surveys of the downstream end of the fan show that at times of excessive sediment influx the channel was only shallowly incised and frequently changed its position across much of the fan-width. In contrast the contemporary active channel deeply incises the fan surface (Fig. 2a). The fan terminates at its junction with Te Weraroa Stream (Figs. 2c and 3) where deposition in the past has splayed upstream and downstream of this confluence and on occasion has completely blocked Te Weraroa Stream to form a temporary lake. Sediment discharged from the Tarndale gully complex has, over the years, deflected Te Weraroa Stream against its true left bank (Figs. 2c and 3) where the valley slope above has become destabilised and continues to fail as a large rotational slump. The morphological development of the fan over the last 20 years provides a context for a more detailed appraisal of cutting and filling of the fan between 2004 and 2007. 3. Methodology and results 3.1. Cross-section surveys The past 20 years of fan behaviour has been established by cross-sectional survey of the fan, carried out (by MM) between benchmarks installed in 1983 (Fig. 3). Survey data were collected using a Sokkhisha Total Station (electronic distance measurement (EDM) precision ±(5 +5 ppmxD) mm, angular resolution Is). Biannual surveys within this period permitted identification of seasonal cut and fill cycles, where surveys were completed in May (end of summer) and November (end of winter). The volume of scour and fill, and the net change in sediment storage was determined as the mean of the upstream and downstream cross-section area times the length of each reach (cf. Griffiths, 1979). Changes in cross-section area between consecutive surveys were based on changes in elevation across the width of the active portion of the fan. The end point method calculates the volume as per Eq. (1). V A, (1) where A\ = area of cross section at downstream end of reach (m2) A2 = area of cross-section at upstream end of reach (m2)L = distance between the sections (m) Net volumetric changes are shown in Fig. 4. These demonstrate a degree of seasonality where seasonal surveys were completed. Generally the winter period is dominated by infilling of the fan, whilst 48 Fig. 2. Photographs of cutting and filling on the Tamdale Fan 2004-2007. Locations of these photos are given in Fig. 5, where both spatial and temporal context are provided, (a) 3 m deep channels incised in the upper Tamdale fan (April 2005, cf. Fig. 5(a), photo I.C. Fuller), (b) direct delivery of sediment to the upper fan by landsliding attributed to the Labour Weekend storm, 22-24 October 2005 (cf. Fig. 5(c), photo I.C. Fuller). Note the dry and hardened surface, photo taken December 2005. (c) formation of a new fan building into the Te Weraroa Stream (November 2006, cf. Fig. 5(f), photo I.C. Fuller). Note the delivery of fines to the Te Weraroa Stream in the colour change at the confluence, (d) up to 5 m deep infilling by single debris flow (April 2007, cf. Fig. 5(g), photo I.C Fuller). the summer period is dominated by incision. These cross-sections are at a relatively coarse spatial and temporal resolution and it is evident that a seasonal cutting and filling may be an oversimplification. To further assess fan behaviour a more intensive approach to survey was adopted from 2004 using a Real Time Kinematic differential Global Positioning System (RTK-dGPS) to generate topographic data for subsequent analysis using digital elevation models (DEMs). 3.2. Digital elevation models DEMs derived from aerial photos have been used previously to assess gully erosion (Betts and DeRose, 1999), although necessarily these are at a lower resolution than is feasible using detailed GPS data. Between December 2004 and August 2007, the entire active fan surface was surveyed nine times using RTK-dGPS. A Trimble® R8 GPS receiver was set up in transmit mode to act as a base station and a second R8 receiver was used as a Rover unit, to deploy RTK-dGPS survey. This setup permits rapid data acquisition using 1 s occupation time per observation and real time coordinate calculation using on-the-fly algorithms (Stewart and Rizos, 2002). The minimum acceptable vertical accuracy of observations was set at 0.05 m. Average vertical accuracy was 0.02 m. Points were surveyed to a precision of 0.001 m. The base station was set up some distance away from the active fan itself, preventing any multipath errors (Kennedy, 2002). DEMs were constructed using Surfer® G1S. These DEMs are based on 1 m grids (Fuller et al., 2003); however as survey data are not collected on a grid basis they were interpolated to create a digital elevation surface. Data interpolation (DEM generation) in this study uses Triangulation with Linear Interpolation (TL1). TL1 is a grid-based version of a triangulated irregular network (TIN). It is constructed of contiguous triangular facets, irregularly sized and spaced (Tsai, 1993; Lee, 1991). TL1 is based upon optimal Delaunay triangulation. All grid nodes within a given triangle are defined by the triangular surface and because the original data are used to define the triangles, the data are honoured very closely (although not exactly) (Surfer, 2002). TL1 does not extrapolate z values beyond the range of data. TL1 is regarded as being most effective when the data are distributed evenly throughout the study area, but sparse data points can manufacture obvious (and unrepresentative) triangular faces (Surfer, 2002). 3.2.1. DEM validity DEM accuracy is assessed using an approach recommended by Fisher and Tate (2006), which derives the error standard deviation (S) 49 (Eq. (2)), and also permits estimation of bias using the Mean Error, which may be positive or negative (Table 1). £[(zD -zref)-M£]' n-1 (2) where zDEM is the measurement of elevation from the DEM, zref is the higher accuracy measurement of elevation for a sample of n points. ME is the mean error (Eq. (3)). ME Zref) (3) Each survey point was used as a zref value, to provide an estimation of error derived from the entire DEM. However, it should be noted that this method of error estimation is dependent upon the surveyed points; error at interpolated points is not assessed, as independent data are not available. Whilst survey points could have been thinned to provide quasi-independent data to check interpolation, as surface sampling was designed to provide the best possible data for interpolation, this would inevitably reduce the quality of the DEM and still not provide a rigorous measure of accuracy. This approach therefore does not necessarily provide an unbiased measure of overall DEM quality, but research elsewhere indicates this approach is fit for purpose (Fuller and Hutchinson, 2007). The errors (S) and bias (ME) for each surface are indicated in Table 1. Generally the ME indicates a consistent underestimation of the surface, with the exception of the first survey in December 2004. That this was slightly positively biased (i.e. interpolated surface lying above the date points), and the subsequent surface slightly negatively biased suggests that when comparing these two consecutive surveys, volumetric estimates will be biased towards scour. 3.2.2. DEM differencing and sediment transfer estimation Elevation changes between successive DEMs may be used to derive sediment gains and losses based on morphological budgets, given their valid representation of surface topography (Fuller et al., 2003). DEMs of difference (Fig. 5) depict patterns of sediment gains and loss, indicating complex patterns of cutting and filling of the Tarndale fan between December 2004 and August 2007. Table 2 summarises volumes and quantities of sediment eroded and deposited in the Tarndale fan between successive survey periods. Quantities are based on a dry bulk density of 1840 kg m~3 (De Rose et al., 1998). Errors associated with these volumetric estimations using TLl-derived DEMs are in the order of ±5% (Fuller and Hutchinson, 2007). It should however be noted that volumes are a lower-bound estimation of total sediment transfers (Fuller et al., 2003), neither do they account for suspended sediment, which forms the majority of sediment generated from the Tarndale gully complex (Gomez et al., 2003b). This assessment of sediment transfer is therefore limited to the coarse O > Cyclone Bola V£FÚ 1 r all w Area of histogram = Increase/decrease in storage volume over period □ Summer ■ Winter 1985 1990 1995 Year 2000 2005 Fig. 4. Net volumetric changes in the Tarndale fan between 1983 and 2004 based on cross-section surveys. 50 Table 1 Error analysis showing Standard deviation and Mean Error for each DEM, derived from Eqs. (2) and (3). Date S ME (m) December 2004 0.116 0.0007 April 2005 0.155 -0.0067 August 2005 0.121 -0.0067 December 2005 0.087 -0.0023 May 2006 0.116 -0.0014 August 2006 0.102 -0.0035 November 2006 0.108 -0.004 April 2007 0.111 -0.0021 August 2007 0.099 -0.0032 fraction, which conditions channel morphology (Leopold, 1992; Martin and Church, 1995). The coarse fraction is here defined as that which thus behaves as bedload; the D50 of which has been Fig. 5. DEMs of difference for successive surveys: (a) December 2004-April 2005, (b) April 2005-August 2005, (c) August 2005-December 2005, (d) December2005-May 2006, (e) May 2006-August 2006, (f) August 2006-November 2006, (g) November 2006-April 2007, (h) April 2007-August 2007. Numbers indicate location of photos in subsequent Figures (Fig 6). Table 2 Sediment volumes and tonnes produced by cutting and filling of the Tamdale fan between successive survey periods, based on DEM differencing to provide a morphological budget (Fuller et al. 2003). Period Volumes (m3) Tonnes Erosion Deposition Net Net Yielded3 December 2004-April 11997 6424 -5573 -10 254 -10 000 2005 April 2005-August 6174 24452 + 18 278 + 33 632 - 2005 August 2005—December 11162 31528 + 20 366 + 37 473 -20 000 2005 December 2005-May 21 075 8078 -12 997 -23 915 -24 000 2006 May 2006-August 13 936 25689 + 11 753 + 21 626 -25 000 2006 August 2006-November 3100 6912 + 3812 + 7014 -5000 2006 November 2006-April 4247 14935 + 10 688 + 19 666 - 2007 April 2007-August 2007 10125 11270 + 1144 + 2105 - a This figure is a lower-bound estimate based on a morphological budget It takes no account of throughput or yield of fines in suspension, which is likely to be far higher (cf. De Rose et al. 1998). Whether eroded sediment is yielded depends on where the erosion takes place. Hence, whilst net change may be positive, suggesting accumulation and minimum yield, if the erosion takes place in the lower fan, that material may be considered as a yield from the fan to the Te Weraroa Stream. measured at 1.4 mm and 60% is finer than 2 mm (Phillips, 1988; Gomez et al., 2001). Patterns of cutting and filling are discussed below. 4. Discussion Superficially, DEM analysis indicates a degree of seasonal dependence, with the fan mostly filling during winter and cutting during summer (Fig. 5). However, the increased frequency of survey demonstrates a greater complexity. Patterns of cutting and filling are briefly discussed for each survey period, highlighting both this complexity and the volumes of sediment transferred in connection with cut and fill activity. 4.1. December 2004-April 2005 Incision generally characterises the upper fan, where up to 3 m deep channels have been cut (Fig. 5a and cf. Fig. 2). The intensity of incision is reduced in the lower fan, to a point where it aggrades in this period in its lowermost reach (Fig. 5a). This suggests that the lower fan stores a proportion of the sediment generated by upper fan incision. Nevertheless, the volume of sediment eroded exceeds that deposited, suggesting a net export of some 10000 tonnes to the Te Weraroa Stream (Table 2), although this may be an overestimation given the bias in the two surfaces (Table 1). However, some deposition also occurred in this summer period in the upper fan, prior to incision, evidenced by the positive elevation change detectable adjacent to incised channels (Fig. 5a). The sequence of events is not straightforward: the upper fan continued to aggrade in early summer (December 2004), before being incised. 4.2. April 2005-August 2005 Infilling is evident across the entire fan and channels cut in late summer have been refilled. Lateral erosion in the lower fan (true right) produces a negative elevation change (Fig. 5b). However, new channels are evident in both upper and mid-fan (Fig. 5b). This indicates a rapid cutting in response to an absence of heavy rain events, notably from July (Fig. 6), preceded by a similarly rapid filling (April-June wetter, cf. Fig. 6b). Given the absence of incision in the lower fan and a tendency to aggrade, the morphological budget 51 (a) Surveys & © 2 >< Q s3* & & & s£ s£ s£ s$ s£ s£ & & ^ $ (b) 1200.00 1000.00 E £ 400.00 1 Total rainfall -mm/day 12 ia — E > n T3 > 4 * E D 2 (0 # ^ ^ ^ ^ ^ ^ ^