Sediment continuity through the upland sediment cascade: geomorphic response of an upland river to an extreme flood event

Article history: Received 15 February 2018 Received in revised form 29 April 2018 Accepted 1 May 2018 Available online xxxx Hillslope erosion and accelerated lake sedimentation are often reported as the source and main stores of sediment in the upland sediment cascade during extreme flood events. While upland valley floodplain systems in the transfer zone have the potential to influence sediment continuity during extreme events, their geomorphic response is rarely quantified. This paper quantifies the sediment continuity through a regulated upland valleyfluvial system(St John's Beck, Cumbria, UK) in response to the extremeStormDesmond (4–6December 2015)flood event. A sediment budget framework is used to quantify geomorphic response and evaluate sediment transport during the event. Field measurements show 6500 ± 710 t of sediment was eroded or scoured from the river floodplains, banks and bed during the event, with 6300 ± 570 t of sediment deposited in the channel or on the surrounding floodplains. b6% of sediment eroded during the flood event was transported out of the 8 km channel. Floodplain sediment storage was seen to be restricted to areas of overbank flow where the channel was unconfined. Results indicate that, rather than upland floodplain valleys functioning as effective transfer reaches, they instead comprise significant storage zones that capture coarse flood sediments and disrupt sediment continuity downstream. Crown Copyright © 2018 Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Upland rivers are active geomorphic systems that generate some of the highest annual global sediment yields (Milliman and Syvitski, 1992). The steep channel gradients, high runoff and dynamic geomorphic processes result in high rates of sediment production, transfer, deposition and geomorphic change (Johnson and Warburton, 2002;Warburton, 2010). These processes are greatest during high magnitude, low frequency, extreme flood events when sediment yields can increase by orders of magnitude, even when averaged over centennial to millennial timescales (Korup, 2012;Wicherski et al., 2017). The geomorphic impacts of these extreme events such as riverbed and bank erosion (Prosser et al., 2000;Milan, 2012;Thompson and Croke, 2013), channel widening (Krapesch et al., 2011), overbank sediment deposition (Williams and Costa, 1988;Knox, 2006), floodplain scour (Magilligan, 1992) and the destruction of protection structures (Langhammer, 2010) can have significant impacts on upland river valleys and surrounding society and infrastructure (Davies and Korup, 2010). Many of these upland systems have been anthropogenically modified to minimise the geomorphic impacts of 1 in 100 yr flood events (Hey and Winterbottom, 1990;Gergel et al., 2002), but under extreme flows managed river corridors can be reactivated.
Previous research has focused on understanding the controls of such geomorphic change during extreme events to help better predict and manage the impacts. For example, studies have explored the potential for geomorphic work through magnitude-frequency relationships (Wolman and Gerson, 1978), hydraulic forces (i.e., discharge, shear stress, stream power (Magilligan, 1992;Thompson and Croke, 2013)), catchment characteristics such as valley confinement (Righini et al., 2017), the role of engineered structures (Langhammer, 2010) and anthropogenic modifications (Lewin, 2013). However, only a few studies (Trimble, 2010;Warburton, 2010;Warburton et al., 2016) have investigated the geomorphic impacts of extreme events in terms of sediment continuity of the upland catchment sediment cascade (USC). Here, sediment continuity is defined as the physical transfer or exchange of sediment from one part of the fluvial system to another, and represents the conservation of mass between sediment inputs, stores and outputs. Sediment continuity is therefore distinct from the concept of sediment connectivity (Hooke, 2003;Bracken et al., 2015) as it describes the pathways for sediment transfer by quantifying the physical movement and storage of sediment mass.
The USC describes the supply, transfer and storage of catchment sediment from source to sink (Chorley and Kennedy, 1971;Slaymaker, 1991;Burt and Allison, 2010). Fig. 1 provides a framework for the USC displaying the main sediment stores that are often characterised in upland sediment budget studies (Reid and Dunne, 1996;Fuller et al., 2002;Brewer and Passmore, 2002). The USC is adapted from Schumm's (1977) simple sediment cascade (SSC) model that divides the fluvial system into the production zone, transfer zone and deposition zone. In many upland regions however, the SSC is modified due to the presence of water bodies such as lakes, reservoirs or impoundments, which restrict sediment continuity between zones (Foster, 2010). Many of these water bodies (N40%) are the product of previous glacial activity that has scoured over-deepened basins (Herdendorf, 1982;Foster, 2010;McDougall and Evans, 2015). These basins occur both towards headwaters, between catchment production and transfer zones, as well as in lowland reaches where they form major long term depositional sites (Petts, 1979;Williams and Wolman, 1984;Kondolf, 1997). The movement of coarse sediment in and between the zones of the USC has been compared to a 'jerky conveyor belt' (Ferguson, 1981;Newson, 1997) where sediment is transferred and stored over a range of temporal scales. Sediment stores can fuel or buffer sediment transport rates and therefore influence sediment continuity and potential geomorphic change downstream; this is particularly relevant during less frequent higher magnitude events where sources and stores of sediment can rapidly change over a short period of time (Davies and Korup, 2010;Fryirs, 2013).
The USC production zone is characterised by mountain torrent and cascade channels that have steep channel slopes (N0.03-0.30) and surrounding hillslopes (N0.15-0.7) (Montgomery and Buffington, 1993). Here, channels are confined by the local valley topography and have no intervening floodplain; hillslopes are strongly (N80%) coupled to the channel (Lewin, 1981;Montgomery and Buffington, 1993;Harvey, 2001;Korup, 2005;Crozier, 2010). Sediment flux in this zone is dominated by suspended sediment, but during flood events bedload and coarse sediment stored on hillslopes can be mobilised, thus contributing to the total sediment load (Ashbridge, 1995). Hillslope erosion processes (mass wasting or water-driven) are the principal sources of sediment, which is deposited either on the hillslopes or in the channel (Montgomery and Buffington, 1993;Fuller et al., 2016). Previous studies have explored sediment dynamics in the USC production zone including: (i) hillslope-channel coupling relationships (Harvey, 2001(Harvey, , 2007Johnson et al., 2008;Smith and Dragovich, 2008;Caine and Swanson, 2013), (ii) variability in sediment supply, transfer and deposition (Johnson and Warburton, 2006), (iii) response of these systems to extreme flood events (Johnson and Warburton, 2002) and (iv) the relative contribution of sediment sources to the channel through sediment budgeting approaches (Warburton, 2010).
In contrast, in the transfer zone ( Fig. 1), sediment sources and deposits differ from those of the production zone as the channel (or piedmont channel) gradient decreases (slopes of b0.001-0.03), floodplain width increases, and the channel becomes unconfined allowing greater channel-floodplain interaction (Lewin, 1981;Church, 2002). Hillslope erosion processes are disconnected from the active channel by floodplains and therefore do not contribute directly to channel sedimentation (Lewin, 1981;Church, 2002). Instead, sediment in this zone is sourced from tributary inputs and reworked from channel bed and bank deposits. Suspended sediment dominates the low to medium flow sediment fluxes, with bedload sediment stored in the channel only mobilised at 50-60% of bankfull flow (Carling, 1988;Knighton, 1998;Fuller et al., 2002). Only during overbank flow is the largest bedload sediment entrained in quantity in this zone (Carling, 1988). Sediment continuity in the transfer zone is heavily influenced by anthropogenic modifications to the system (Fryirs et al., 2007;Lewin, 2013). The presence of upstream reservoirs or impoundments disrupt coarse sediment supply from headwaters, and influence the potential for sediment transport downstream through flow regulation (Petts and Thoms, 1986;Kondolf, 1997). Many of these systems have become "genetically modified" over time (Lewin, 2013) with channels artificially confined by flood protection structures to safeguard adjacent land, reducing channel-floodplain interactions. Consequently, sediment continuity and potential for sediment storage on the floodplains during extreme flood events is heavily modified by anthropogenic activity (Wohl, 2015).
Previous research has discussed the impacts of lakes, dams and impoundments on downstream sediment transport in the USC transfer Fig. 1. The upland sediment cascade (USC) framework displaying sediment stores and the relative sediment continuity through each store during non-flood conditions. The USC framework is modified from Schumm's (1977) Simple Sediment Cascade model. zone (Gurnell, 1983;Kondolf, 1997;Petts and Gurnell, 2005). More recently, Sear et al. (2017) modelled the response to the 2009 and 2015 Cumbria floods on the Lower River Derwent, downstream of Bassenthwaite Lake, showing how the modified confined channel reverted to a course dictated by the wider valley morphology. However, the continuity of sediment transfer through intervening modified valley systems has only rarely been directly surveyed or evaluated in detail after extreme flood events (i.e., Johnson and Warburton, 2002;Warburton, 2010) and few studies have looked at how these systems recover following these extremes (Milan, 2012).
Understanding sediment continuity during extreme events in upland valley systems will become increasingly important for hazard management given projected increases in winter precipitation from predicted climate change (Raven et al., 2010;van Oldenborgh et al., 2015). However, extreme flood events are difficult to predict (Lisenby et al., 2018) and there are few direct measurements from these events. Consequently, their impacts have to be inferred from historical information and estimates of the quantity of sediment stored and transported are generally poorly constrained.
This paper quantifies the geomorphic response of an upland river valley system (transfer zone) to Storm Desmond, an extreme flood event that hit Cumbria, Northwest UK in December 2015. Specifically we (i) quantify the geomorphic impacts of the extreme event on the upper floodplain valley system of the USC; (ii) estimate bedload sediment transport rates during the flood; (iii) evaluate system recovery one year after the flood event and (iv) place findings within the wider context of sediment continuity through the USC. This study is the first to quantify the role of the floodplain zone in the USC in response to an extreme event and thus will enable better understanding of sediment continuity in upland regions.

Study site
This study focused on St John's Beck, an 8 km channelised, regulated gravel bed river downstream of Thirlmere Reservoir, Central Lake District, UK (OS National Grid Reference (NGR): NY 318203, catchment area including Thirlmere Reservoir is 53.4 km 2 , effective catchment area is 12 km 2 ) (Fig. 2a). St John's Beck is a tributary to the River Greta that flows through the town of Keswick before discharging into Bassenthwaite Lake (area = 5.1 km 2 ). St John's Beck ranges in altitude from 178 m OD at the Thirlmere Reservoir outlet to 130 m OD where it joins the River Greta (Fig. 2a). St John's Beck lies in the upper floodplain transfer zone of the USC (Fig. 2b). The channel has a Strahler (1952) stream order of 3, mean channel slope of 0.005 and mean channel width of 12 m. St John's Beck lies in a previously glaciated valley (Vale of St John's) that is underlain by Ordovician Borrowdale Volcanic rocks in the north of the catchment and the Skiddaw group in the south. The land surrounding the channel is predominantly mixed woodland and pasture used for livestock grazing. St John's Beck is a Site of Special Scientific Interest and lies in the Derwent and Bassenthwaite Lake Special Area of Conservation. The river is protected to support salmon, lamprey species, otters and floating water plantain (Wallace and Atkins, 1997;Reid, 2014).
St John's Beck has a wandering planform which has been restricted laterally due to channelisation in the late nineteenth century following the impoundment of Thirlmere Reservoir (area = 3.3 km 2 ). The channel is confined by the natural valley topography in the upstream reaches. Floodplain valley width increases 1.8 km downstream from Thirlmere Reservoir (Fig. 2a), however the river channel has been modified and restricted from movement here (1.8-5 km downstream) through bank reinforcement and flood protection levees. Flood protection levees were built to protect farmland and a major link road from flooding. Long term flow regulation has influenced sediment transport rates in St John's Beck and as a result the system displays clear zones of aggradation. There are four first order tributaries that flow into St John's Beck. Flow and sediment are intercepted from two of these tributaries, which drain the Helvellyn mountain range and are directed to Thirlmere Reservoir (Reid, 2014;Bromley, 2015). The third and fourth first order tributaries are constrained by the presence of a road and a sediment trap and therefore are not a major source of sediment to St John's Beck.

The Storm Desmond flood event
Extreme flood events in the Lake District have been documented from 1690 to the present (Watkins and Whyte, 2008) (recent floods summarised in Table 1). This study describes the geomorphological impacts of the Storm Desmond (4-6 December 2015) flood event. Storm Desmond, a North Atlantic storm, was associated with a mild and moist slow moving low pressure system located northwest of the UK that brought severe gales and exceptionally persistent heavy rainfall over northern UK (Met Office, 2016). Northern England experienced the wettest December on record (in a series from 1910), following the second wettest November, after 2009(McCarthy et al., 2016. The average December rainfall doubled in northern England, with the Lake District receiving three times its average monthly rainfall (McCarthy et al., 2016). Storm Desmond produced record-breaking rainfall maximums in the UK: 341.4 mm rainfall was recorded in a 24 h period at Honister Pass (NGR NY 225134), Western Lake District, and 405 mm of rainfall was recorded in a 38 h period at Thirlmere (study catchment), central Lake District (NGR NY 313194). The storm was the largest in the 150 yr local Cumbrian rainfall series (1867-2017), and exceeded previous records set in the 2005 and 2009 Cumbrian floods. The estimated return period for the rainfall event was 1 in 1300 years (CEH, 2015) based on the FEH13 rainfall frequency model (Stewart et al., 2014). The UK climate projection change scenarios for northwest England predict winter flood events like this will occur more often in the future because of increases in rainfall intensity due to climate change (Watts et al., 2015).

Storm Desmond impacts
Storm Desmond caused widespread disruption across northern England, and in particular in upland areas in the Lake District region. The event captured national attention when extreme weather conditions prompted a full scale emergency response to extreme flooding, erosion and sediment movement by upland rivers. Over 5000 homes were flooded, access routes were destroyed (257 bridges destroyed) and key infrastructure was affected, including the erosion of the main A591 trunk road through the central Lake District. The latter was estimated to cost the local economy £1 million per day (BBC, 2016). In the production zone of the USC, saturated hillslopes and high porewater pressures triggered landslides in a number of valleys, with sediment eroded and transported through mountain torrents (Warburton et al., 2016). Geomorphic impacts in the upper floodplain system of the USC included the erosion of riverbed and banks, floodplain scour, scour around man-made structures (bridges, levees) and extensive deposition of coarse sediment across floodplains. Storm Desmond caused severe flooding and substantial geomorphic change along St John's Beck (Fig. 3).

Hydrological regime in St John's Beck
Flooding is not unusual in St John's Beck, historic accounts describe a "most dreadful storm… with such a torrent of rain, [which] changed the face of the country and did incredible damage in [St John's in the Vale]" in 1750, (Smith, 1754). This historical event has characteristics similar to that of Storm Desmond, with large boulders of sediment being transported and deposited on floodplains along the transfer zone. Long term rainfall records available for the St John's Beck Catchment

Methods
This study analyses geomorphic data collected during two field campaigns at St John's Beck. The first survey was completed after the Storm Desmond flood (April-May 2016) to capture the geomorphic impacts of this event before clean-up operations and reworking of flood sediments occurred. The second survey was conducted in June 2017 to assess short-term system recovery following the flood. All field data were digitised and analysed in a GIS in British National Grid coordinates. A 5 m resolution digital elevation model (DEM) (Digimap, 2016), preflood aerial imagery, 2009-2011, (from Bluesky International Limited, resolution 0.25 m) and post-flood event, May 2016, (from the Environment Agency, resolution 0.2 m) were used for validating field measurements and to assess valley topographic and local controls of the geomorphic impacts observed.

Channel geometry and bed material
A Leica Geosystems Real Time Kinetic differential GPS (RTK dGPS) 1200, was used to survey channel cross section geometry, floodplain geometry and thalweg long profile during the 2016 and 2017 surveys. Cross section sites were chosen along the 8 km river where there was a clear change in channel geomorphology identified by a walk-over reconnaissance of the catchment in 2016. A total of 22 sites for cross section surveys were chosen along St John's Beck. Cross section 1 was located near the St John's Beck gauging station (1 km downstream from Thirlmere Reservoir), so all data collected could be discussed in relation to the flow and rainfall records (Figs. 4b, c, and 5). The last cross section was located near the confluence with the River Greta (7.8 km downstream). Ten of the cross section sites were located along a 1.3 km length reach where significant riverbank erosion and overbank flood sediment deposition occurred during Storm Desmond. Survey pegs were positioned at the endpoints of each cross section in 2016 and used as control points to allow resurvey in 2017. Cross section profile RTK dGPS measurements had a mean accuracy of ±0.02 m and standard deviation of 0.06 m in the 2016 survey, and a mean accuracy of ± 0.03 m and standard deviation of 0.03 m in the 2017 survey. Bankfull channel cross-sectional area was calculated at each cross section and changes in channel bankfull capacity (m 2 yr −1 ) were calculated by differencing the data collected over the survey periods. Thalweg long profile was surveyed using the RTK dGPS. Average profile point spacing was 8 m (mean accuracy of ±0.02 m and standard deviation of 0.01 m) in the 2016 survey and 12 m (mean accuracy of ±0.03 m and standard deviation of ±0.01 m) in the 2017 survey. Channel surface bed material was measured at each cross section following the pebble count method for grain size distribution (GSD) in the 2016 and 2017 field campaigns. The b-axis of 100 particles were randomly measured (particle under tip of the toe method; Wolman, 1954) along the width of each cross section. The median diameter grain size (D 50 ) and the 90th percentile (D 90 ) were calculated and used to understand system response and sediment transfer following the event.

Bedload transport
Bedload sediment transport during Storm Desmond was estimated using the Bedload Assessment for Gravel-bed Streams (BAGS) software (Pitlick et al., 2009) applying a surface-based bedload transport equation (Wilcock and Crowe, 2003). The input parameters were: the GSD of the channel bed surface, cross-sectional data including floodplains, cross section averaged bed elevation slope, flow discharge in the form of a flow exceedance curve for the event, and Manning's "n" values for a clean winding channel (0.04) and short grass floodplains (0.03) estimated from Chow (1959). Sensitivity to Manning's "n" values was assessed using Chow (1959) minimum and maximum values for the channel and floodplains. Morphological change between cross sections was calculated by subtracting the downstream cross section bedload transport rate from the upstream value to identify net erosion and deposition reaches.
Historical bedload sediment transport rates were also estimated using the BAGS model (i) as an average daily transport rate for the long-term daily discharge record 1935-2015, and (ii) for the top five discharge events in the long term (15 min interval) flow record. Whilst we assume that the cross-sectional profiles and grain size distribution are the same as the post-Desmond channel, this analysis allows us to assess the importance of the Storm Desmond event on sediment transport rates in relation to the longer term system history.

Geomorphic impacts of the Storm Desmond event: sediment budget analysis
A sediment budget framework was used to quantify the geomorphic impacts of the Storm Desmond event and identify the dominant stores of sediment along St John's Beck. Sediment budgets focus on quantifying the erosion, deposition and transfer of sediment through a channel or reach over an event or time period (Reid and Dunne, 1996;Brewer and Passmore, 2002;Fuller et al., 2003). Sediment budgets represent the conservation of mass and can be summarised as (Slaymaker, 2003): where O s is the sediment output (yield) of the reach, I s is input of sediment from dynamic sediment sources, and S s is sediment stored on floodplains, channels etc. This framework is useful to understand local sediment continuity in response to a particular event and indicate whether a system is balanced (Reid and Dunne, 2003). The main geomorphic depositional (S s ) and erosional (I s ) features identified after Storm Desmond along St John's Beck were: floodplain sediment deposits, in-channel bars, floodplain scour, channel bed scour and riverbank erosion (Fig. 3). Floodplain scour is differentiated from bank erosion as it is associated with the stripping of the floodplain surface (vegetation) and removal of large blocks of sediment (Nanson, 1986); whereas bank erosion is defined as the removal of sediment from the bank by hydraulic action or through mass failure (Odgaard, 1987;Knighton, 1998). The volume and sediment size distribution of erosional and depositional components were measured using the RTK dGPS, and pebble count technique (Wolman, 1954) and their spatial extent was validated using the pre-and post-event aerial photographs. Channel bed scour was active during the event, however, it was not directly measured as no cross sections were monumented prior to Storm Desmond. During flood events some reaches can experience scour whilst other reaches aggrade (Reid and Dunne, 1996). The location of channel bed scour was assumed to occur where riverbank erosion or floodplain scour was observed after Storm Desmond; this was quantified using the post-event air photo and field data in GIS. The depth of channel bed scour was estimated according to Carling's (1987) scourdepth relation for gravel bed rivers: where d s is depth of scour (m) and Q is the event peak discharge (m 3 s −1 ). Volumes of sediment eroded and deposited for each geomorphic component were converted to sediment mass using local values of coarse sediment bulk density of 1860 ± 17 kg m 3 derived from the mean bulk density of 30 measured samples from the channel bed and floodplain sediment deposits.
Sediment input and output of St John's Beck during the event was estimated by converting the BAGS estimated event bedload sediment transport rates into (cross section 1, 1 km downstream) and out of St John's Beck (cross section 22, 7.8 km downstream) into the event sediment yield.
Error in sediment budgets represents a combination of survey measurements and calculations, so standard methods of error analysis are difficult to apply. Often, sediment budget error is calculated as an unmeasured residual by subtracting the erosion and deposition components (Kondolf and Matthews, 1991;Reid and Dunne, 2003). As a result, sediment budgets may balance only because errors are hidden in the residual terms (Kondolf and Matthews, 1991). To avoid misrepresentation of the sediment balance, in this study the standard error was calculated for each measurement technique for each geomorphic component. The standard errors were summed and then converted to a percentage before being converted to mass (t) for each component. For example, floodplain deposit mass error represents a combination of errors from the RTK dGPS, depth of deposit, and bulk density error measurements. The standard error from these measurements was calculated and then summed to calculate the total error percentage before being converted to the mass error (t).

Lateral channel confinement ratio
Channel confinement describes the extent to which topography, such as hillslopes, river terraces and artificial structures, limit the lateral mobility of a river channel (Nagel et al., 2014). Lateral channel confinement ratio (C) was calculated as: where w f is the floodplain width and w c is the active channel width. Floodplain width (pre-and post-Storm Desmond) is defined as the horizontal distance from the top of the channel bank to the base of the hillslope (Gellis et al., 2017); this is determined using the 2009-2011 and 2016 aerial photographs, the 5 m resolution DEM and the 2016 field data. The active channel width was measured (1) prior to Storm Desmond using the 2009-2011 aerial photographs, and (2) after Storm Desmond using the RTK dGPS channel cross section measurements and May 2016 aerial photographs. Channel and floodplain width were measured at the 22 cross section sites. Hall et al. (2007) documented that confined channels have a confinement ratio of ≤3.8 and unconfined channels a ratio of N3.8. Channel confinement can influence the potential for sediment erosion and deposition; for example, Thompson and Croke (2013) found that in a high magnitude flood event in the Lockyer Valley, Australia, erosion was concentrated in the confined reaches, and deposition was concentrated in unconfined reaches with floodplains acting as a major store of sediment. Such behaviour may be affected by the presence of structures such as  (2) artificial confinement, where reaches of the channel have been modified through reinforced riverbanks, the presence of walls, levees, or road embankments that prevent the channel from migrating laterally; and (3) the post-Storm Desmond confinement taking into consideration the active channel width following the extreme event.

Stream power and shear stress
At the reach scale average shear stress, Eq. (4) (Du Boys, 1879), critical shear stress, Eq. (5) (Gordon et al., 1992), unit stream power, Eq. (6) (Bagnold, 1966) and critical unit stream power Eq. (7) (Bagnold, 1966;Williams, 1983;Petit et al., 2005) were calculated for the Storm Desmond flood to understand the potential magnitude of sediment transport rates and geomorphic impacts observed during the event using the one-dimensional uniform flow approximations: where τ is the reach averaged shear stress (N m −2 ), ρ is the density of water (kg m −3 ), g is the acceleration of gravity (m s −2 ), S is channel bed slope (m m −1 ) and d is the maximum water depth during the event (m). τ c is the critical shear stress (N m −2 ) and D i is the grain size (mm). Here we use the channel D 50 and D 90 . ω is the unit stream power (W m −2 ), Q corresponds to the peak discharge (m 3 s −1 ) during Storm Desmond and w (m) is the bankfull width during the flood. ω c is the critical unit stream power (W m −2 ) for particle motion based on Williams' (1983) relation for gravel transport in rivers with grain sizes between 10 and 1500 mm. Calculations were applied at the cross section locations and the critical shear stress (τ N τ c ) and critical stream power (ω N ω c ) entrainment thresholds estimated to understand the potential for sediment mobility during the event. Shear stress and stream power calculations were also calculated using the June 2017 survey data (bankfull cross section profiles, grain size data, and mean daily discharge (0.085 m 3 s −1 ) to quantify variation in shear stress and stream power during non-overbank flows.

Geomorphic response to the storm Desmond event
Storm Desmond flood impacts along St John's Beck were concentrated in the channel and on the surrounding floodplains. The spatial distributions of both erosional and depositional impacts of Storm Desmond are shown in Fig. 6a. Generally, erosion and deposition impacts were observed in spatially similar locations, for example, where bank erosion or scour occurred overbank deposition was observed. Significant erosion and deposition impacts were observed 1.7-3.6 km downstream of Thirlmere Reservoir (Fig. 6b). Geomorphic impacts were less pronounced 3.6-8 km downstream of Thirlmere Reservoir; impacts here were often concentrated locally at meander bends (e.g., as seen at 5.2 km downstream from Thirlmere Reservoir, cross section 18). Fig. 6b shows a detailed map of the reach where significant geomorphic impacts (1.7-3.6 km downstream) were observed after Storm Desmond. Overbank floodplain deposits and channel bars measured 2.1-2.5 km downstream (between cross sections 7 to 10) occur where the channel is laterally unconfined. The channel in this reach (2.1-2.5 km downstream) was identified as aggradational (low channel capacity, channel bed nearly level with banks) in a reconnaissance survey (approach after Thorne, 1998) of the site prior to the flood. Bank erosion and scour was concentrated on the artificially-confined reach 2.5-3 km downstream (cross sections 10 to 13). Local lateral riverbank recession exceeded 12 m and caused the destruction of flood protection levees 2.7 km downstream of Thirlmere Reservoir (see cross section 11 Fig. 6b). Material eroded at cross section 11 was subsequently deposited on the floodplains downstream.
The dominant geomorphic features surveyed after the event were overbank floodplain sediment deposits. Floodplain sediment deposits located 1.8 km downstream (near cross section 5) were sourced from a tributary and not from St John's Beck. The tributary sediment did not enter St John's Beck due to a wall and sediment trapping structure, therefore, the mass of sediment measured here (300 t) is excluded from the sediment budget analysis. A total of 105 floodplain deposits were identified from St John's Beck, equating to a sediment mass of 4700 ± 300 t. Flood sediment deposits were generally composed of a single layer of sediment with a mean deposit depth 0.09 m ± a standard deviation of 0.07 m; the maximum flood deposit depth measured was 0.3 m located 2.7 km downstream of Thirlmere Reservoir. The mean grain size of sediment deposit D 50 was 32 mm and D 90 was 90 mm. The 10 largest clasts from the deposits had a mean grain size of 147 mm ± a standard deviation of 12.5 mm. Flood deposit grain size decreased with distance from the channel. The farthest flood deposit from the channel bank (70 m distance) had a D 50 of 22 mm and D 90 of 63 mm. The proximal flood deposits (2 m distance from the channel) had a mean D 50 of 39 mm ± a 17 mm standard deviation and D 90 of 111 mm ± a standard deviation of 35 mm. Table 2 shows the variation in grain size between the flood sediment deposits and the channel bed sediments. Channel bed sediment D 50 is greater than the floodplain sediment deposits, however, this pattern is reversed for sediment D 90 . Floodplain sediment deposits are composed of material from the channel bed and from eroded features (such as artificial levees and stone walls), which generally have coarser grain sizes that could account for this variation.
Riverbank erosion and floodplain scour were the main processes accounting for a loss of sediment during Storm Desmond. Based on the field data collected, 2300 ± 270 t of sediment was eroded from the riverbanks. Floodplain scour contributed to the removal of 1300 ± 50 t of sediment during the event, 40% of sediment removed through scour was over the reach (2.2-3.6 km downstream) where significant sediment deposition was observed. Local scour of 350 ± 13 t undermined and destroyed the access bridge to Low Bridge End Farm (see crosssection 10, 2.5 km downstream of Thirlmere Reservoir, Fig. 6). The depth of channel bed scour was estimated at 0.13 m according to Carling's (1987) scour depth equation, and this equated to a mass of 2900 ± 470 t. Fig. 7 displays the total mass of sediment eroded and deposited along St John's Beck during Storm Desmond. The greatest mass of sediment eroded and deposited occurs from 1.7 to 3.6 km downstream where the floodplain width increases from 7 to 450 m and channel  (Fig. 3c) and sediment deposition was observed on the inside of a meander bend 5.2 km downstream where overbank flows were permitted during Storm Desmond. Local bank erosion and overbank sediment deposition was observed on bends 6.8 and 7.3 km downstream. Tree debris were observed surrounding St John's Beck following Storm Desmond. Tree debris did not cause a blockage around the access bridge to Low Bridge End Farm. However, tree debris were observed in the channel near cross section 10 (2.5 km downstream) (see Fig. 3b). The limited occurrence of woody debris in the channel inhibits the formation of log jams and only has local impacts on sedimentation.

Estimates of bedload sediment transport rate
The mean event bedload sediment transport rate for the 22 cross sections was 160 t ± a standard error of 60 t. Sediment transport rates fluctuate downstream with clear reaches of low and high sediment transfer (Fig. 8a). For example, 1.5-2 km downstream of Thirlmere Reservoir high sediment transport rates during the event (range = 220-500 t) are estimated; these are attributed to a local increase in channel slope. The maximum estimated transport rate during the event was 1200 t at 2.5 km downstream of Thirlmere Reservoir where the channel widens and local slope increases (slope 0.01) downstream of a ford, near the access bridge to Low Bridge End Farm that was destroyed during the event (Fig. 3f). The sediment input into St John's Beck during the event is estimated at 7 t (1 km downstream of Thirlmere Reservoir, cross section 1) and the sediment output (7.8 km downstream of Thirlmere reservoir, cross section 22), during the event is estimated as 370 t.
Zones of erosion and deposition along St John's Beck have been identified by differencing sediment transport rates between the surveyed cross sections (Fig. 8b). A total of 10 deposition and 11 erosion zones are defined. The zone of greatest erosion and deposition is located from 1.8 to 4 km downstream from Thirlmere Reservoir (Fig. 8b), which corresponds closely with field measurements of erosion and deposition during the event (Fig. 6).
The mean daily bedload sediment transport rate (calculated as the mean transport rate from the 22 cross sections using the 1935-2015 discharge record), is 0.05 t day −1 with a standard deviation of 0.09 t day −1 . The estimated annual bedload sediment input is estimated at 0.5 t yr −1 (at cross section 1) and the bedload sediment yield (at cross section 22) is 38 t yr −1 for St John's Beck long term discharge record. The bedload sediment output during Storm Desmond (370 t) exceeds the annual value by a factor of 9. Table 3

Channel confinement index
St John's Beck displays different degrees of lateral confinement downstream (Fig. 9). The natural channel confinement pattern shows that the channel becomes gradually unconfined downstream (Fig. 9). For example, in the upstream reach (0 to 1.8 km downstream of Thirlmere Reservoir) the channel is topographically confined (confinement ratios range from 0.1 to 0.6) and from 4.4 to 8 km downstream the channel is topographically unconfined (confinement ratios range  from 5 to 65). The channel has been artificially confined from 1.8 to 4.4 km downstream by flood protection levees, reinforced banks and walls that restrict lateral channel movement. The mean natural floodplain width has been reduced by 90% due to the presence of artificial structures along the artificially confined reach 1.8 to 4.4 km downstream. During Storm Desmond, many of the artificially-reinforced banks and flood protection levees were scoured or eroded increasing the active channel width and allowing channel-floodplain interactions (Fig. 9). After Storm Desmond the mean confinement ratio increased from 0.95 to 17 along the artificially confined reach (1.8 to 4.4 km downstream), indicating the system reverted to a natural floodplainchannel width relationship (Fig. 9).

Shear stress and stream power
Shear stress and stream power are used to understand the energy expenditure for erosion and sediment entrainment during the event (Fig. 10). The shear stress values estimated for Storm Desmond are shown in Fig. 10a. The shear stress values estimated should be regarded as minimum values because they assume shear stress is the same on the channel and floodplain and the equations assume steady uniform flow, which was unlikely during the event. The mean shear stress value is 149 N m 2 with a standard deviation of 78 N m 2 . The peak shear stress value (426 N m 2 ) was estimated 2.7 km downstream of Thirlmere Reservoir; near where the access bridge was destroyed and mass overbank coarse sediment deposition occurred. The minimum shear stress values are estimated 1.1 to 1.3 km downstream (30-60 N m 2 ) where local slope is 0.001. The mean shear stress value exceeded the mean critical entrainment thresholds for particle D 50 (48 ± a standard deviation of 14 N m 2 ) and D 90 (124 ± a standard deviation of 30 N m 2 ) (Fig. 10a), suggesting full mobility of the GSD during the event. The mean shear stress value estimated using the 2017 survey data (62 N m 2 with a standard deviation of 40 N m 2 ) does not exceed the threshold for mean particle D 90 (114 N m 2 ) entrainment and only exceeds 60% of the cross section particle D 50 entrainment threshold during bankfull flow conditions.
The unit stream power values estimated along St John's Beck using the peak Storm Desmond discharge value range from 25 to 354 W m −2 , with a mean of 230 W m −2 and a standard deviation of 132 W m −2 (Fig. 10b). The values are within the range of stream power values documented for those causing erosion during flood events and sediment transport (Baker and Costa, 1987;Magilligan, 1992;Fuller, 2008;Marchi et al., 2016). A value of 300 W m −2 is commonly referred to as a threshold for producing floodplain erosion (Baker and Costa, 1987;Magilligan, 1992;Fuller, 2008). Significant erosion and Table 3 Bedload sediment transport estimates for the top five discharge events from the 15 min interval flow series data for St John's Beck. The event bedload transport rates are calculated as the mean transport rate from the 22 cross sections, and the event sediment yield is calculated at cross section 22. scour was observed 2.5 km downstream where an access bridge was destroyed and where stream power was estimated at 420 W m −2 . The mean unit stream power estimate (230 W m −2 ) exceeds the critical unit stream power value for particle D 50 (13 W m −2 ) and D 90 (54 W m −2 ) entrainment, suggesting mobilisation of the coarsest grains. The mean unit stream power, estimated using the 2017 data and mean daily discharge, is 0.26 W m −2 ± a standard deviation of 0.12 W m −2 ; this value does not exceed the critical stream power threshold for channel bed particle D 50 and D 90 entrainment.

System resurvey in 2017
Resurveys of St John's Beck longitudinal profile, cross section profiles and grain size in 2017 provide an indication of how the system is recovering 1.5 yr after the extreme flood event (Fig. 11). There were no significant changes in the mean channel bed slope between the 2016 and 2017 survey, however, there were local changes where there is an increase or decrease in bed elevation height (Fig. 11a). Local changes in channel bed elevation result in changes in bankfull channel capacity (Fig. 11b). For example, at a distance of 1 to 2.4 km from Thirlmere Reservoir there is a general increase in bed elevation suggesting the deposition of sediment; a pattern further evidenced by a decrease in channel capacity. Overall a decrease in bankfull channel cross-sectional area was observed (at 15 cross sections) 1.5 yr after Storm Desmond. Thirteen of these cross-sections are located 1 to 2.7 km downstream from Thirlmere Reservoir (Fig. 11b). The largest change and reduction in channel capacity (2.7 km downstream of Thirlmere Reservoir, cross section 11) was 32.8 ± 0.03 m 2 caused by the rebuilding of flood protection levees that reduced channel width to its pre-Storm Desmond size. A total of seven cross-sections displayed either no change or an increase in cross-sectional area and channel capacity. Cross-section 9, 2.4 km downstream from Thirlmere Reservoir, shows an increase in channel capacity associated with anthropogenic removal of sediment from the channel bed after the flood event. The percentage change in grain size between the 2016 and 2017 surveys illustrates a general coarsening of bed D 50 and fining of D 90 downstream post Storm Desmond (Fig. 11c).
6. Discussion 6.1. Geomorphic impacts of the extreme flood event along the upland sediment cascade The 2015 Storm Desmond event constitutes the largest recorded event in the available long term flow and rainfall records for the St John's Beck catchment (Fig. 5). The results presented here illustrate the geomorphic work of the flood in terms of sediment erosion and storage along the upper floodplain transfer zone of the USC. The main impacts were associated with erosion of river channel banks and floodplain scour allied with extensive sediment deposition on the floodplains. The summary sediment budget (Fig. 12) shows erosion (6500 ± 710 t) was generally balanced by deposition (6300 ± 570 t) along the upper floodplain zone. b6% of the total sediment eroded during the event was transferred out of the reach. Hence, the upper floodplain zone acted as a significant sink for locally-eroded sediment during the extreme event.
The geomorphic impacts of Storm Desmond were influenced by the physical characteristics of the upper floodplain transfer zone. Unlike steep headwater catchments dominated by slope-channel linkages and hillslope processes (Harvey, 2001), geomorphic impacts of the event along St John's Beck were controlled by floodplain-channel interactions. Tributaries were only a minor source of sediment as these were disconnected from the channel by sediment trapping structures and therefore are not reported in the sediment budget in Fig. 12. Sediment was sourced from transient stores, i.e., channel bars) and through erosion of the channel bed and banks and stored in channel bars and on the surrounding floodplains (Fig. 6).
Valley confinement (natural and artificial) controlled the spatial positioning of erosional and depositional storm impacts along St John's Beck (Fig. 9). In the upstream reaches (0 to 1.8 km downstream) the channel was confined by the natural valley topography and geomorphic impacts were comprised of local erosion or sediment bar deposition. Where the natural floodplain valley width increases from 3 to 160 m (1.8 km downstream) and there is an associated decrease in channel slope, rapid floodplain sediment deposition occurred (Fig. 7). In contrast, artificially confined reaches (2.7 to 3.6 km downstream) were associated with bank erosion or scour due to local increases in channel bed slope. Major riverbank erosion was observed along an artificially confined reach 2.7 km downstream of Thirlmere Reservoir; here riverbanks were eroded until the channel became unconfined (Fig. 9) with extensive floodplain sedimentation. Similar effects have been observed by Magilligan (1985), Nanson (1986), Butler and Malanson (1993), Lecce (1997), Fuller (2007, 2008, who all identified a concentration of erosion on constricted reaches. The transition between confined and unconfined reaches therefore plays an important role in controlling the spatial pattern of erosion and deposition impacts of these events.

Sediment continuity through the upland sediment cascade
The sediment continuity concept focuses on the principle of mass conservation of sediment within a system (Slaymaker, 2003;Hinderer, 2012). The USC sediment continuity has been described as a 'jerky conveyor belt', where sediment can spend a longer time in storage than in  transfer (Ferguson, 1981;Walling, 1983;Newson, 1997;Otto et al., 2009). This study has highlighted that sediment continuity is disrupted or "discontinuous" at the event scale due to storage. b6% of sediment eroded during Storm Desmond was transported out of St John's Beck (Fig. 12). Elsewhere, sediment budget studies have shown similar inefficiencies in sediment transfer, often referring to this as the 'sediment delivery problem' (Trimble, 1983;Walling, 1983;Phillips, 1991;McLean et al., 1999;Fryirs, 2013). For example, in the Coon Creek Basin, USA, b7% of sediment left the basin between 1853 and 1977 (Trimble, 1983). In the River Coquet, UK, annual sediment budget within-reach sediment transfer was identified but there was minimal net export of sediment downstream (Fuller et al., 2002). In three UK upland catchments, Warburton (2010) demonstrated sediment transfer is inefficient in the production zone by comparing sediment budgets on an annual, landslide event and flood event timescale. Despite variations in catchment area and the timescale of enquiry, these examples demonstrate there is attenuation of sediment downstream due to sediment storage. This study highlights the importance of the floodplain as a major store of sediment at the event scale causing sediment attenuation at the channel outlet.
The Storm Desmond event sediment yields were higher than estimated sediment yields for previous flood events along St John's Beck (Table 3), indicating the event was significant in generating and transporting large quantities of sediment downstream. The estimated mean shear stress and unit stream power values for Storm Desmond exceeded the thresholds for particle entrainment, suggesting sediment on the channel bed was mobilised and transported during the event (Fig. 10). Despite this, the event sediment yield is lower than the total quantity of sediment eroded. Sediment transfer during extreme events, where overbank flows are produced, is reduced on the floodplains (because of variations in roughness, slope, local topography) compared to the channel, resulting in sediment deposition (Trimble, 1983;Moore and Newson, 1986). Consequently, sediment continuity through the upper floodplain transfer zone during extreme events will ultimately be controlled by the conveyance of sediment across floodplains, and the propensity for sediment deposition during overbank flows. Future flood events may promote exchanges in sediment stores and movement of sediment downstream in pulses or waves, thereby influencing sediment yield (Nicholas et al., 1995). However, if a future similar magnitude event were to occur along St John's Beck, it is likely that the reach sediment output would again be lower than the total sediment eroded along the river corridor due to deposition on the floodplains.
Previous studies have described the potential linkages between sources and stores of sediment in terms of connectivity or disconnectivity (Hooke, 2003;Fryirs, 2013;Bracken et al., 2015). However, few of these studies have quantified the mass exchange of sediment between different landscape units during flood events (Thompson et al., 2016) and assessed their impact on sediment yield. This study is among the first to effectively quantify sediment attenuation in the upper floodplain zone of the USC during an extreme event.

System recovery
Fluvial systems can take decades (Wolman and Gerson, 1978;Sloan et al., 2001) to millennia (Lancaster and Casebeer, 2007) to recover from extreme events, with some systems never fully recovering to the preflood condition. The channel re-survey one year after Storm Desmond showed that 70% of cross sections had a reduced channel capacity reflecting sediment aggradation in the channel (Fig. 11). A reconnaissance survey prior to Storm Desmond identified distinct reaches of sediment aggradation in the system (in particular, 2 to 2.5 km downstream of Thirlmere Reservoir), suggesting the river is displaying characteristics similar to the pre-flood system. Long term flow regulation and upstream sediment trapping by Thirlmere Reservoir has influenced sediment continuity, implying that the sediment regime is already disturbed by the legacy of anthropogenic modification (Wohl, 2015). Phillips (1991) states that stores of sediment may develop in fluvial systems so the system can maintain sediment yields when sediment from upstream is reduced. The critical shear stress and critical stream power entrainment thresholds for channel bed particle D 90 estimated using the 2017 survey data were not exceeded during daily flows after storm Desmond indicting coarse sediment immobility. It is likely that the finer material was transported in 2017 and deposited downstream in aggradational zones where channel dimensions change (i.e., reduction in slope, width and depth), resulting in further aggradation downstream and apparent coarsening in reaches where the fine sediment was partially mobilised. Therefore local aggradation observed could be a response to long-term system disturbance and transport-limited flows. The most significant changes observed along St John's Beck one year after the flood were associated with anthropogenic modifications to the system through the rebuilding of flood protection levees, reinforced river banks and removal of sediment from the channel bed and floodplains (2 to 4 km downstream); these modifications took place after the 2016 field campaign. Distal floodplain deposits were located 70 m from the channel and therefore can only be remobilised during overbank flows with similar peak discharges where the critical entrainment thresholds are exceeded. Consequently, system recovery and sediment transfer depends on the conveyance capacity of the valley floodplains in addition to the stream channel capacity (Trimble, 2010). If sediment was not anthropogenically removed from the floodplains, it would have a long residence time in this store and only be remobilised during overbank extreme flows similar to Storm Desmond. Flood levees were rebuilt 2.7 km downstream to the pre-flood position, it is likely that if these levees were not restored the river would permanently occupy the post-Storm Desmond position; a natural "re-wilding" process (Fryirs and Brierley, 2016).

Conclusions
This paper has quantified the geomorphic response of an upper floodplain river system (transfer zone) to an extreme high magnitude flood event: Storm Desmond, 2015. The results highlight that sediment continuity along upland rivers is complex and to fully understand the response of these systems to extreme events, sediment continuity in the context of the upland sediment cascade needs to be understood (Fig. 1). Based on our results, the primary conclusions of this work are: 1. Sediment continuity through the upper floodplain transfer zone was highly disrupted during Storm Desmond, with b6% of the eroded sediment being transported out of the system. 2. Floodplains acted as a major sink of coarse sediment during the flood, storing 72% of the eroded sediment, although these floodplains can also be a source of sediment through scouring and erosion processes. 3. Spatial patterns of erosion and deposition were controlled by valley confinement; where the channel is naturally unconfined overbank floodplain deposits were prominent, in contrast, in artificially-confined reaches, bank erosion and scour were dominant geomorphic impacts. 4. The event exceeded critical entrainment thresholds for channel bed particle D 50 and D 90 transporting sediment that had aggraded in the channel. Critical entrainment thresholds were not exceeded during daily flows for all particle sizes along St John's Beck in the 2017 survey. 5. Channel capacity decreased 1.5 yr after the event and channel bed grain size had coarsened due to aggradation in the channel.
This study has quantified the importance of the upper floodplain zone in regulating sediment output during extreme events. The results suggest that rather than envisioning upper floodplain zones as effective transfer reaches they are actually major storage zones that capture flood sediments and disrupt sediment continuity downstream. The intervening valley floodplain geomorphology (confinement, slope) plays a major role in influencing the spatial location of erosion and deposition impacts.