General enquiries on this form should be made to:Defra, Science Directorate, Management Support and Finance Team,Telephone No. 020 7238 1612E-mail: research.competitions@defra.gsi.gov.uk

SID 5 Research Project Final Report

SID 5 (Rev. 3/06) Page 1 of 23

NoteIn line with the Freedom of Information Act 2000, Defra aims to place the results of its completed research projects in the public domain wherever possible. The SID 5 (Research Project Final Report) is designed to capture the information on the results and outputs of Defra-funded research in a format that is easily publishable through the Defra website. A SID 5 must be completed for all projects.

This form is in Word format and the boxes may be expanded or reduced, as appropriate.

ACCESS TO INFORMATIONThe information collected on this form will be stored electronically and may be sent to any part of Defra, or to individual researchers or organisations outside Defra for the purposes of reviewing the project. Defra may also disclose the information to any outside organisation acting as an agent authorised by Defra to process final research reports on its behalf. Defra intends to publish this form on its website, unless there are strong reasons not to, which fully comply with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.Defra may be required to release information, including personal data and commercial information, on request under the Environmental Information Regulations or the Freedom of Information Act 2000. However, Defra will not permit any unwarranted breach of confidentiality or act in contravention of its obligations under the Data Protection Act 1998. Defra or its appointed agents may use the name, address or other details on your form to contact you in connection with occasional customer research aimed at improving the processes through which Defra works with its contractors.

Project identification

1. Defra Project code WQ0120

2. Project title

Sediment sourcing in the Demer basin, Belgium: implications for catchment diffuse pollution management

3. Contractororganisation(s)

ADAS UK Ltd.                         

54. Total Defra project costs £ 17,005(agreed fixed price)

5. Project: start date................ 01 April 2007

end date................. 31 March 2008

SID 5 (Rev. 3/06) Page 2 of 23

6. It is Defra’s intention to publish this form. Please confirm your agreement to do so...................................................................................YES NO (a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They

should be written in a clear and concise manner and represent a full account of the research project which someone not closely associated with the project can follow.Defra recognises that in a small minority of cases there may be information, such as intellectual property or commercially confidential data, used in or generated by the research project, which should not be disclosed. In these cases, such information should be detailed in a separate annex (not to be published) so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report without including references to any sensitive or confidential data, the information should be included and section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No" answer.In all cases, reasons for withholding information must be fully in line with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.

(b) If you have answered NO, please explain why the Final report should not be released into public domain

Executive Summary7. The executive summary must not exceed 2 sides in total of A4 and should be understandable to the

intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together with any other significant events and options for new work.The need to manage diffuse sediment pollution is moving up the policy agenda. Mitigating sediment pressures on watercourses is important because sediment can play a key role in governing the transfer and fate of nutrients (e.g. phosphorus) and contaminants (e.g. pesticides, faecal indicator organisms). Enhanced sediment loadings have the potential to impact detrimentally upon freshwater ecology as demonstrated by the spawning gravel siltation problem. It is therefore important to reduce sediment pressures on freshwater ecosystems.

Improved sediment management requires reliable information on the key sources of the sediment problem. Such information should document the relative significance of sediment sources and should be based upon an understanding of sediment mobilisation and delivery at catchment scale. The catchment has been identified as the principal spatial unit for diffuse pollution management by the EU Water Framework Directive.

Traditional approaches to documenting sediment sources are confounded by a range of problems, including spatial and temporal sampling issues and the costs involved. In addition, conventional approaches, e.g. erosion pins, require additional data on sediment delivery to watercourses in order that information on erosion can be converted into an understanding of the key sources contributing to sediment loadings and pressures measured at a downstream location. Estimation of sediment delivery ratios involves many uncertainties. Diffuse pollution models commonly fail to represent all potential key sediment sources by focusing upon specific aspects of the catchment sediment budget e.g. sediment mobilisation and delivery from agricultural fields. Sediment source apportionment can therefore demand the integration of several models. Coupled modelling can be resource demanding.

Sediment fingerprinting offers a valuable alternative direct means of establishing catchment scale sediment sources. The fingerprinting approach is founded on the link between the geochemical properties of sediment and those of its sources. Comparison of sediment source material and downstream sediment samples using a range of fingerprint properties permits estimation of the relative contribution of individual sources to the sediment flux. Potential sediment sources can be classified as spatial units represented by geological zones or tributary sub-catchments, or alternatively, as source types comprising surface soils supporting different land use and channel banks/subsurface sources.

Against this background, this project was designed to provide further confirmation of the utility of the fingerprinting approach as a sediment policy support tool. Accordingly, a revised quantitative fingerprinting technique, incorporating statistical analysis, numerical mass balance computer modelling

SID 5 (Rev. 3/06) Page 3 of 23

and uncertainty analysis was used to examine:a) Spatial sources of sediment pressures in the Demer, basin, Flanders, Belgiumb) Sediment source types in two representative sub-catchments of the Demer basin, Flanders,

Belgium

Spatial sediment sources were defined in terms of the eight tributary sub-catchments comprising the Demer basin. Suspended sediment originating from each tributary sub-catchment was collected using simple time-integrating samplers operating in situ. The geochemical fingerprints of the suspended sediment samples collected from the tributary sub-catchments were compared with those of the sediment samples collected at the overall drainage basin outlet. The findings suggested that spatial sediment sources in the Demer basin can be readily distinguished using the fingerprinting approach. Further work is required to improve the representativeness of the spatial data by taking more explicit account of water conveyance and storm water travel times through the Demer basin.

Sediment sourcing in the Mangelbeek sub-catchment was designed to investigate key sediment sources in an area representative of the Tertiary sands across northern Flanders. Potential sediment sources were classified as surface soils supporting cultivation and pasture, channel banks/subsurface sources, authigenic Fe production due to groundwater upwelling in the river channel and channel bed sediment remobilisation (a secondary sediment source). The findings (for the period 14/03/2007-22/05/2008) suggested that sediment sources are in the order; channel banks/subsurface sources (27±2%), cultivated topsoils(24±2%), remobilised channel bed sediment (21±2%), authigenic Fe (20±2%) and pasture topsoils (8±2%). On the basis of the sediment source data, sediment mitigation options in the Mangelbeek sub-catchment need to focus upon protecting channel banks and interrupting sediment delivery from cultivated surface soils. The mitigation of sediment loss from pasture fields requires less attention.

Sediment sourcing in the Gete sub-catchment was used to document sediment sources in an area representative of the loess deposits in southern Flanders. Potential sediment sources were categorised as surface soils supporting cultivation and pasture, channel banks/subsurface sources and channel bed sediment remobilisation. Over the duration of the sampling campaign in the Gete sub-catchment (14/03/2007-22/05/2008), the mean relative contributions from the individual sediment source types were estimated in the order; cultivated topsoils (81±2%), channel banks/subsurface sources (11±2%), and pasture topsoils and channel bed sediment remobilisation (both 4±2%). On the basis of these data, sediment management strategies in the Gete sub-catchment need to target the reduction of soil erosion and subsequent sediment delivery to watercourses from cultivated fields and, to a lesser extent, the protection of eroding channel bank sections.

Future work could usefully benefit from:a) Repeat sediment sourcing surveys to examine the efficacy of individual sediment mitigation

options or the integrated impact of mitigation strategies deploying several control methodsb) Extension of the sediment sourcing framework to apportion sediment loss from higher resolution

sediment sources e.g. poached gateways versus feeder areas versus wider pasture sources, in order to support the targeting of capital grant options

The fingerprinting approach clearly provides the basis for a useful sediment policy support tool.

SID 5 (Rev. 3/06) Page 4 of 23

Project Report to Defra8. As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with

details of the outputs of the research project for internal purposes; to meet the terms of the contract; and to allow Defra to publish details of the outputs to meet Environmental Information Regulation or Freedom of Information obligations. This short report to Defra does not preclude contractors from also seeking to publish a full, formal scientific report/paper in an appropriate scientific or other journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms. The report to Defra should include: the scientific objectives as set out in the contract; the extent to which the objectives set out in the contract have been met; details of methods used and the results obtained, including statistical analysis (if appropriate); a discussion of the results and their reliability; the main implications of the findings; possible future work; and any action resulting from the research (e.g. IP, Knowledge Transfer).

ContextDiffuse sediment inputs to rivers are widely perceived to be an important environmental issue. Whilst the presence

of elevated levels of suspended sediment will reduce light penetration in the river water column, thereby impacting upon primary productivity, the fine (silt and clay) fraction of sediment fluxes has been shown to represent an important vector for the transport and delivery of nutrients, organic contaminants, and trace and heavy metals (Allan, 1986; Sibbesen and Sharpley, 1997; Miller 1997; Dawson and Macklin, 1998; Russell et al., 1998; Meharg et al., 1999; Camusso et al., 2002; Warren et al., 2003). Sediment-associated transport dominates (>90%) the global land-ocean flux of many natural and anthropogenically-derived substances mobilised within, and subsequently delivered through, river systems (Frank, 1981; Meybeck and Helmer, 1989; Horowitz et al., 2001). The dispersal and fate of sediment-associated nutrients and contaminants is therefore heavily controlled by the transport, storage, remobilisation and delivery of suspended sediment in river basins. Cave et al. (2005), working on the Humber estuary, England, suggested that sediment deposition within the estuarine zone has the capacity to store 55-97% of As, 15-27% of Cu, 17-50% of Pb and 11-12% of Zn annual input loads. Working on lowland permeable catchments in the UK, Collins et al. (2005) estimated that channel bed storage of fine sediment-associated contaminants was equivalent to a significant proportion of the suspended sediment-associated contaminant flux measured at the study cathment outlets. On the basis of catchment sediment yields of 5-15 t km-2 yr-1, channel bed fine sediment storage accounted for between 20-77% and 7-26% (C), 8-82% and 3-27% (Cd), 11-61% and 4-20% (Co), 4-65% and 1-22% (Cr), 13-58% and 4-19% (Cu), 19-68% and 6-23% (N), 9-76% and 3-25% (total P), 12-41% and 4-14% (Pb), and, 12-48% and 4-16% (Zn) of the respective suspended sediment-associated fluxes in the Frome/Piddle, Pang/Lambourn and Tern catchments.

Sediment quality and more importantly quantity also directly or indirectly impact habitat condition and ecological status. High suspended sediment concentrations can affect the feeding and health of aquatic organisms, contributing to the loss of species diversity and dysfunction of community structure (Carpenter et al., 1998; Mainstone and Parr, 2002). The deposition and accumulation of fine-grained sediment on river channel beds is increasingly recognised as an important environmental problem (Collins and Walling, 2007a, b). Channel bed siltation primarily degrades habitat quality by reducing hyporheic exchange (Packman and Mackay, 2003). Excessive sediment accumulation adversely affects two critical properties of fish spawning gravels, namely, permeability and porosity (Acornley and Sear, 1999; Baxter and Hauer, 2000). Permeability governs the rates of dissolved oxygen supply and metabolic waste removal, both of which critically influence egg-to-hatching success (Iwamoto et al., 1978; Turnpenny and Williams, 1980; Alonso et al., 1996; Theurer et al., 1998; Naden et al., 2003; Berry et al., 2003; Greig et al., 2005). Porosity controls the intra-gravel movement and eventual emergence of newly hatched fry and is reduced by excessive sediment accumulation and concretion of the channel substrate (Weaver and Fraley, 1993; Crisp, 1993; Shackle et al., 1999). River channel sedimentation can also have a deleterious impact on macrophyte communities (Graham, 1990; Clarke and Wharton, 2001) and invertebrate populations (Cummins and Lauff, 1969; Ward et al., 1998).

Given the wider environmental significance of sediment pressures, the need for further sediment research to assist catchment management is rising up the policy agenda. On account of sediment-associated environmental issues, it is widely acknowledged that sediment management and mitigation strategies require reliable information on sediment sources. Management of sediment mobilisation at source provides scope for prevention rather than cure.

Documenting catchment scale sediment sourcesReliable information on catchment suspended sediment sources is an essential prerequisite for assisting the design

and implementation of targeted abatement strategies for controlling sediment and associated diffuse pollution problems (United States Environmental Protection Agency, 1999; Collins et al., 2001). Equally, improved datasets on sediment sources are needed to assist the interpretation of catchment suspended sediment budgets and response over time to changes in management (Dedkov and Moszherin, 1992; Reid and Dunne, 1996; Walling et al., 2001). It is therefore important to document catchment sediment sources.

Existing approaches to assembling information on catchment sediment sources comprise two key categories (Collins and Walling, 2004). The indirect approach to sediment source assessment is founded on the use of a number of techniques to measure sediment mobilisation in situ. But, on account of being developed to assess soil erosion rather than sediment sources per se, these methods take no explicit account of the substantial uncertainties in linking potential catchment sediment sources to the river channel. Areas of significant erosion will not represent sediment sources, unless there is clear

SID 5 (Rev. 3/06) Page 5 of 23

connectivity with watercourses thereby permitting eroding areas to contribute to downstream sediment fluxes. Consequently, sediment sources can only be inferred on the basis on erosion data, unless the linkages between erosion, sediment transport, deposition and sediment flux can be readily quantified. Given the uncertainties associated with quantifying sediment transfers from land to water, information on erosion must be interpreted in conjunction with an understanding of the remaining components of the sediment delivery system for the purpose of providing meaningful data on sediment provenance.

Indirect assessment of sediment sources can be undertaken using a range of techniques. For example, numerous studies have employed surveying based on profilometers (Shakesby, 1993), erosion pins (Couper et al., 2002), cross-profiling (Springer et al., 2001) and GPS (Malet et al., 2002). Alternatively, both terrestrial and aerial photogrammetry has been used to monitor a range of sediment sources including eroding channel banks (Barker et al., 1997) and gullying (Nachtergaele and Poesen, 1999). In other cases, either bounded (Vacca et al., 2000) or unbounded (Megahan et al., 2001) erosion plots have been deployed to obtain data. Geomorphological mapping has also provided an indirect means of elucidating sediment sources (Hasholt and Hansen, 1993). It is important to note, however, that the deployment of these traditional methods for documenting catchment suspended sediment sources is frequently constrained by a number of important problems, including the representativeness of data collection, logistical considerations and the costs involved (Loughran and Campbell, 1995; Collins and Walling, 2004).

Due to such problems, the fingerprinting approach has attracted increasing attention as a reliable alternative direct means of assembling information on catchment sediment sources. Sediment source fingerprinting is founded upon the link between the geochemical properties of suspended sediment and those if its sources. Assuming potential sediment sources can be readily distinguished on the basis of their constituent properties or ‘fingerprints’, the provenance of sediment can be established using a comparison of its properties with those of the individual potential sources.

The sediment fingerprinting approachThe discrimination of individual potential sediment sources using the fingerprinting approach has traditionally involved

a wide range of fingerprint properties (Collins and Walling, 2002, 2004). The choice of properties has, to some extent, typically reflected access to the necessary laboratory analytical equipment as well as previous or published experience. Some studies have used mineral-magnetism to identify sediment sources on account that mineral-magnetic measurements are simple, cheap, rapid and non-destructive (Walden et al., 1997; Caitchen, 1998). Alternatively, some source fingerprinting investigations have used mineralogy or colour as a means of distinguishing potential sediment sources in catchments with heterogeneous geology and pedology (Garrad and Hey, 1989). In other cases, sediment geochemistry (Lewin and Wolfenden, 1978; Jones et al., 1991), environmental radionuclides (He and Owens, 1995; Wallbrink et al., 1998), organic constituents (Oldfield and Clark, 1990; Peart, 1995), stable isotopic properties (Douglas et al., 1995) or particle size measurements (Kurashige and Fusejima, 1997; Hillier, 2001) have been used to discriminate individual sediment sources.

Due to the frequent need to distinguish several potential sediment sources, it is now widely accepted that the quest for a single diagnostic property is inappropriate on account of the problem of spurious source-sediment matches (Collins and Walling, 2002). In consequence, most recent source fingerprinting studies have used so-called ‘composite fingerprints’ comprising a range of different diagnostic properties (Collins and Walling 2002, 2004). Composite fingerprints comprise individual properties influenced by differing environmental controls and which thereby improve source discrimination by affording a substantial degree of independence. Such fingerprints can represent several diagnostic properties from either a particular property subset e.g. several radiometric (He and Owens, 1995), mineral-magnetic (Oldfield and Clark, 1990) or geochemical (Collins and Walling, 2002) properties, or a combination of geochemical, radiometric and organic constituents (Walling et al., 1993; Collins et al., 2001; Collins and Walling, 2002). In order to satisfy dimensionality, the number of fingerprint properties should exceed the number of potential sediment sources being discriminated (Foster and Lees, 2000; Collins and Walling, 2004).

Sediment source fingerprinting assumes that the selected fingerprint properties are readily transported and deposited in association with suspended sediment and that selective erosion and sediment delivery processes do not transform the properties (via enrichment, depletion, dilution) beyond what can be corrected for using appropriate procedures. Composite fingerprints should be identified using statistical verification (Collins et al., 1997a, 2000; Collins and Walling, 2002). Many investigations using the fingerprinting approach have used a simple qualitative comparison between the fingerprint properties of different potential sources and sediment samples as a means of elucidating sediment provenance (Peart, 1993; Walling and Kane, 1984; Walling and Amos 1999). But, in order to provide more useful quantitative information on sediment contributions from individual sources, composite fingerprints are now generally used in conjunction with a multivariate numerical mixing model (Collins et al., 1997a, 2001; Wallbrink et al., 2003; Krause et al., 2003; Motha et al., 2004). Sediment mixing models can be based on linear programming (Yu and Oldfield, 1989, 1993; Caitcheon, 1993, 1998) or optimisation algorithms (Collins et al., 1997a; Walling et al., 1999; Walling, 2005).

Application of the sediment fingerprinting approach to document catchment suspended sediment sources necessitates collection of representative samples of individual potential sediment sources. The latter can be defined in a variety of ways. In some investigations, especially those in large-scale river drainage basins, it has proved most meaningful to investigate the spatial provenance of suspended sediment sources, defined in terms of individual tributary sub-catchments (Collins et al., 1996; Walling et al., 1999) or discrete geological zones (Collins et al., 1998; Walling et al., 1999; Owens et al., 2000; Bottrill et al., 2000). In smaller catchments, it is commonly more appropriate to characterise sediment provenance in terms of individual source types comprising either surface and subsurface categories (Peart and Walling, 1986, 1988) or surface soils supporting different land use and eroding channel banks (Collins et al., 1997b; Walling et al., 1999; Collins et al., 2000; Russell et al., 2001; Krause et al., 2003; Motha et al., 2004; Walling and Collins, 2005). Carter et al. (2003) recently used the approach to investigate suspended sediment source types (topsoils supporting woodland, pasture and cultivation, channel banks, road dust and solids from sewage treatment works) in an urban river system in northern England. Sediment source fingerprinting affords a convenient basis for investigating spatial provenance and source type in an integrated manner (Collins et al., 1997b). As well as documenting contemporary suspended sediment sources, the fingerprinting approach provides a unique means of reconstructing longer-term sediment provenance and thus for examining linkages between soil

SID 5 (Rev. 3/06) Page 6 of 23

erosion patterns and land use change (Collins et al., 1997c; Owens et al., 1999) or the occurrence of extreme flood events (Collins et al., 1997d). The approach has recently been used to examine the contribution of channel bed sediment remobilisation to suspended sediment flux at the outlets of lowland groundwater-fed catchments in the UK (Collins and Walling, 2006). Recent work in conjunction with the England Catchment Sensitive Farming Delivery Initative (ECSFDI) has used the approach to examine cross sector sediment and particulate phosphorus sources on the Somerset Levels (Collins, 2008a), sediment inputs to watercourses from damaged road verges in the Hampshire Avon (Collins, 2008b) and to assess both primary and secondary sediment source inputs in the upper Kennet catchment (Collins, 2008c).

In tandem with the adoption of statistical and numerical data processing techniques for fingerprinting catchment sediment sources, other important developments are associated with the use of various corrections and weightings during sediment source ascription. Since the properties of soil and sediment samples are strongly controlled by particle size composition and organic matter content (Horowitz, 1991), it is necessary to correct for differences in these characteristics. The selectivity of sediment delivery processes means that sediment samples are typically enriched in fines and organic matter content relative to the individual contributing source areas of the catchment. Approaches to correct for contrasts in particle size and organic matter content have varied in complexity. The most basic approach has been to restrict laboratory analyses to only the <63 µm (<0.063 mm) fraction of source and sediment samples, thereby ensuring a focus upon the dominant size class of suspended sediment (Motha et al., 2002). Given that the composition of the <63 µm fraction tends to differ between samples, fingerprint property concentrations measured on this size fraction have been corrected using specific surface area information (Collins et al., 1997a, 1998; Gruszowski et al., 2003). Specific surface area provides a useful surrogate for particle size and is governed by the entire composition of a given size fraction. More complex approaches to correcting for particle size composition have been based on detailed information on the precise relationship between grain size composition and the concentrations of individual fingerprint properties (He and Walling, 1996; Russell et al., 2001). These approaches avoid the assumption that there is a consistent linear relationship between concentration and particle size composition for all properties. But, the identification of correction factors for individual properties requires substantial investment in laboratory resources. Alternatively, some researchers have adjusted the fingerprint property concentration data for source materials using information on the grain size characteristics of sediment and the concentration information for different size fractions of the source samples. Under these circumstances, source material fingerprint property concentrations are adjusted to reflect the same particle size composition as that measured for sediment (Slattery et al., 1995; Motha et al., 2002). Less attention has been directed towards correcting for contrasts in the organic matter content of samples. Organic matter content adjustments typically rely upon a simple ratio between the organic carbon content of source material and sediment samples (Collins et al., 1997a, 1998), or the adjustment of source material fingerprint property concentrations to reflect a similar organic matter content to that measured for sediment (Motha et al., 2002). Correcting for organic matter content frequently reduces the errors associated with numerical sediment source ascription (e.g. Walling et al., 2003), although the risk of double correction, in tandem with the use of a particle size correction factor should be carefully explored during each fingerprinting study.

In addition to corrections for particle size and organic matter content, the varying levels of precision of laboratory analyses for individual sediment properties has also been taken into account in order to ensure that greater emphasis is placed on those properties affording the greatest precision (Collins et al., 1997a, 1998). Due to the need to take explicit account of the natural variability of source material properties, uncertainty testing has been incorporated into the quantitative source apportionment procedure, using a selection of Bayesian statistics and Monte Carlo routines (Rowan et al., 2000; Small et al., 2002; Motha et al., 2004; Douglas et al., 2003; Collins and Walling, 2007 b, c).

The Demer study and its scientific objectivesAs part of ongoing sediment policy support work in Flanders, Belgium, the Demer drainage basin has been selected

to demonstrate the utility of the fingerprinting approach for providing Catchment Officers with reliable information on the sources of the sediment problem. The Demer basin has been selected for the sediment policy support work as its southern part is characterized by loess deposits and is therefore highly susceptible to soil erosion, whereas its northern part is characterized by iron-rich Tertiary sands, which contribute iron-rich authigenic sediment to particulate fluxes. The Demer drainage basin is thus broadly representative of Flemish river basins experiencing problems associated with diffuse sediment pollution.

River basin managers can make use of available models to predict soil erosion and resulting sediment delivery pressures (e.g. Morgan et al., 1998; Verstraeten et al., 2002; Van Rompaey et al., 2003; Collins et al., 2007). However, validation of the predictions by such models at catchment scale is frequently difficult, especially in situations where the model framework fails to represent the entire sediment budget and where the models have been developed to support wider strategic as opposed to catchment scale characterisation. Since management decisions need to be based on reliable evidence, other techniques that provide information on sediment sources, such as sediment fingerprinting, potentially provide a useful alternative approach. The Demer sediment policy support project therefore provided a useful opportunity for further testing the utility of the fingerprinting approach to assist management decisions at drainage basin and catchment scale. The need to manage diffuse pollution problems at the catchment scale has been emphasised by the EU Water Framework Directive (European Parliament, 2000).

As a means of testing the fingerprinting approach as part of the Demer study, work focused upon:a) Using sediment fingerprinting to apportion spatial sediment sources, characterised in terms of individual

tributary sub-catchments, as a means of providing the first tier of information required to help target sediment mitigation options (spatial data)

b) Using sediment fingerprinting to apportion individual sediment source types in two sub-catchments representative of the loess deposits and Tertiary sands observed across Flanders, as a means of providing the second tier of information required to help target sediment mitigation options (land use inputs)

c) Using sediment fingeprinting to apportion primary and secondary sediment sources in the two representative sub-catchments, as a means of further elucidating sediment dynamics and the capacity of the systems to flush fine channel bed sediment deposits.

SID 5 (Rev. 3/06) Page 7 of 23

Focusing upon the Demer drainage basin provided scope for extrapolating the results to other river basins in wider Flanders experiencing diffuse sediment pollution.

The Demer sediment policy support project involves the following partners:Flemish Environment Agency (VMM)Flanders Hydraulics ResearchUniversity of GhentADAS UK LtdUniversity of Antwerp

Establishing the spatial provenance of sediment pressures in the Demer drainage basinThe River Demer, a tributary of the River Dijle, is part of the Scheldt drainage basin. The Demer basin covers 2334

km², of which 2163 km² is upstream of the gauging station in Aarschot, which was selected as the overall outlet for this investigation. Geologically, the Demer drainage basin can be divided into two significantly different areas: the southern part, characterised by loess deposits, and the northern part characterised by iron-rich Tertiary sands producing authigenic iron (Vanlierde et al. 2007). For the purpose of the sediment sourcing exercise, spatial sediment sources were represented by eight tributary sub-catchments (Figure 1), of which three (the Mangelbeek, Zwartebeek and the Hulpe) drain the northern part of the basin and three (the Gete, the Velpe and the Herk) the southern portion. The final two sub-catchments (the upper Demer and the Motte) drain areas underlain by both of the principal geological formations.

Field sampling and laboratory workIsokinetic time integrating samplers were installed on each of the eight tributaries selected to represent potential

spatial sediment sources in the Demer drainage basin. These simple, affordable samplers (Phillips et al., 2000), provide a means of avoiding the logistical difficulties associated with the need to sample multiple sites and operate according to the principles of sedimentation. Each sampler was attached to an I-profile, to prevent displacement during flood events. Previous work has demonstrated that these traps provide representative samples of suspended sediment elemental composition (Russell et al., 2000). The samplers operate in situ without the need for power and collect composite samples of sediment continuously during the period of operation. Natural variations in sediment properties during either individual storm events or a series of floods are therefore captured, whilst the devices provide a means of collecting sufficient sample mass for laboratory analyses. Deployment of the time-integrating devices negates the need for quick response site visits during storm events and therefore ensures that sediment flux is continuously sampled. The samplers (Figure 2) comprise a PVC pipe (98 mm internal diameter, 1 m length) with two end caps containing a central inlet / outlet pipe (4 mm internal diameter). During insertion, the sampler is filled with clean native water and submerged in alignment with water flow. The sampler is typically secured to uprights, which are either driven into the river bed or concreted into heavy blocks. Following submersion, water flow enters the sampler via the inlet tube and upon progressing into the main chamber, represented by the PVC pipe, its velocity in reduced in excess of 600, thereby encouraging sedimentation. Because most fine-grained sediment is transported in the form of aggregates, the sedimentation within the sampler collects a heterogeneous mix of the primary particle sizes comprising the local absolute grain size distribution for suspended sediment. Settling velocities for aggregates exceed those for individual primary particles, thereby ensuring that sediment deposition within the sampler occurs more readily than predicted by Stokes Law. The time-integrating samplers were emptied on a regular (approximately monthly) basis when low flow conditions permitted safe entry to the river channel. Sediment retrieved by the samplers was de-watered using a combination of settling and centrifugation. Representative samples of suspended sediment flux at the Aarschot outlet monitoring station (Figure 1) could not be obtained using time-integrating samplers because of the high ambient water depth and flow velocities and resulting safety considerations. As a result, suspended sediment was sampled using an Alfal Laval Emmie continuous-flow centrifuge. The outlet sampling stations for each discrete tributary sub-catchment are listed in Table 1. Details of the time-integrated samples used to characterise sediment originating from the individual spatial sediment sources in the Demer drinage basin are presented in Table 2. Table 3 summarises the suspended sediment sampling undertaken at the overall drainage basin outlet at Aarschot.

SID 5 (Rev. 3/06) Page 8 of 23

Figure 1: The Demer drainage basin showing the tributary sub-catchment and overall outlet sampling stations

Figure 2: A schematic of the time-integrating suspended sediment sampler

Table 1: Summary of the tributary sub-catchment sampling stations in the Demer drainage basinTributary sub-catchment Sampling station

Motte RillaarHulpe MolenstedeVelpe HalenGete Halen

Zwartebeek LummenMangelbeek Lummen

Herk Kermtupper Demer Hasselt

SID 5 (Rev. 3/06) Page 9 of 23

13

5

48

6 7

9

2

Key1 Basin outlet: Demer at Aarschot2 Motte at Rillaar3 Hulpe at Molenstede4 Zwartebeek at Lummen5 Velpe at Halen6 Gete at Halen7 Herk at Kermt8 Mangelbeek at Lummen9 upper Demer at Hasselt

Table 2: Summary of the sampling periods used to characterise sediment originating from the individual tributary sub-catchment spatial sediment sources in the Demer drainage basin

Period No. Dates1 14/03/2007-20/04/20072 20/04/2007-24/05/20073 24/05/2007-21/06/20074 21/06/2007-6/08/20075 6/08/2007-29/08/20076 29/08/2007-27/09/20077 27/09/2007-6/11/20078 6/11/2007-18/12/20079 18/12/200728/01/2008

10 28/01/2008-28/02/200811 28/02/2008-10/04/200812 10/04/2008-22/05/2008

Table 3: Summary of the suspended sediment sampling at the overall Demer drainage basin outlet at AarschotSample No. Date

1 13/04/20072 14/09/20073 24/10/20074 4/12/20075 21/03/08 (10.20-11.20 am)6 21/03/08 (11.40-12.40 pm)7 21/03/08 (12.55-13.45 pm)8 21/03/08 (14.00-14.30 pm)

Following freeze drying, the time-integrated suspended sediment samples were sieved using a 63 µm nylon mesh and only the <63 µm fraction was used for analyses of fingerprint properties, in recognition that clay and silt particles represent the most chemically active fractions (Horowitz, 1991). Energy dispersive X-ray fluorescence (EDXRF) analysis required pressed pellets (minimum thickness of 6 mm; typically 6 g of sediment mixed with 1.2 g of wax). When less than 6 g of sediment was available, a smaller quantity of sediment (minimum 5 g) was used, but the 5:1 sample to wax ratio was maintained. The sediment/wax mixture was homogenized in a mixing mill prior to being compressed into pellets in a semi-automatic press using pressurised air (ca. 2 minute duration) and 25 mm diameter stainless steel dies. Organic matter content was determined using loss on ignition (LOI) at 5500C for three hours, whereas absolute particle size analysis was measured using a Malvern Mastersizer 2000 laser diffraction granulometer, following exposure to ultrasound. Output from the grain size analysis was used to estimate specific surface area, assuming spherical particles. Sediment sample density was measured using a pycnometer (Accupyc II 1340 - Micromeritics).

An Epsilon 5 (PANalytical, Almelo) high-energy EDXRF machine using a polarizing beam was deployed to determine the elemental composition of the time-integrated samples collected to represent the individual spatial sediment sources as well as sediment collected from the overall basin outlet at Aarschot. This instrument has a 600 W Gd-anode with exciting voltages of 25 - 100 kV and 0.5 - 24 mA current. The High Purity Ge-detector (HPGe) has an energy range of 0.7 - 200 keV and a resolution at Mn Kα of < 165 eV. Excitation conditions were chosen based on a previous application with soil and sediment samples, using standard reference material for calibration purposes. A Compton correction was used to provide a simple and reliable method of matrix correction for the heavy elements. Output from the EDXRF analyses were validated using ICP-AES and a correlation for both low- and high-Z elements.

Spatial sediment source discriminationSediment fingerprints require statistical verification to confirm their suitability for distinguishing the sediment sources

under scrutiny. The two-stage statistical selection procedure proposed by Collins et al. (1997a) was used to identify a statistically robust composite fingerprint for distinguishing the time-integrated sediment samples collected from the eight tributary sub-catchments in the Demer study area.

In stage one, the non-parametric Kruskal-Wallis H-test was deployed to confirm the ability of individual tracer properties for distinguishing the spatial source suspended sediment samples. Deployment of the Kruskal-Wallis H-test is founded on the logical assumption that the selection of a robust composite fingerprint requires confirmation of the power of individual constituents to discriminate the source samples under scrutiny. The Kruskal-Wallis H-test is the non-parametric equivalent of analysis of variance and provides a distribution-free procedure for examining contrasts between sample sets. It has a power efficiency of ca. 95.5%, thereby rendering it suitable for testing in conjunction with relatively small sample sets (Hammond and McCullagh, 1978). Greater inter-group contrasts generate larger test statistics and where these exceed the critical value, Ho (i.e. the null hypothesis stating that measurements of the fingerprint property exhibit no significant differences between the sediment source categories) is rejected. The Kruskal-Wallis H-test is applied to the values of a specific property for the sediment source dataset as a whole. Consequently, a statistically significant output is suggestive of source inter-category contrasts, rather than confirming differences between all possible pairs of source categories (Fowler and Cohen, 1990). Stage one of the procedure provides a basis for eliminating redundant fingerprint properties. The results in Table 4 confirmed that a total of 22 properties passed the test, by generating test statistics in excess of the critical value (14.07) at 95% confidence. The highest H-values were computed for Zn (79.388), Fe (77.355) and Ba (77.269). The lowest H-value

SID 5 (Rev. 3/06) Page 10 of 23

(32.379) was computed for V. All properties generated p-values of 0.000. Since no properties failed the Kruskal-Wallis test, all constituents progressed to stage two.

During the second stage of the selection procedure, Discriminant Function Analysis (DFA) was performed to identify from the properties passing the Kruskal-Wallis H-test, the optimum (i.e. smallest) combination of constituents, or composite fingerprint, for correctly classifying the suspended sediment samples representing the individual tributary sub-catchment spatial sediment sources. DFA estimates discriminant function coefficients indicative of the explanatory power of fingerprint properties. A multivariate stepwise selection algorithm, based on the minimisation of Wilks’ lambda, was used to identify the optimum (i.e. smallest) combination of properties, or composite fingerprint, for discriminating the sediment samples collected to charactertise the tributary sub-catchments. During the stepwise selection procedure, properties satisfying two principal test criteria, i.e. the partial F ratio and tolerance level, are entered in order of their explanatory power. Default values were used for both the partial F ratio (1.0) and tolerance level (0.001). As a means of avoiding the preferential selection of individual properties for inclusion in the final composite fingerprint, all parameters were assigned the default inclusion level (1.0). Stepwise selection ceases when all source material samples are classified correctly, or when none of the remaining constituents available for inclusion in the composite signature improve sample discrimination. A composite fingerprint was identified in recognition that the use of a single property can be confounded by spurious source-sediment matches and that the discriminatory power of geochemical signatures improves with the addition of different properties potentially responding to a range of environmental controls (Collins and Walling, 2002). The final results of the DFA for discriminating spatial sediment sources in the Demer study area are presented in Table 5. In order to satisfy the dimensionality required by the sediment mixing model (cf. Foster and Lees, 2000) used to apportion spatial sediment source contributions, the final composite signature comprised a total of 13 individual properties. A total of 100% of the tributary suspended sediment samples were classified correctly on the basis of the optimum composite fingerprint. This performance was highly encouraging since previous sediment source tracing studies examining spatial sources have attempted to distinguish fewer (ca. 5) sub-catchments (e.g. Collins et al., 1998; Walling et al., 1999).

Table 4: The results of the Kruskal-Wallis H-test in relation to discriminating spatial sediment sources in the Demer drainage basin

Fingerprint property H-value P-valueAl 73.860 0.000Ba 77.269 0.000Ca 66.723 0.000Ce 75.891 0.000Cr 41.475 0.000Cu 57.720 0.000Density 70.916 0.000Fe 77.355 0.000K 55.608 0.000LOI 76.287 0.000Mn 54.369 0.000Ni 64.427 0.000P 75.989 0.000Pb 73.732 0.000Rb 69.460 0.000S 68.336 0.000Si 75.175 0.000Sr 63.949 0.000Ti 72.211 0.000V 32.379 0.000Zn 79.388 0.000Zr 73.779 0.000critical value @ 95.5% confidence = 14.07 (for n-1 = 7 degrees of freedom)

Table 5: The optimum composite fingerprint for discriminating spatial sediment sources in the Demer drainage basin

Step Fingerprint property selected % tributary sub-catchment sediment samples classified

correctly1 Ce 64.02 Zn 74.23 P 91.04 Ba 95.55 Pb 97.86 Rb 98.97 Al 98.98 K 1009 Ti 98.9

10 Cr 10011 Zr 10012 Si 10013 Cu 100

SID 5 (Rev. 3/06) Page 11 of 23

Spatial sediment source apportionmentA slightly modified version of the multivariate mixing model developed by Collins et al. (1998) was used to apportion

spatial sediment sources in the Demer study basin. Optimised estimates of the relative contributions of the individual tributary sub-catchment spatial sources to each sample of suspended sediment collected from the overall basin outlet at Aarschot were provided by minimizing the sum of squares of the weighted relative errors, viz.:

(1)

where: = concentration of fingerprint property for each basin outlet suspended sediment sample from Aarschot; = the

optimised percentage contribution from tributary sub-catchment spatial sediment source ; = mean concentration of

fingerprint property for spatial sediment source ; = particle size correction factor for tributary sub-catchment spatial

source ; = organic matter content correction factor for tributary sub-catchment spatial source ; = tracer

discriminatory weighting; = number of fingerprint properties comprising the optimum composite fingerprint; = number of spatial sediment sources. Two linear constraints were used to provide boundary conditions for the sediment mixing model, in order to ensure that the relative contribution from each potential tributary sub-catchment spatial source was non-negative (Equation 2) and that the contributions from the individual spatial sources summed to unity (Equation 3).

(2)

(3)

The multivariate sediment mixing model incorporated correction factors for contrasts in the grain size composition and organic matter content of the suspended sediment samples collected from the tributary sub-catchments and the overall study basin outlet. The former correction was estimated using the ratio of the specific surface area of each individual basin outlet suspended sediment sample collected from Aarschot to the mean specific surface area of the corresponding samples collected to characterise each tributary sub-catchment spatial source (Collins et al., 1997a). The correction factor for organic matter content was calculated in the same manner. The combined use of the particle size and organic matter content correction factors was examined carefully in order to check for over-correction, but in all cases, both factors improved model performance. A weighting was included to take account of the differing levels of spatial source discrimination associated with each fingerprint property identified in the optimum composite signature. In the original model (Collins et al., 1998), the tracer specific weighting reflected the differing levels of precision associated with laboratory analyses for individual fingerprint properties. The goodness-of-fit was assessed by comparing the actual fingerprint property concentrations measured in the basin outlet suspended sediment samples with the corresponding values predicted by the numerical spatial sediment source mixing procedure. The mean relative errors (i.e. average for all properties included in the optimum composite fingerprint) ranged up to 13%, confirming that the mixing model provided meaningful estimates of the measured elemental composition of the basin outlet suspended sediment samples collected at Aarschot (Collins et al., 1998; Walling and Collins, 2000).

It is important to recognise uncertainty during the application of the sediment fingerprinting approach and a number of methods have been used to address this issue (Rowan et al., 2000; Krause et al., 2003; Motha et al., 2004; Collins and Walling 2007a,b). During this study, a Monte Carlo routine was adopted to assess the uncertainty associated with the characterisation of the tributary sub-catchment spatial sediment sources using suspended sediment samples. The mean and standard deviation calculated for each property in the composite fingerprint, using the set of time-integrated suspended sediment samples collected to represent each individual spatial source, provided a basis for computing a cumulative Normal distribution. The mixing model solutions pertaining to each basin outlet sediment sample from Aarschot were subsequently repeated for 1000 realisations by randomly selecting spatial source elemental composition from the Normal distributions. Finally, the standard error of the mean for the repeat iterations was used to generate 95% confidence limits for the relative spatial sediment source proportions.

ResultsFigure 3 presents the output from the mixing model for the centrifuge samples collected during four discrete flood

events at Aarschot (Table 3). On each sampling occasion, the Motte sub-catchment represented the principal spatial sediment source. For example, at the time of sampling on 13/04/2007 (flood event 1), the Motte sub-catchment was estimated to contribute a mean of 68±2% of the suspended sediment load sampled at Aarschot, with the Monte Carlo routine suggesting a range in the contributions from this spatial source of 12-100%. At the time of sampling on 14/09/2007 (flood event 2), the mean relative contribution from the Motte sub-catchment to the suspended sediment flux sampled at Aarschot was computed at 73±2%, with a corresponding range of 0-100%. The mixing model suggested that the second most important spatial sediment source was the Hulpe sub-catchment (Figure 3). For example, this spatial source was estimated to contribute a respective mean of 5±2% and 7±2% of the suspended sediment load sampled at Aarschot on 13/04/2007 (flood event 1) and 4/12/2007 (flood event 4) (Figure 3). The corresponding ranges were computed at 0-13% on both sampling

SID 5 (Rev. 3/06) Page 12 of 23

occasions. Over the period 13/04/2007 – 4/12/2007, the overall mean relative contributions from the discrete spatial sediment sources were estimated at 70±2% (Motte), 6±2% (Hulpe), 6±2% (Velpe), 5±2% (Gete), 4±2% Zwartebeek), 4±2% (Mangelbeek), 4±2% (Herk) and 1±1% (upper Demer).

Closer inspection of the centrifuge sampling during flood events 1-4, relative to the continuous discharge record at Aarschot, suggested that sampling had coincided with the rising limbs of the four storm events. Under such circumstances, the Motte sub-catchment could be expected to represent the major spatial source of the outlet sediment samples given its proximity to the Aarschot sampling station. It was therefore deemed useful to undertake intra-storm suspended sediment sampling at Aarschot to take better account of storm water travel times and the time-variant inputs from the individual spatial sediment sources. Accordingly, Figure 4 summarises the mixing model output for four discrete sediment samples collected during a single storm event on 21/03/2008. The results suggested that as time proceded, the mean relative contribution from the Motte tributary sub-catchment decreased from 70±2% at 11.20 am to 62±2% at 14.30 pm. At the same time, the corresponding mean inputs from the Hulpe and Velpe sub-catchments increased from 8±2% to 13±2% and 5±2% to 8±2% (Figure 4). This pattern is consistent with storm water arriving at Aarschot from further up the system as time progressed. As a result, if field resources on the 21/03/2008 had permitted, the relative contributions from the tributary sub-catchments more distal from Aarschot (Figure 1) could have been expected to have increased relative to the input from the Motte sub-catchment. The work undertaken during this project has therefore successfully identified a composite fingerprint for discriminating and apportioning the eight tributary sub-catchments comprising the Demer drainage basin. Futher sampling is, however, required to characterise the typical mean relative contributions from the individual spatial sediment sources, taking full account of storm water conveyance through the channel system.

1 2 3 4

0

10

20

30

40

50

60

70

80

% c

ontri

butio

n

Flood event

Motte sub-catchment

Hulpe sub-catchment

Velpe sub-catchment

Gete sub-catchment

Zwartebeek sub-catchment

Mangelbeek sub-catchment

Herk sub-catchment

upper Demer sub-catchment

Figure 3: The mean relative contribution from each tributary sub-catchment spatial sediment source to the suspended sediment load sampled at Aarschot on 13/04/2007 (1), 14/09/2007 (2), 24/10/2007 (3) and 4/12/2007 (4)

SID 5 (Rev. 3/06) Page 13 of 23

21/3/

08 10

.20-1

1.20

21/3/

08 11

.40-1

2.40

21/03

/0812

.55-1

3.45

21/3/

08 14

.00-1

4.30

0

10

20

30

40

50

60

70

% c

ontr

ibut

ion

Sampling time

Motte sub-catchment

Hulpe sub-catchment

Velpe sub-catchment

Gete sub-catchment

Zwartebeek sub-catchment

Mangelbeek sub-catchment

Herk sub-catchment

upper Demer sub-catchment

Figure 4: Intra-storm information on the spatial provenance of the suspended sediment flux sampled at Aarschot during the storm event on 21/03/2008

Establishing suspended sediment source types in the Mangelbeek and Gete sub-catchmentsIn addition to examining spatial sediment sources within the Demer drainage basin, this project also investigated

sediment source types in the Mangelbeek and Gete sub-catchments (Figure 1). Potential primary sediment sources in the Mangelbeek sub-catchment were categorised as cultivated topsoils, pasture topsoils, channel banks/subsurface sources and authigenic Fe production in the river channel. Authigenic Fe was not included as a potential primary sediment source in the fingerprinting exercise for the Gete sub-catchment. Bed sediment remobilisation was included as a potential secondary sediment source in each sub-catchment. The Mangelbeek sub-catchment was selected to be generally representative of the northern portions of the Demer study basin underlain by iron-rich Tertiary deposits. In contrast, the Gete sub-catchment was used to represent the loess areas in the southern portions of the Demer drainage basin.

Field sampling and laboratory workTable 6 summarises the source material sampling campaign in the Mangelbeek sub-catchment. The

channel bed sediment sampling strategy for the Mangelbeek is summarised in Table 7. Tables 8 and 9 present corresponding information for the Gete sub-catchment. Bed sediment sampling sites were selected to represent important storage zones within the channel system (Goyvaerts et al., 1996; De Pauw and Heylen, 2001). Channel bed sediment samples were retrieved using a 2 L Van Veen grab suspended from a small boat. Bed sediment samples were collected from representative 50 m reaches using a zig-zag pattern. Between 25-40 sub-samples were collected at each reach and bulked prior to laboratory analyses. Representative samples of suspended sediment flux at the Lummen monitoring station on the Mangelbeek and Halen on the Gete tributary (Figure 1) were collected using time-integrating samplers (Tables 10 and 11). Samples collected from cultivated and pasture fields comprised topsoil (0-2 cm) susceptible to mobilisation by rainfall and water erosion and subsequent delivery to neighbouring weatercourses. Channel bank samples comprised material from the entire vertical extent of the bank profile at any sampling point. Subsurface samples comprised material collected from rill or gully walls and cuttings. Representative samples of authigenic Fe were collected from groundwater upwelling sites. Each topsoil, bank or authigenic Fe source material sample comprised a composite of up to 10 smaller samples collected from the vicinity of each sampling site, thereby improving the representativeness of the fingerprint property dataset. All samples were freeze-dried before laboratory analysis for fingerprint properties which was restricted to the <63 µm fraction.

Table 6: Summary of sediment source sampling in the Mangelbeek sub-catchmentPotential sediment source No. of samples collected

Cultivated topsoils 10Pasture topsoils 10Channel banks/subsurface sources 10Authigenic Fe 6Channel bed sediment 24

SID 5 (Rev. 3/06) Page 14 of 23

Table 7: Summary of the channel bed sediment sampling campaign in the Mangelbeek sub-catchmentFlemish Environment Agency site i.d. Sampling dates

E003632 24/05/200720/09/20075/03/2008

453000 22/05/200712/09/200727/02/2008

453200 21/05/200711/09/200715/02/2008

453500 21/05/200711/09/200715/02/2008

454050 21/05/200711/09/200715/02/2008

454200 21/05/200711/09/200715/02/2008

454600 21/05/200711/09/200715/02/2008

E003632 21/05/200711/09/200715/02/2008

Table 8: Summary of sediment source sampling in the Gete sub-catchmentPotential sediment source No. of samples collected

Cultivated topsoils 18Pasture topsoils 18Channel banks/subsurface sources 18Channel bed sediment 39

Table 9: Summary of the channel bed sediment sampling campaign in the Gete sub-catchmentFlemish Environment Agency site i.d. Sampling dates

427000 22/05/200712/09/200727/02/2008

429000 4/06/200712/09/200728/02/2008

430000 1/06/200719/09/20075/03/2008

431000 24/05/200719/09/20075/03/2008

433400 4/06/200712/09/200728/02/2008

433600 1/06/200720/09/200727/02/2008

433900 22/05/200712/09/200727/02/2008

434000 22/05/200712/09/200727/02/2008

435000 4/06/200720/09/2007

SID 5 (Rev. 3/06) Page 15 of 23

28/02/2008435800 24/05/2007

20/09/20075/03/2008

439000 1/06/200719/09/200728/02/2008

443000 24/05/200719/09/20075/03/2008

E002723 1/06/200719/09/200728/02/2008

Table 10: Summary of the time-integrated suspended sediment sampling periods on the Mangelbeek sub-catchment

Suspended sediment sample Sampling period1 14/03/2007-20/04/20072 20/04/2007-24/05/20073 24/05/2007-21/06/20074 21/06/2007-6/08/20075 6/08/2007-29/08/20076 29/08/2007-27/09/20077 27/09/2007-6/11/20078 6/11/2007-18/12/20079 18/12/2007-28/01/2008

10 28/01/2008-28/02/200811 28/02/2008-10/04/200812 10/04/2008-22/05/2008

Table 11: Summary of the time-integrated suspended sediment sampling periods on the Gete sub-catchmentSuspended sediment sample Sampling period

1 14/03/2007-20/04/20072 20/04/2007-24/05/20073 24/05/2007-21/06/20074 21/06/2007-6/08/20075 6/08/2007-29/08/20076 29/08/2007-27/09/20077 27/09/2007-6/11/20078 6/11/2007-18/12/20079 18/12/2007-28/01/2008

10 28/01/2008-28/02/200811 28/02/2008-10/04/200812 10/04/2008-22/05/2008

Source type discriminationThe two-stage statistical selection procedure proposed by Collins et al. (1997a) was used to identify a robust

composite fingerprint for distinguishing individual primary and secondary sediment sources in the Mangelbeek and Gete sub-catchments. Table 12 presents the results of the Kruskal-Wallis H-test for the Mangelbeek sub-catchment. A total of 15 properties passed the test, thereby yielding test statistics in excess of the critical value of 9.49 at 95% confidence. Ba, Ce, Ni, Pb, Sr, Ti and Zr failed the test and were therefore not included in stage two. The highest H-values (39.405 and 24.770) were generated by V and Al, respectively (Table 9). Table 13 presents the corresponding final results of the DFA. A combination of density, V, LOI, Cu, K and Rb successfully distinguished 98.3% of the source material samples collected to characterise primary and secondary sediment sources in the Mangelbeek sub-catchment.

Table 12: The results of the Kruskal-Wallis H-test in relation to discriminating sediment source types in the Mangelbeek sub-catchment

Fingerprint property H-value P-valueAl 24.770 0.000Ba 5.802 0.214*Ca 10.376 0.035Ce 3.642 0.457*Cr 12.736 0.013

SID 5 (Rev. 3/06) Page 16 of 23

Cu 17.415 0.002Density 29.392 0.000Fe 23.015 0.000K 18.624 0.001LOI 15.864 0.003Mn 14.243 0.007Ni 9.010 0.061*P 10.693 0.030Pb 5.994 0.200*Rb 14.197 0.004S 24.641 0.000Si 18.676 0.001Sr 6.382 0.172*Ti 2.616 0.624*V 39.405 0.000Zn 23.092 0.000Zr 0.734 0.947*critical value @ 95.5% confidence = 9.49 (for n-1 = 4 degrees of freedom)* = not statistically significant at p=<0.05

Table 13: The optimum composite fingerprint for discriminating potential primary and secondary sediment source types in the Mangelbeek sub-catchment

Step Fingerprint property selected % source type samples classified correctly

1 Density 65.02 V 81.73 LOI 83.34 Cu 90.05 K 95.56 Rb 98.3

The results of the statistical verification procedure for the Gete sub-catchment are shown in Tables 14 and 15. A total of 16 properties passed the Kruskal-Wallis H-test, thereby generating test statistics in excess of the critical value of 7.82 at 95% confidence (Table 14). Al, Ba, Cs, Fe, Mn, Ni and Pb failed to distinguish the source material samples collected to characterise the Gete sub-catchment. The optimum composite fingerprint successfully distingushed 76.7% of the source material samples and comprised a combination of Zr, Sr, Ca, V, K, Ce, Cu, Rb and Ti (Table 15).

Table 14: The results of the Kruskal-Wallis H-test in relation to discriminating sediment source types in the Gete sub-catchment

Fingerprint property H-value P-valueAl 1.457 0.692*Ba 2.302 0.512*Ca 23.948 0.000Ce 63.282 0.000Cr 46.276 0.000Cs 4.458 0.216*Cu 9.139 0.027Density 17.045 0.001Fe 1.556 0.669*K 27.125 0.000LOI 12.211 0.007Mn 6.047 0.109*Ni 2.148 0.542*P 19.900 0.000Pb 6.040 0.110*Rb 16.593 0.001S 63.290 0.000Si 10.469 0.015Sr 34.572 0.000Ti 33.795 0.000V 34.181 0.000Zn 15.751 0.001Zr 56.440 0.000critical value @ 95.5% confidence = 7.82 (for n-1 = 3 degrees of freedom)* = not statistically significant at p=<0.05

SID 5 (Rev. 3/06) Page 17 of 23

Table 15: The optimum composite fingerprint for discriminating potential primary and secondary sediment source types in the Gete sub-catchment

Step Fingerprint property selected % source type samples classified correctly

1 Zr 46.62 Sr 57.33 Ca 63.14 V 66.05 K 70.96 Ce 70.97 Cu 71.88 Rb 74.89 Ti 76.7

Source type source apportionmentA revised version of the multivariate mixing model described by Collins et al. (1997a) was used to apportion sediment

source types in the Mangelbeek and Gete sub-catchments. The original mixing model optimises estimates of the relative contributions from the potential sediment source types by minimising the sum of squares of the weighted relative errors, viz.:

(4)

where: = concentration of fingerprint property in catchment outlet time-integrated suspended sediment sample; = the

optimised percentage contribution from source type category ; = mean concentration of fingerprint property in source

type category ; = particle size correction factor for source type category ; = organic matter content correction factor

for source type category ; = tracer specific weighting; = number of fingerprint properties comprising the optimum

composite fingerprint; = number of sediment source categories. The revised mixing model algorithm also optimises estimates of the relative contributions from the potential sediment

sources by minimising the sum of squares of the weighted relative errors, but includes revised weightings, viz.:

(5)

where: = concentration of fingerprint property in catchment outlet time-integrated suspended sediment sample; = the

optimised percentage contribution from source type category ; = mean concentration of fingerprint property in source

type category ; = particle size correction factor for source type category ; = organic matter content correction factor

for source type category ; = weighting representing the spatial variation of fingerprint property in source type

category ; = tracer discriminatory weighting; = number of fingerprint properties comprising the optimum composite

fingerprint; = number of sediment source type categories.Two linear boundary conditions are imposed on the mixing model iterations to ensure that the relative contributions (

) from the individual sediment source types are non-negative (Equation 6) and that these contributions sum to unity (Equation 7):

(6)

(7)

The particle size correction factor is included in the sediment source type mixing model since it is widely understood that grain size exerts an important influence on element concentrations in soil and sediment samples (Horowitz and Elrick, 1987; Horowitz, 1991; Stamoulis et al., 1996; Queralt et al., 1999). In consequence, the fingerprint properties of source material and sediment samples cannot be directly compared, even after sieving, unless a correction factor is utilised. Due to particle size selectivity during sediment transportation from source to river channel, the typical sediment sample is enriched in fines compared to the corresponding samples collected to represent the individual sources. In order to calculate a particle size correction factor, specific surface area (m2 g-1) is used as a surrogate measure for grain size composition because it exerts a key control on element concentrations (Horowitz, 1991). During the application of the mixing model for source type apportionment, the ratio of the specific surface area of the individual time-integrated suspended sediment samples to the

SID 5 (Rev. 3/06) Page 18 of 23

corresponding mean value for each individual source type is used. Although this approach assumes a linear relationship between fingerprint property concentration and specific surface area, it provides a pragmatic means of addressing the need to take explicit account of particle size selectivity.

The source type mixing model algorithm also includes an organic matter content correction since the latter also influences element concentrations in soil and sediment samples. This correction is calculated in the same manner as the equivalent for particle size, but using information on organic carbon content. Because the influence of particle size and organic matter content on element concentrations can be closely related, the combined use of the correction factors was carefully examined in order to ensure that the over-correction of the source material datasets was avoided. Sensitivity tests confirmed that the combined use of the particle size and organic matter content correction factors was appropriate for all applications of the revised sediment source type mixing model.

A weighting to reflect the spatial variation of individual tracers in each source was incorporated in the revised sediment source type mixing model. This new weighting was included to ensure that the fingerprint property values for a particular source characterised by the smallest standard deviation exerted the greatest influence upon the optimised solutions. It is logical that as the standard deviation of the fingerprint property values increases, the uncertainty associated with the source ascription also increases. The weighting was calculated using the inverse of the root of the variance associated with each fingerprint property measured for each source. The spatial variation weighting provided a means of representing the compound effect of a number of sources of uncertainty, including the variance of the tracer datasets for specific sources and the differing levels of precision associated with laboratory measurements of those tracers. Sensitivity tests, without the spatial uncertainty associated with the source material sampling, demonstrated that inclusion of the spatial variation weighting in the revised sediment mixing model reduced the average range of solutions for the contributions from each source type comprising the Mangelbeek and Gete sub-catchments for all flood events by 12.9% and 13.3%, respectively. The range of solutions pertaining to any specific sediment source type for the individual flood events was constrained by as much as 29% in the case of the Mangelbeek and 26% in the case of the Gete sub-catchment. These results clearly indicated that inclusion of a weighting for spatial variation in tracer properties is worthwhile.

The revised sediment source type mixing model algorithm also incorporated a weighting to reflect tracer discriminatory power (Equation 5). This weighting was based on information on the discriminatory efficiency of each individual tracer included in each composite fingerprint provided by the results of the DFA. Sensitivity analysis, suggested that the tracer discriminatory power weighting constrained the range of source type contributions generated by the Monte Carlo iterations by an average of 1% and 6.8% for the Mangelbeek and Gete sub-catchments, respectively.

The uncertainty of the results obtained using the revised sediment source type mixing model was investigated using a Monte Carlo framework. The mean and standard deviation of each fingerprint property for each source type were used to generate cumulative Normal distributions using a random number generator. A non-negativity constraint was implemented during the generation of the simulated Normal distributions. In future, the sediment data associated with replicate time-integrating traps could be used to inform uncertainty testing. The set of linear equations representing the composite fingerprint for either the Mangelbeek or Gete sub-catchment was repeatedly solved 5000 times. As a result, the computation of source apportionment data for the set of time-integrated suspended sediment samples collected for each sub-catchment amounted to 60000 model iterations.

The robustness of the optimised mixing model solutions was interrogated using the relative mean error (RME) or goodness of fit, viz.:

(8)

The average RME for the apportionment of the time-integrated suspended sediment samples was estimated at 15.4% (range 10.3%-20.3%) for the Mangelbeek and 8.6% (range 2.2%-25.4%) for the Gete sub-catchment. These estimates confirmed that the revised mixing model is capable of simulating meaningful time-integrated suspended sediment mixtures.

ResultsFigure 5 illustrates the mixing model output for the sediment source fingerprinting exercise in the Mangelbeek sub-

catchment. Eroding channel banks and subsurface sources were consistently the most important sediment source type, with their relative contributions ranging from 25±2% (27/09/2007-6/11/2007) to 31±2% (6/11/2007-18/12/2007). Actively eroding channel banks were found to be widespread along the channel system of the Mangelbeek during the source material sampling exercise. The important contribution of channel banks to downstream suspended sediment flux has been noted by previous investigations (e.g. Church and Slaymaker, 1989; Trimble, 1993; Novotny and Olem, 1994; Trimble, 1997; Collins et al., 1997a; Rosgen, 2001; Nelson and Booth, 2002; Evans et al., 2003; Collins and Walling, 2004). Eroding surface soils supporting cultivation consistently represented the second most important sediment source type with the contribution from this source being as high as 25±2% for the periods 24/05/2007-21/06/2007, 21/06/2007-6/08/2007, 6/11/2007-18/12/2007, 18/12/2007-28/01/2008, 28/01/2008-28/02/2008 and 28/02/2008-10/04/2008 (Figure 5). Autumn and spring sowing exposes bare tilled soils to the risk of water erosion and sediment mobilisation (Evans and Cooke, 1986; Boardman, 1990, 1993; Evans, 1993, 2002; Walling, 2005; Collins and Walling, 2007c). Authigenic Fe precipitation within the channel system was computed to contribute between 19±2% (20/04/2007-24/05/2007, 24/05/2007-21/06/2007, 21/06/2007-6/08/2007, 6/11/2007-18/12/2007, 18/12/2007-28/01/2008, 28/01/2008-28/02/2008) and 23±2% (28/02/2008-10/04/2008, 10/04/2008-22/05/2008) of the suspended sediment load sampled at Lummen. The important role of authigenic Fe in catchment sediment dynamics in Flanders has been noted by Vanlierde et al. (2007). Channel bed sediment remobilisation represented an important secondary sediment source with its relative contribution to sampled suspended sediment loadings being as high as 23±2% (18/12/2007-28/01/2008). The important contribution of remobilised channel bed sediment to sediment loadings in lowland agricultural catchments has been noted by Collins and Walling (2006). Eroding surface soils beneath pasture represented the least important suspended sediment source in the Manglebeek sub-catchment (maximum input of 9±2%) reflecting the limited spatial extent of this land use. Sediment loss from pasture fields reflected the impact of poaching and cattle treading in

SID 5 (Rev. 3/06) Page 19 of 23

gateways and around feeder rings in particular (cf. Morgan, 1985; Heathwaite and O’Sullivan, 1991; Heathwaite, 1994; Ministry of Agriculture, Fisheries, and Food, 1999; Evans, 1997; Collins and Walling, 2007c).

Over the duration of the sampling period (14/03/2007-22/05/2008), the mean relative contributions from the individual source types were estimated in the order channel banks/subsurface sources (27±2%), cultivated topsoils(24±2%), remobilised channel bed sediment (21±2%), authigenic Fe (20±2%) and pasture topsoils (8±2%). The lack of seasonal variation in sediment source contributions noted by previous research (e.g. Collins et al., 1997a) is likely to reflect the low connectivity resulting from the flat topography of the Mangelbeek sub-catchment. Under such conditions, the scope for time-variant source proportions is constrained, since the same few areas of erosion connected to the channel system will be contributing to downstream sediment transport. Further work is clearly required to convert the source proportions to actual loadings of sediment delivered to the channel network (e.g. Collins, 2008a). On the basis of the sediment source data, sediment mitigation options in the Mangelbeek sub-catchment need to focus upon protecting channel banks and interrupting sediment delivery from cultivated surface soils. The mitigation of sediment loss from pasture fields requires less attention.

0%10%20%30%40%50%60%70%80%90%

100%

14/03

/2007

-20/0

4/200

7

20/04

/2007

-24/0

5/200

7

24/05

/2007

-21/0

6/200

7

21/06

/2007

-06/0

8/200

7

6/08/2

007-2

9/08/2

007

29/08

/2007

-27/0

9/200

7

27/09

/2007

-6/11

/2007

6/11/2

007-1

8/12/2

007

18/12

/2007

-28/0

1/200

8

28/01

/2008

-28/0

2/200

8

28/02

/2008

-10/0

4/200

8

10/04

/2008

-22/0

5/200

8mea

n

Sampling period

% c

ontr

ibut

ion

Remobilised channel bedsediment

Authigenic Fe

Channel banks/subsurfacesources

Pasture topsoils

Cultivated topsoils

Figure 5: Sediment source information for the Mangelbeek sub-catchment

The corresponding results for the the sediment fingerprinting exercise in the Gete sub-catchment are presented in Figure 6. In contrast to the Mangelbeek sub-catchment, the mixing model iterations suggested that cultivated topsoils represented the dominant sediment source type, contributing up to 90±2% (14/03/2007-20/04/2007, 27/09/2007-6/11/2007 and 10/04/2008=22/05/2008). Widespread evidence of rill and gully erosion was observed in cultivated fields during source material sampling. The loess soils observed in the southern portions of the Demer study basin, represented by the Gete sub-catchment, are highly susceptible to erosion. Channel banks and subsurface sources were predicted to represent the second most important sediment source type in the Gete sub-catchment (Figure 6). The maximum contribution from this source type (32±2%; 24/05/2007-21/06/2007) is likely to reflect a period of channel disturbance e.g. associated with dredging or bank works, especialy as the input from remobilised bed sediment was also higher (13±2%) for the same period (Figure 6). Eroding pasture surface soils contributed up to 15±2% (6/08/2007-29/08/2007) with sediment loss from grassland again reflecting poaching damage around feeder areas. Channel bed sediment remoblisation was computed to be less significant in the Gete sub-catchment, contributing between 1±1%-13±2% of the suspended sediment load sampled at the Halen monitoring station.

Over the duration of the samping campaign in the Gete sub-catchment (14/03/2007-22/05/2008), the mean relative contributions from the individual sediment source types were estimated in the order cultivated topsoils (81±2%), channel banks/subsurface sources (11±2%), and pasture topsoils and channel bed sediment remobilisation (both 4±2%) (Figure 6). On the basis of these data, sediment management strategies in the Gete sub-catchment need to target the reduction of soil erosion and subsequent sediment delivery to watercourses from cultivated fields and, to a lesser extent, the protection of eroding channel bank sections.

SID 5 (Rev. 3/06) Page 20 of 23

0%10%20%30%40%50%60%70%80%90%

100%

14/03

/2007-

20/04

/2007

20/04

/2007-

24/05

/2007

24/05

/2007-

21/06

/2007

21/06

/2007-

06/08

/2007

6/08/2

007-2

9/08/2

007

29/08

/2007-

27/09

/2007

27/09

/2007-

6/11/2

007

6/11/2

007-1

8/12/2

007

18/12

/2007-

28/01

/2008

28/01

/2008-

28/02

/2008

28/02

/2008-

10/04

/2008

10/04

/2008-

22/05

/2008

mean

Sampling period

% c

ontri

butio

n

Remobilised channel bedsediment

Channel banks/subsurfacesources

Pasture topsoils

Cultivated topsoils

Figure 6: Sediment source information for the Gete sub-catchment

Policy implicationsThe need to manage diffuse sediment pollution is rising up the policy agenda due to the pivotal role sediment plays in

governing the transfer and fate of nutrients and contaminants and on account of its widely documented detrimental impacts upon freshwater ecology. Improved sediment management should target the key sources of the problem and should focus upon catchment scale understanding and mitigation. Against this background, it is increasingly recognised that sediment fingerprinting provides a valuable toolkit for assessing the sources of the sediment problem. Reliable information on sediment sources provides an invaluable evidence base to Catchment Officers for targeting mitigation options. Stakeholder engagement with respect to sediment management demands an evidence base on the principal sources of sediment pressures.

The Demer basin study in Belgium was specifically designed to provide further proof of the applicability of the fingerprinting approach for helping to target the management of the sediment problem. The findings suggest that the approach can be readily applied in larger drainage basins, such as the Demer, comprising several spatial sediment sources defined in terms of individual tributary sub-catchments. Equally, the results confirm that sediment fingerprinting can be used to apportion both primary and secondary sediment sources in lowland mixed land use catchments.

The findings of the Demer basin study suggest that:a) Sediment fingerprinting can be used to assist the spatial targeting of sediment mitigation options, highlighting the

most important tributary sub-catchmentsb) Sediment mitigation plans in areas of Flanders represented by the Mangelbeek sub-catchment need to focus, in

particular, upon improving the protection of channel banks and managing erosion on, and subsequent sediment delivery from cultivated fields

c) Sediment mitigation plans in areas of Flanders represented by the Gete sub-catchment need to focus, in particular, upon improving the protection of, and interrupting the delivery of sediment to watercourses from, cultivated topsoils and, to a lesser extent, channel banks

d) Authigenic Fe represents an important primary sediment source in areas of Flanders represented by the Mangelbeek sub-catchment

e) Bed sediment remobilisation can represent an important secondary sediment source in Flemish river basins

Potential future workGiven the increasing focus upon managing key sources of harmful sediment pollution by many Catchment Officers

across the EU, it would prove highly beneficial to compare current and future sources of sediment pressures. Present and future surveys of the provenance of sediment could provide the basis for demonstrating change due to mitigation stategies. Thus, for example, in a priority catchment where channel banks are being fenced or buffer strips at the bottom of cultivated fields are being encouraged, repeat sourcing surveys could be used for pre- and post-remediation source apportionment, thereby providing catchment managers with a measure of change. Such data would prove invaluable in engaging catchment stakeholders. Equally, empirical information on the efficacy of sediment mitigation options is urgently required to update the Diffuse Pollution Inventory User Manual (Cuttle et al., 2007). Sediment source apportionment surveys can be readily expanded to incorporate additional particulate-associated pollutants such as phosphorus, thereby further suporting the updating of the User Manual. Emphasis upon pollutant sources avoids the problems associated with delayed response at water quality monitoring points further downstream. It is important for mitigation programmes to provide evidence for positive outcomes in the short–term in order that stakeholders remain engaged and supportive.

Existing sediment source apportionment studies provide useful generic information on sediment sources that can be used by Catchment Officers to help target sediment mitigation strategies. But, there is clearly a need for improving the

SID 5 (Rev. 3/06) Page 21 of 23

resolution of sediment source data in order to help target mitigation options more effectively in relation to individual generic sources. In the case of pasture sediment sources, for instance, it would be extremely valuable for Catchment Officers to be able to apportion mitigation efforts between poached gateways, poached areas around feeder rings and water troughs and wider pasture sources. Equally, it would be extremely useful to be able to distinguish the relative inputs from poached channel banks and riparian areas from the corresponding sediment contribution derived from bank sources due to natural fluvial erosion. Future work should endeavour to deliver sediment source information at this resolution in order that capital grant options can be targeted more reliably.

Knowledge transferThe findings from the Demer basin sediment sourcing project have, or will be, used to support knowledge transfer in the

following manner:a) Paper presented at the 11th International Association for Sediment Water Science (IASWS) meeting in Esperance,

Australia, February 2008b) Paper submitted to Marine and Freshwater Science detailing experience in spatial sediment sourcing in the Demer

Basinc) Results and their wider implications for sediment policy support to be presented and discussed at a policy workshop

in Flanders, Belgium, planned for autumn 2008, involving the Flemish Environment Agency, Flanders Hydraulics Research, Universities of Ghent and Antwerp, ADAS UK Ltd., catchment stakeholder groups including farmers, local councils and water companies

d) Further scientific papers to be submitted to a selection of appropriate Journals

References to published material9. This section should be used to record links (hypertext links where possible) or references to other

published material generated by, or relating to this project.

SID 5 (Rev. 3/06) Page 22 of 23

The references cited in this report are listed in Appendix 1.

SID 5 (Rev. 3/06) Page 23 of 23