Quantifying Sediment and Associated Pollutants Sources in Agricultural Catchments Using Isotopic Techniques

Abstract

This chapter provides an overview of nuclear tools used in sediment source apportionment studies, summarising essential information about the use of fallout radionuclides, elemental geochemistry, and compound-specific stable isotopes. Exemplar studies are described to show how the method can inform decision making at farm and basin scales to control diffuse water pollution by agriculture.

7.1 Overview of Sediment and Associated Pollutants in Agricultural Catchments

In the context of agricultural pollution, sediment derived from soil erosion is widely regarded as a major pollutant both in its own right and in its role as a vector for particle-reactive agrochemicals. The widespread and costly environmental impacts of sediment transfer from agricultural land into downstream aquatic ecosystems are recognised by river basin managers and policy makers worldwide (Evans 2010), yet the complexity of sediment production and delivery processes renders it a complex challenge to mitigate. Soil erosion by water poses a serious threat to water and nutrient retention, biodiversity, and plant primary productivity on agricultural land (Pimentel 2006). Downstream, sediment, and nutrient export to rivers and lakes contributes to eutrophication and contamination of water resources by algae and associated natural toxins, affecting the functioning of wetland and lake ecosystems. Furthermore, siltation enhances the habitats that promote water-borne disease vectors and siltation of reservoirs compromises water supply and hydroelectric power (HEP) generation (Boardman and Poesen 2006). Herein, it is clear that enhanced runoff and soil erosion have a major impact on food, water and energy security (Blake et al. 2018). Future land-management decisions require quantification of the anthropogenic amplification of runoff and soil erosion processes, e.g., sediment yield from different agricultural land uses, so that hotspots of soil erosion can be targeted and controlled specifically by mitigation measures.

Application of tools to support source apportionment should be grounded in a thorough understanding of hydrological and soil erosion processes (Morgan 2005) that lead to sediment generation from agricultural land for specific study sites. The mode of overland flow development, i.e., infiltration excess vs saturation overland flow can have a critical bearing on tracer behaviour in some circumstances and the particle size selectivity of soil erosion (Issa et al. 2006) can strongly influence source signatures.

Sediment dynamics in river systems involve complex processes and their quantitative assessment faces many uncertainties. River sediment budgeting approaches provide understanding of sediment mobilisation, transport, storage, and yield (Walling and Collins 2008), a useful framework within which to evaluate evidence on sediment and pollutant source dynamics. River sediment budgets can be understood, in simplistic terms, as the mass balance between the sediment sources, deposition areas, and outputs. One of the most significant findings from budgetary studies is the importance (and magnitude) of sediment storage in river systems. From the total amount of sediment produced (i.e., eroded) in upland surfaces (i.e., sources), only a fraction makes its way to the basin outlet (i.e., sediment yield). This discrepancy has been named the ‘sediment delivery problem’ (Walling 1983), and it has been argued that various sediment storage mechanisms operating within a catchment explain this discrepancy (Trimble 1983; Walling 1983; Fryirs 2013). Since then, increasing research has been carried out to explain and disentangle the mechanisms, pathways, and fates of eroded sediment within a river basin, and evidence on source dynamics is a crucial ingredient to support management decisions at the basin scale.

In many catchments, sediment spends significant time stored in riverine compartments which can compound challenges of source apportionment within specific time frames. Its delivery is therefore controlled by storage and sporadic remobilisation at various timescales (Muñoz-Arcos et al. 2022). These storage units can be defined as transient, short-lived landforms that are frequently reworked by episodic events such as bars, lateral deposits, and the streambed where they play a key role in the (dis)connectivity of catchment sediment cascades (Fryirs 2013). Spatial and temporal aspects of sediment connectivity from hillslope to channel are important considerations in all source apportionment studies (Fig. 7.1).

Fig. 7.1

Scales of complexity in sediment tracing applications (Koiter et al. 2013)

Excess fine sediment (silt and clay fractions < 63 µm) poses a risk to receiving waters owing to both its physical presence as well as the potential for transport of co-contaminants. Many surface water systems are sensitive to changes in the transport and deposition of fine sediments owing to the impact of increased turbidity in the water column and siltation of channel bed habitats. For example, there is potential for siltation to affect oxygen demand and subsequently impact upon salmonid spawning grounds (Jensen et al. 2009; Sear et al. 2017). Excess suspended and stored sediment may also affect other, sessile biota such as the freshwater pearl mussel (Margaritifera margaritifera), which is critically endangered in Europe, according to the International Union for the Conservation of Nature (IUCN). Freshwater pearl mussels are sensitive to changes in water parameters, preferring oligotrophic and low turbidity conditions. For successful recruitment of juvenile freshwater pearl mussels, it is essential that the sediment in which juveniles establish has an open, well-oxygenated structure, enabling clear exchange between the interstices and the overlying water column. Excess fine sediment can negatively impact upon channel bed habitat by embedding coarse material and preventing exchange between interstices and surface (Geist and Auerswald 2007).

Transport of fine sediment from agricultural settings to receiving waters can also impact upon water quality through increased inputs of co-contaminants, which may exert direct or indirect effects depending on toxicity and bioaccumulations factors, and influence upon trophic structure. In relation to the latter, surface waters can be sensitive to elevated nutrient inputs associated with increased sedimentation, particularly with regard to phosphorus (P), often the growth-limiting nutrient in freshwater systems. Agricultural inputs can potentially increase the bioavailable pool of P, leading to eutrophication and associated ecological impacts (Withers and Jarvie 2008). Given that P is strongly associated with sediments, there is now a clear recognition of the need for improved management practices to reduce soil and associated phosphorus inputs to river channels (Ballantine et al. 2009; Ockenden et al. 2014). Significant improvements have been made to reduce the amount of dissolved P that enters watercourses from point source discharges across the EU. However, legacy sediment stored within channel and floodplain systems has the potential to act as a secondary source of P to the water column following disturbance (van der Perk et al. 2007) or in response to changes in condition of overlying waters (Jarvie et al. 2005b; Neal et al. 2012). Desorption and remobilisation processes are highly dependent upon environmental factors such as pH and oxidation–reduction conditions but, nevertheless, could offset the benefit of reductions in P inputs in the short term (Burns et al. 2015).

Sediment associated contaminants from agricultural settings are not limited to nutrients but can be wide ranging and related to biological (faecal) contamination, metals, and an array of organic compounds used in crops such as pesticides, herbicides and fungicides, and veterinary pharmaceutical treatments. Many organic compounds are highly particle reactive and persistent and are, thus, now restricted in their usage. However, the persistence of legacy inputs in sediment storage zones may present a continued threat to aquatic systems (Holden et al. 2017).

7.2 Approaches and Theory for Sediment Source Tracing

7.2.1 Core Principles of the Methodology

Among the sediment source tracing methods, the sediment fingerprinting approach has been widely used to apportion sediment sources at different landscapes and catchment characteristics (Walling 2005; Owens et al. 2016; Collins et al. 2017, 2020). Sediment fingerprinting approaches are based on identification of statistically significant differences in soil properties between target river basin sediment source areas. If signatures of downstream mixtures can also be derived, and the key assumption that signatures are not altered during hydrological and fluvial transport (i.e., conservatively), then qualitative or quantitative comparison between source and mixture fingerprints can elucidate dominant source contributions (Fig. 7.2). Differences in source properties and ‘fingerprints’ can occur due to a range of natural and anthropogenic processes generally linked to the geological substrate, cultivation practice, land use, and atmospheric depositions or pollution.

Fig. 7.2

Example experimental designs to link primary sources to sediment mixtures (Blake et al. 2018) where rivers flow downward, filled circles represent nodes at which sediment mixture (Mix = M) samples could be collected, and dashed grey lines delineate watersheds (S) denoted by subscript numbers, and subscript letters indicate unique sources. (A) Simple watershed with three sources, SA-C, and one mixture location at the outflow, M. (B) Longitudinal system with four sources, SA-D, and multiple mixture locations at the outlet of each nested subwatershed, M1–4. (C) Distributed system with mixtures at the outflow of each of three subwatersheds, M1, M2, and M4, four sources (SA-D), as well as mixtures on the main channel: M3 and M5. Note not all sources are present in all subwatersheds

Properties that have been used to discriminate sources include: (i) fallout radionuclides (FRNs) (e.g., ¹³⁷Cs, ⁷Be, and excess ²¹⁰Pb_xs defined as the difference in activity concentrations between ²¹⁴Pb and total ²¹⁰Pb) which effectively label surface soil material permitting discrimination of surface, subsurface, and cultivated sources (e.g., Mabit et al. 2008; Wilkinson et al. 2009); (ii) major and minor element geochemistry which is related to geological substrate but is also modified by soil formation processes and weathering and therefore can discriminate between land use (e.g., Collins et al. 2010; Laceby and Olley 2015); (iii) Compound Specific Stable Isotopes (CSSIs) which are organic fingerprints related to soil plant cover (e.g., Blake et al. 2012; Bravo-Linares et al. 2018; Upadhayay et al. 2018b); (iv) mineral magnetic properties which are sensitive to soil formation processes and pollution (e.g., Ley and Devon 1998; Blake et al. 2006); (v) contaminants from industrial or other anthropogenic activities, e.g., heavy metals (e.g., Rothwell et al. 2005); and (vi) sediment colour (e.g., Pulley and Collins, 2021, Barthod et al. 2015).

Collins et al. (2020) provide comprehensive reviews of the wider methodology and key considerations in source apportionment applications for the user community. Here, we focus specifically on practical application of techniques that fall within the ‘nuclear tools’ domain: fallout radionuclides (FRNs), X-ray fluorescence (XRF) elemental geochemistry, and Compound Specific Stable Isotopes (CSSI).

7.2.2 The ‘Nuclear Tools’ Tracer Toolkit

Fallout radionuclides (FRNs) are delivered to surface soil via wet and dry fallout mechanisms. The three FRNs applied in source apportionment studies are ¹³⁷Cs, ²¹⁰Pb, and ⁷Be. Cs-137 is a legacy radioisotope derived from thermonuclear weapons testing in the twentieth century with most fallout occurring during the 1950s and 1960s, and locations affected by fallout from nuclear power plant accidents, notably the Chernobyl incident in the 1980s. Fallout ²¹⁰Pb is the unsupported component of total ²¹⁰Pb present in soil and sediment. The supported component is in equilibrium with its geogenic parents in the uranium series. The unsupported fallout component enters the atmosphere via diffusion of its intermediate parent ²²²Rn from regional bedrock and weathered substrates. Be-7 is produced by cosmic ray spallation of oxygen and nitrogen in the upper atmosphere.

All three FRNs are delivered to the surface predominantly by rainfall with the assumption of instantaneous binding at the soil surface. As with all tracer tools, such assumptions need careful consideration and evaluation in the specific context of any study site. For example, FRNs have been shown to be preferentially associated with fine-grained materials (He and Walling 1997; Taylor et al. 2012) raising critical questions about size selectivity of transport and influence on concentration-based tracer signatures (Laceby et al. 2017a). The different half-lives of the three isotopes and hence residence time in the soil profile is a key controlling factor of their utility as source tracers. With its half-life of ~ 30 years, ¹³⁷Cs delivered to the soil profile are subjected to downward translocation by both bioturbation and geochemical diffusion and migration. Furthermore, ploughing of the soil surface mixes ¹³⁷Cs uniformly through the plough layer. Similarly, the ~ 22-year half-life of ²¹⁰Pb means its fallout component in surface soil is subject to similar processes although it is less geochemically mobile than ¹³⁷Cs, and it tends to have a shallower depth profile in the soil. Ploughing of course leads to uniform mixing as for ¹³⁷Cs. The short half-life of ⁷Be, 53 days, means it has a uniquely different profile in the soil compared to the longer-lived FRNs. The FRN remains present in the uppermost millimetres of the soil since activity concentrations of material bioturbated are rendered below detection limits by radioactive decay. If a soil surface is ploughed, the ⁷Be activity of the surface is diluted to below detection limits effectively resetting the surface signature. The ⁷Be signature of the soil surface is also highly dependent on fallout dynamics linked to spatial and temporal rainfall patterns (Taylor et al. 2016).

There are two useful models of FRN-based source signatures. The first (Wallbrink and Murray 1993) was developed for uncultivated scenarios representing degraded land subject to sheet wash and incision processes (Fig. 7.3a). In this context, variability in the activity concentrations of ⁷Be and ¹³⁷Cs of downstream sediment can be linked back to sediment production processes. The second (Walling and Woodward 1992) is based on typical agricultural conditions (Fig. 7.3b) with a mixed cultivated and uncultivated land cover allowing apportionment to land use and process drivers.

Fig. 7.3

Models of FRN application as a source tracers a for degraded landscapes subject to sheet wash and incision (Wallbrink and Murray 1993) and b agricultural settings with mixed cultivated and uncultivated land cover (Walling 2012)

XRF elemental geochemistry

Analysis of stable elements offers complementary data to CSSI and FRNs to provide multi-parameter datasets capable of improved source discrimination. Within a river catchment, source materials can be identified a priori based upon hydrological connectivity of land (source) units or via field observation. Element profiles may vary naturally in line with underlying geology or be altered by contaminant inputs such is the case for road-derived or mining-related materials for example. Agricultural practices can influence surface soil signatures through enhanced weathering of cultivated soils, such that differences between surface and subsurface element concentrations can often be detected. This may be further enhanced by atmospheric deposition at the soil surface (Smith and Blake 2014). Field survey sample numbers are dependent on the size of the study area, the range of potential sources, and gaining representative samples involves repeat sampling of each source type. Subsequently, tracing studies often require large sample numbers to be processed and characterised for element concentrations. The choice of analytical technique, therefore, requires careful consideration. Techniques such as inductively coupled plasma optical emission spectrometry (ICP-OES) and inductively coupled plasma mass spectrometry (ICP-MS) have the advantage of low detection limits (≤ µg/kg), which offer potential for a wide range of tracers to be measured. However, these techniques measure elements in the aqueous phase, which requires the use of hazardous (e.g., strong acids) and time-consuming digestion methods prior to measurement. X-ray fluorescence (XRF) spectrometry offers an alternative and potentially rapid approach for measuring a broad range of elements in solid materials such as soils and sediments. XRF spectrometry provides total element concentrations and can, if required, be undertaken in a non-destructive manner on loose powder material, which is particularly useful where sample mass is limited. As well as laboratory-based instrumentation, the use of portable XRF instruments offers an opportunity for in situ screening of contaminant hotspots and source materials (Turner and Taylor 2018).

Compound Specific Stable Isotopes (CSSI) offer complementary/specific information to FRNs and geochemical approaches. Elemental geochemistry, fallout radionuclides, and mineral magnetism are sediment fingerprints that have been successfully applied to identify and apportion sediment sources within a catchment. However, they are limited when determining specific land use-derived sediment sources. The CSSI technique offers the possibility of refining other fingerprinting techniques by potentially allowing discrimination of different land uses based on the stable carbon isotopic composition of fatty acids δ(¹³C)(_FA). CSSI refers to isotopes of individual compounds in complex mixtures rather than the isotopic signature of the bulk sample. Sources of different land uses can be identified based on stable isotope characteristics of specific molecules in vegetation communities. Soil contains a wide variety of organic compounds, which vary significantly in stability and structure, and therefore, the isotopic signature of a bulk soil is not expected to be conservative. Consequently, the need to target specific organic compounds appears. Plants synthesise complex organic molecules during photosynthesis using carbon dioxide. They can produce the same compounds (e.g., fatty acids) but with different isotopic compositions because of isotopic fractionation which will reflect the different photosynthetic pathways of the plants (Chikaraishi and Naraoka 2003; Reiffarth et al. 2016). Plant-derived FAs are incorporated into the soil through rhizodeposition and decomposition of organic matter transferring the δ(¹³C) signature of the vegetation from which they are released. Fatty acids become bound to mineral and clay soil particles making them persistent for long periods of time (Gibbs 2008; Reiffarth et al. 2016; Upadhayay et al. 2017a). Although FA concentrations may change overtime due to degradation by microorganisms, volatilisation, dilution, and dispersion, they do not cause significant isotopic fractionation (Gibbs 2008). However, the use of the very long-chain fatty acids (VLCFAs) is particularly recommended for this purpose since they are predominantly plant-derived and more resistant to degradation in soils and sediments which avoids the influence of external sources such as contamination by microorganisms and algae and reduces the risk of isotopic fractionation (Reiffarth et al. 2016; Upadhayay et al. 2017b). Due to this fingerprinting method based upon qualitative data (δ(¹³C)(FA)), a further correction by the concentrations of each fatty acid in the source material considered in the fingerprinting approach is required (Upadhayay et al. 2018a). Different relative tracer concentrations in potential sources influence apportionment calculations, and the need to consider concentration-dependency in isotope mixing is strongly recommended (Upadhayay et al. 2018a).

7.2.3 Linking Source Materials to Downstream Mixtures

Quantitative comparison of source signatures to downstream mixtures is achieved through application of mixing models. With increasing attention to challenges of uncertainty and representativeness alongside the need to harmonised approaches, methodologies have emerged with exemplar structures and data to support user application. Recent innovations include application of MixSIAR (Stock et al. 2018) in river basin contexts (Blake et al. 2018); FingerPro purpose designed for the soil sediment continuum (Lizaga et al. 2020a), and the SIFT framework (Pulley and Collins 2018). Recent publications all provide code and /or open-source software to support user implementation. Nevertheless, a key element before performing any calculation is the careful choice of tracers (see also Chap. 3). Tracer selection is a crucial step for any fingerprinting study in which tracer conservativeness, source discrimination, and mixing model performance should be evaluated. Several strategies have been employed to select the most suitable tracers for sediment source apportionment such as boxplots, biplots, multivariate analyses, Kruskal–Wallis, and discriminant function analysis (DFA) tests, among others (Collins et al., 2020). For discussion about the performance of different tracer selection methods, the reader is directed to other works (Palazón and Navas 2017; Smith et al. 2018; Lizaga et al. 2020b).

7.3 Sample Preparation Techniques and Measurement

7.3.1 Understanding the Origin of Sediment Source Fingerprints

In all tracer applications, successful implementation must be grounded in a thorough process understanding of the origins of tracer properties and the expected differences between sources. A hypothesis-led approach to source characterisation is recommended to connect the study to established theoretical contexts but also, through the hypothesis testing process, to identify attributes that are unique to the study site in question.

For example, in the context of FRNs and questions of sediment contributions from different land use within the catchment, the starting point is to hypothesise the relative proportions of the three FRNs in each land use based on local environmental knowledge of catchment land use history over recent decades and months. A strategic sample design programme can then be implemented to evaluate consistency of signature within source type areas and confirm differences between types (Fig. 7.3).

In the context of XRF geochemical signatures, a useful starting point is the consideration of the broadest control on differences within the local geology, i.e., weathering profiles and surface versus subsurface material. From here the influence of land use can be hypothesised, e.g., mixing of soil by ploughing to create new signatures and/or amendment by agrochemicals.

In the context of CSSI applications, source samples are targeted to those representing the major land uses in the studied catchment. Sampling should focus mainly on those land uses which are supposed to be susceptible to erosion. To account for spatial variability in δ(¹³C)(_FA) values, the collection of enough replicates of every source is recommended to properly represent the land use. The CSSI sampling is focused on the uppermost soil layers as this layer is the most susceptible to erosion. Therefore, the sampling depth is usually 20 mm, i.e., the soil layer in which most of the organic matter is found. In some cases, for assessment of gully erosion, road erosion, and riverbank erosion, the sampling could involve also much deeper layers.

A sampling programme that represents all potential sources is of course the goal in any study, but users must also be aware that a missing source will not be identified by the mixing model—the model will always report to a sum of unity whether all sources have been represented by sampling or not.

7.3.2 Sampling Strategies

Sampling strategies should naturally be adapted to the specific challenges of each scale of study (Fig. 7.1). With increasing spatial scale comes increased field sampling, processing and analytical demand. Herein, it is useful to consider two broad approaches to source definition that can be tailored to catchment-specific sample designs.

The primary source approach (Fig. 7.2) is a straightforward characterisation of each primary source within the system of interest and comparison to mixtures downstream in lower-order channels. The use of spatially integrated composite samples from primary sources is widely recognised as a means to ensure representative characterisation of properties. For example, one sample might comprise 10–30 smaller samples taken from a designated area within the source zone. Replication is essential, and the user must satisfy themselves that the number of samples for a source classification is both environmentally and statistically representative to permit effective source discrimination. At the large catchment to basin scale, characterisation of all primary sources can become unwieldy and stretching resources as well as introduce complexity in unmixing model development. In this case, a tributary approach might prove useful where the mixtures in higher-order channels or receiving water bodies are compared back to the sediment mixtures in transit or storage in lower-order tributaries upstream (see also Chap. 2). The advantage of the approach is that hydrological processes above the tributary source end member will have created a spatially, and potentially temporally, integrated sample at the subcatchment outlet. A disadvantage is where management practice and erosion response within the lower-order tributary catchment areas are heterogeneous and complex. A hybrid model might use specific tributaries of known land use and/or source type as representative of a specific source end member, assuming other controls on soil properties e.g., geology are uniform. These challenges are explored in the case studies (Sect. 7.4).

7.3.3 Laboratory Analysis

Measuring FRNs via gamma spectrometry

Detail on gamma spectroscopic methods to determine FRN activity concentrations in soils and sediment is provided in detail by Iurian and Millward (2019). For all tracer applications, users need to carefully consider the particle size fraction that they wish to compare between source and sink (Laceby et al. 2017). Critically, one must compare like with like and further evaluate particle size assemblages to verify.

Once screened to the relevant particle size (e.g., < 63 µm) samples must be sealed in gas-tight and stored for 21 days to permit equilibration between ²¹⁴Pb and its parent radioisotope ²²⁶Ra prior to measurement by gamma spectrometry (Appleby 2001). Activity concentrations of the target radionuclides are measured using a low background HPGe Gamma spectrometry system. Instruments should be calibrated using soil material spiked with certified mixed radioactive standards supplied by a certified provider. All calibration relationships are generally derived using proprietary software and should be verified by inter-laboratory comparison tests, e.g., with reference materials supplied by the IAEA via worldwide proficiency tests. Total ²¹⁰Pb is measured by its gamma emissions at 46.5 keV and its unsupported component calculated by subtraction of ²²⁶Ra activity, which in turn was measured by the gamma emissions of ²¹⁴Pb at 295 and 352 keV. ¹³⁷Cs is determined by its gamma emissions at 662 keV (with correction for ²¹⁴Bi emissions). Short-lived ⁷Be is analysed via its emission at 477 keV. Count times typically range 24–48 h subject to activity concentrations.

XRF spectrometry for elemental geochemistry

X-ray fluorescence (XRF) spectroscopy enables total element concentrations to be determined without the need for time-consuming and hazardous digestion procedures and has, thus, been widely applied in aquatic and marine contaminant research. XRF spectrometry is often employed as a complementary analytical technique in tracing studies with the potential to provide data for a wide range of elements (typically Na to U) at mg/kg concentrations. There are two major branches of XRF spectrometry: energy dispersive XRF (ED XRF) and wavelength dispersive XRF (WD XRF). Both techniques employ the same fundamental principles of ionisation in the sample material although the techniques differ in their dispersion and detection of secondary X-rays. WD XRF spectrometers provide the advantage of improved resolution and sensitivity for low energy elements (particularly Na to P), whereas the simultaneous nature of ED XRF spectra analyses provides a time advantage over WD XRF (the latter typically employing sequential detection of elements). Despite the differing detection methods, sample matrix and the manner in which the sample is presented to the spectrometer are equally important to both techniques. This is particularly relevant for the analysis of trace elements in sediments, which are often analysed as pressed powder pellets. Sample homogeneity is a key consideration for the operator, and it is crucial that the sample preparation ensures that particle size in the sample is uniform and below ~ 50 µm to reduce particle size and shadowing effects in the analysis layer (Willis et al. 2011). If the particle composition is also heterogeneous, such that an analyte is present in more than one mineral form, mineralogical effects can occur whereby the intensity of the analyte line is affected by differing attenuation properties of the particles. These matrix effects can only be eliminated by fusion methods, which involve mixing the sample with a flux material, typically lithium tetra/metaborate, at high temperature. The resulting glass disc provides a homogenous geometry for analysis (Willis et al. 2014). Owing to the flux to sample ratios used in the fusion process, XRF analysis of fused materials is usually limited to major elements, which will remain detectable even when diluted by the addition of flux.

Sample preparation and extraction for CSSI signatures

Fatty acid (FA) extraction is the first key step in the process. Several methods can satisfactorily extract FAs, namely Accelerated Solvent Extraction (ASE); Microwave-Assisted Extraction (MAE); Ultrasonic-Assisted Extraction (UAE) and Soxhlet. A detailed description and assessment of the performance, advantages and disadvantages of these methods can be found elsewhere (Jeannotte et al. 2008; Hewavitharana et al. 2020). FAs bound to the soil matrix can be extracted using hot/pressurised nonpolar solvents such as dichloromethane (DCM), or a mixture of solvents (e.g., DCM/Methanol) depending on the extraction method used. Samples must be dried before performing any extraction. This can be done by drying at 60 °C in a laboratory oven or using a freeze-dryer. Drying with anhydrous sulphate in direct contact with the wet sample or air-drying is also an alternative when no instruments are available. When using Soxhlet, soil, and sediment samples plus a pinch of anhydrous Na₂SO₄ (used to remove sample moisture) are put in cellulose filter containers, about 250–300 mL of DCM is added to each Soxhlet balloon, and heaters are turned on keeping temperature at 40 °C. The solvent will evaporate and condense in the reflux flask, drip onto the filter container and then the extract is collected into a round-bottom flask. When using ASE, soil and sediment samples are placed in stainless-steel cells and a proper extraction programme is selected. The DCM is injected into the cell, and the sample is heated and pressurised with N₂ (at 100 °C and 2000 PSI). The sample extract is then collected in a sealed vial. For MAE, samples are placed in extraction chambers, and depending on the sample amount, DCM:Metanol (4:1) solvent mixture is added and a proper extraction programme is selected (combining temperature, time and power). When using UAE, samples are transferred into Schott Duran flasks. The sample (10–20 g of dry sample) is then soaked in 100 mL of DCM and placed in an ultrasonic bath for at least 30 min. For both, MAE and UAE, samples must be filtered after the extraction process to remove particles using a suitable filter paper (e.g., Grade 1 Whatman). In addition, it is suggested that the extraction procedure for both (MAE and UAE) can be performed twice to assure quantitative (i.e., complete) extraction of the analytes. In this case, filtrates are combined. To reduce the solvent volume in the solution, a rotary evaporator is used. The temperature is set accordingly (around 40 °C depending on the solvent used), and the reduction process is interrupted when 2–5 mL of the solution remains in the rounded flask. The concentrated sample is then quantitatively transferred to a 10 mL vial using a Pasteur pipette. The vial is then inserted into a water bath at 40 °C, and the extract is subjected to a current of ultrapure N₂ to complete dryness.

Fatty acid derivatisation is the next step. Derivatisation for gas chromatography (GC) analysis is performed to increase volatility, improve separation and reduce tailing of the analytes. Because FAs have high polarity and low volatility, derivatisation is highly recommended (Brondz 2002). In this process, the polar carboxylic acid is converted to a methyl ester. The derivatisation process (i.e., esterification) results in a less polar molecule called Fatty Acid Methyl Ester (FAME). Although, there are several methods reported in the literature for FA derivatisation with reproducible results (Brondz 2002), the method described below is based on the use of methanol as derivatisation agent catalysed with boron trifluoride (BF₃). It is essential to remember that a portion of the methanol used during the derivatisation process must also be sent to the analytical laboratory to obtain the carbon isotopic value of the added methyl group. This will be further used to correct the isotopic values of the FAMEs. It is advisable that all reagents should be prepared daily before starting the derivatisation procedure. To the dry extract, 1 mL of 5% of BF₃ in methanol (BF₃-MeOH) is added and mixed on vortex for 2 min. The samples are then placed in a test tube rack in a fan oven/block heater at 70 °C for 20 min and cool down at room temperature. Then, 1 mL distilled water and 1 mL of hexane/DCM (4:1) mixture are added to the samples and mixed on vortex for 1 min. Posteriorly, samples are let to stand to allow the solvent layers to separate. Carefully, the upper layer (organic phase) is transferred from the 10 mL screw cap test tube to a 2 mL vial by using a Pasteur pipette. The procedure is repeated, but this time adding only 1 mL of hexane/DCM mixture and transferring the upper layer again. At the end, the vial will contain about 2 mL of the FAME extract. It must be stored at 4 °C until analysis or fully dried with ultrapure N₂ if samples are being analysed in an external laboratory.

The δ(¹³C) values of the FAMEs are adjusted using the following formula to account for the addition of the methyl group added (Gibbs 2008):

$$\delta (^{13} )C_{{{\text{FA}}}} = \frac{{\delta (^{13} C)_{{{\text{FAME}}}} - \left( {1 - X} \right)\delta (^{13} C)_{{{\text{methanol}}}} }}{X},$$

where $X$ is the fractional contribution of the free fatty acid to the methyl ester (e.g., 22/(22 + 1) for a FA of 22 carbons); δ(¹³C)_methanol is the isotopic composition of the methanol used during the derivatisation process; δ(¹³C)_FA refers to the corrected isotopic composition of the FA, and δ(¹³C)_FAME is the isotopic composition of the FAME provided by the instrument.

IRMS analysis for CSSI

A gas chromatography-combustion-isotope ratio mass spectrometer (GC-C-IRMS) is used to analyse the δ(¹³C) value of fatty acid esters. Samples are injected into the gas chromatograph. Then the sample is vaporised at 280 °C. Thus, the analysed samples in a GC-C-IRMS system must be volatile. Subsequently, the FAMEs are pushed with ultrapure helium gas to pass through the chromatographic column, whose function is to separate them by molecular weight/boiling point and affinity with the stationary phase of the GC column.

From the GC column, the FAMEs go to the combustion reactor, whose function is to convert the compounds into CO₂ and H₂O. This reactor consists of an Alumina tube (Al₂O₃) at 1000 °C. Its internal part contains a wire (composed of copper, nickel, and platinum) that is purged with oxygen allowing oxidative combustion, where nickel acts as a catalyst. The combustion products pass through a moisture trap, where the water present in the gas mixture is removed. This trap consists of a selectively permeable membrane with a dry helium counterflow. The resulting CO₂ is directed to the IRMS device and δ(¹³C) for each fatty acid can be obtained. From an isotope ratio mass spectrometer (IRMS), it is possible to measure isotopic composition at low enrichment and natural abundance levels. It means that low variations in exceedingly small amounts of the heavier (or less abundant) isotope are detected in the presence of large amounts of the lighter isotope with high precision and accuracy.

Particle size analysis

Tracers often display preferential association with particle size classes in source and channel material, with enrichment in finer particle sizes common for many elements and FRNs (Taylor et al. 2014; Laceby et al. 2017b). Where particle sorting occurs during sediment transport and storage, it is important for the user to account for particle enrichment during tracing studies. Many studies account for sorting effects by directly analysing transported size fractions in both the source and stored sediment material. Other studies apply particle size correction factors based upon empirically derived relationships between particle size and tracer concentrations (Smith and Blake 2014; Laceby et al. 2017a). Particle size analyses of source and transported material may, therefore, be required. The traditional pipette method is based upon the use of a settling column and Stokes’ Law to determine particle size by subsampling the suspension at a predefined depth and time. A more contemporary approach combines the use of Stokes’ law with a paralleled X-ray beam to determine sediment concentration in the settling column (SediGraph). Laser diffraction is a commonly applied approach for particle size analysis offering an effective means of analysing large sample numbers and has been shown to provide good agreement with digital image analysis (Bittelli et al. 2019).

7.4 Applying Nuclear Tools for Tracking Sediment Pollution from Agriculture (Case Study)

7.4.1 Applying FRNs to Identify Agricultural Runoff Source and Connectivity

Context

The River Mease, located in central UK, is a mixed agricultural river basin in a temperate maritime climate. The land use comprises a range of cultivated areas growing fodder maize, potatoes and cereal crops with areas of pasture for livestock grazing where field sizes are of the order of 5–15 ha. The topography is gently undulating with low to moderate slopes. At the time of study (Blake et al. 2013), the main river system was classified as being in an ‘unfavourable condition’ for aquatic habitat with respect to high silt and phosphorus concentrations in the riverbed substrate.

Sediment and associated diffuse phosphorus pollution from agriculture were considered to be significant contributors to unfavourable condition. Water quality modelling had suggested that in the region of 40 % of dissolved phosphorous, inputs to the catchment were derived from diffuse sources, with agricultural sources likely to be the main contributor. In parallel, sewage treatment works and domestic septic tanks had also been identified as potentially important sources of phosphorus, wherein the former was being tackled with improvements to phosphorus removal. The debate surrounding agricultural versus domestic sources of phosphorus was an important management question but was constrained by evidence being limited to phosphorus in the dissolved phase. Given that the greatest proportion of phosphorus delivered from hill slope to channel is in particulate form (Owens and Walling 2002) and potentially exchangeable with surface water once in aquatic storage (Jarvie et al. 2005a, b), a key research question arose regarding sediment sources in relation to agricultural activities.

Sampling and analysis

To characterise source materials, ten spatially integrated samples were collected from representative areas of each primary source type that were relevant to FRN characterisation. These were cultivated soil, uncultivated soil, subsurface drains, soil material deposited on roads within the agricultural areas (considered as secondary sources) and stream channel banks.

To characterise sediment mixtures, samples of sediment stored within the channel bed matrix of six tributaries were collected in triplicate using a stilling well methodology early in the winter ‘wet season’ period. Suspended sediment samples were collected via time-integrated samplers for early and later stage rainstorm events during the main winter period. All samples were freeze-dried and gently disaggregated and sieved to < 63 µm. Samples were sealed into 50 mm Petri dishes and stored for 21 days to permit equilibration between ²¹⁴Pb and its parent radioisotope ²²⁶Ra prior to measurement by gamma spectrometry (Appleby 2001). Activity concentrations of the target radionuclides were measured using a low background EG&G Ortec planar (GEM-FX8530-S; N-type) HPGe gamma spectrometry system at the University of Plymouth Consolidated Radio-isotope Facility. The instrument was calibrated using soil material spiked with certified mixed radioactive standards supplied by AEA Technology Plc. All calibration relationships were derived using EG&G Gamma Vision software and verified by inter-laboratory comparison tests with materials supplied by the IAEA, namely the worldwide proficiency test using moss soil (IAEA-CU-2009-03). Total ²¹⁰Pb was measured by its gamma emissions at 46.5 keV and its unsupported component calculated by subtraction of ²²⁶Ra activity, which in turn was measured by the gamma emissions of ²¹⁴Pb at 295 and 352 keV. ¹³⁷Cs was determined by its gamma emissions at 662 keV (with correction for ²¹⁴Bi emissions). An additional element to this study was analysis for short-lived ⁷Be, analysed via its emission at 477 keV. Count times were typically 48 h due to low activity concentrations and low sediment sample mass.

Results and discussion

The three fallout radionuclides measured provided important information on sediment source dynamics. ⁷Be concentrations were observed to be high in all surface soil samples. The other FRN signatures of cultivated and uncultivated surface soil showed a pattern seen typically in previous UK studies (Walling and Woodward 1992) whereby uncultivated soils carried higher activity concentrations of ¹³⁷Cs and ²¹⁰Pb_xs due to the shallow depth profiles of both radionuclides in undisturbed soil. Cultivated soil represents a mix of the surface and shallow subsurface soil materials which effectively lowers the activity concentrations of both FRNs. Channel banks materials had measurable amounts of ¹³⁷Cs and ²¹⁰Pb, but lower than surface soils but undetectable activity concentrations of ⁷Be. Material from field drain excavations had very low FRN concentrations across all three FRNs. The road-transported sediment in agricultural areas exhibited the greatest ²¹⁰Pb_xs activity concentration of all materials sampled. These high activities were likely due to scavenging of ²¹⁰Pb_xs by soil in transit from surface pluvial runoff on the impervious surface (i.e., binding to suspended sediment in preference to the surface) and high particulate-water interaction time (Charlesworth and Foster 2005). Given that the road dust samples were collected from rural lanes at the bridging points where channel access was gained, they are highly representative of material transported to the channel via the road network. The material itself was likely to comprise sediment transported by run-on from agricultural fields, the FRN signature of which is transformed by ²¹⁰Pb_xs enhancement as above.

The FRN activity concentrations (Table 7.1) of sediment mixtures in the six tributaries showed notable differences between the channel-stored sediment and suspended sediment in terms of ⁷Be concentration. The predominant undetectable-to-low activity concentrations of ⁷Be in the channel-stored sediment sampled at the beginning of the wet season indicated either a long residence time (i.e., greater than 5 half-lives, ~ 250 days) or a predominant channel bank source. The activity concentrations of ¹³⁷Cs and ²¹⁰Pb of these samples were within the bank to cultivated land zone (see Fig. 7.4). The switch to ⁷Be rich material in most of the suspended sediment material collected indicated first, and most importantly, that a large proportion of the material in transit through the wet season to January was freshly eroded from an exposed soil surface, although there was a wide range in these values. This range could have reflected varying contributions of remobilised ‘old’ sediment from the bed or contributions from subsurface and bank sources (which have been shown to be deficient in ⁷Be). The suspended sediment sample from catchment 2 was an exception, with undetectable ¹³⁷Cs and low ⁷Be suggesting a substantial contribution was made from a subsurface source, whereas the other samples from this time indicated surface soil with varying contributions from cultivated and uncultivated land. In the later period, this system was more in line with the other sites in terms of freshly eroded material in transit. It was notable that the three catchments with the highest ⁷Be content in the second set are the three that had the greatest particulate phosphorus concentrations again implying fresh sediment delivery from intensively farmed land. Across all suspended sediment samples, ⁷Be showed a weak but positive correlation (r = 0.33), with particulate phosphorus which strengthened markedly (r = 0.8) when low particulate phosphorus outliers from catchments 1 and 5 were not considered. The close affiliation of ⁷Be, indicating recent surface erosion, and particulate phosphorous concentrations in catchments 2, 3, 4, and 6 suggested mobilisation of soil from intensively farmed land that was well connected to the stream network via roads (linked to enhanced ²¹⁰Pb_xs).

Table 7.1 FRN activity concentrations (Bq kg⁻¹) of channel-stored and suspended sediments sampled in phase 2

Fig. 7.4

FRN signatures for primary source materials where outlines represent range extent and uncertainty is one standard deviation of the mean, shown by the symbols

The indication from the FRN data is that the material stored in the subcatchment channels has a high residence time (deficient in ⁷Be cf Wilson et al. 2007) with activity concentrations of other FRNs in line with catchment soil sources. While this material might comprise bank slumped material, which has been shown to also carry FRN signals similar to cultivated land in this system due to exposure in the past or banks comprising reworked catchment materials, there was little evidence of bank erosion in the field.

The ⁷Be content and particulate phosphorus loading of the suspended material collected during the major storm events, along with the other FRN signals, demonstrated a substantial contribution from surface soils in catchments 2, 3, 4, and 6, wherein roads were an important connectivity pathway linking the agricultural runoff, containing sediment and phosphorus pollution, to the rivers.

7.4.2 Applying XRF Signatures to Evaluate Contribution of Agricultural Runoff to Aquatic Habitat Siltation

Context

Freshwater Pearl Mussels (FWPM), M. margaritifera, are among the most critically threatened freshwater bivalves worldwide. In addition to their important roles in particle processing, nutrient release, and sediment mixing, they also serve as an ideal target species for evaluation of aquatic ecosystem functioning especially of their symbiotic relationship with Atlantic salmon Salmo salar and brown or sea trout Salmo trutta. Poor water quality, particularly eutrophication, and siltation are considered major contributory factors in the decline of the species; hence, management of diffuse water pollution from agriculture is a key priority in catchments that host FWPM habitats. The river Clun, UK, is one such system where FWPM have been identified to be under threat (Blake et al. 2016).

FWPM populations have declined dramatically throughout its range in Europe and North America, and recruitment of juvenile mussels only occurs in a small fraction of the streams where populations still persist. There are a range of environmental threats FWPM populations linked to direct pollution and damage to habitat. Indeed, there had been a decline of more than 90 % in European populations by the 1990s, and it is accepted that this has continued or even increased since (Geist and Auerswald 2007). Currently, the main concern for FWPM populations in general is that juveniles do not seem to be reproducing, so populations are becoming more elderly. Poor water quality, particularly eutrophication, and siltation are considered major contributory factors in the decline of the species; hence, management of diffuse water pollution from agriculture is a key priority in catchments that host FWPM habitats.

The River Clun is located in central UK and drains a catchment 272 km². The principal land use within the catchment is sheep and cattle grazing although there are notable areas of cultivated land. In the lower reaches, the river is protected as a Special Area of Conservation (SA-C) and Site of Special Scientific Interest (SSSI) due to the presence of one of the few lowland populations of FWPM in the UK. At the time of study, adult FWMPs were reported to be reproducing in the Clun since glochida had been found on fish but juvenile/young adult FWPMs were, however, generally absent in surveys leading to the hypothesis that they were not surviving in the channel bed substrate to emerge as adults. It was hypothesised that siltation was a key factor so sediment fingerprinting tools were applied to identify the domain sources of polluting sediment.

Sampling and analysis

The sediment fingerprinting approach was applied at two levels in the study basin to determine the principal sources of fine sediment present in the failing FWPM beds in the higher-order channel.

Firstly, at the broad spatial scale, the main tributaries (Fig. 7.5) were treated as integrated source end members and a fingerprinting approach applied through analysis of fine sediment captured and stored in the channel at the stream outlets. This was then compared to material stored in the FWPM reaches. Importantly, this framework also included mainstem channel banks as a potential sediment source. Therefore, this approach allowed the identification of the relative contribution from the different tributaries and mainstem bank erosion to the sediment found within the FWPM beds.

Fig. 7.5

Schematic of subcatchment scale experimental design showing link between target FWPM reaches on the main channel and subcatchment areas sampled using stream outlet sediment (providing an integrated source signature)

Secondly, the subcatchment scale phase of the sediment fingerprinting attempted to compare sediment collected from the tributary outlets to primary sources in the wider catchment namely: (i) cultivated soil, (ii) uncultivated soil, (iii) channel bank erosion, (iv) farm tracks, (v) road verges, and (vi) road-transported material. Therefore, this approach allowed the identification of the relative contribution from the different sources to the sediment found at the tributary outlets.

Using these two sets of results allows the identification of the tributaries that are significant sources of sediment to the FWPM beds and, within those tributaries, the identification of the primary sources that are most significant.

Channel-stored sediment and suspended sediment were collected in triplicate from the outlet of each tributary. The stilling well method (Lambert and Walling 1988) was used to recover fine sediment stored on and within the uppermost part of the channel bed matrix that was disturbed by water column agitation. Samples were collected during each quarter of the project to characterise spring, summer, autumn, and winter conditions. Suspended sediment was collected using time-integrated samplers (Phillips et al. 2000). These were installed to collect summer (drier season) and winter (wetter season) sample sets. Sampling at the FWPM beds was restricted due to ecological sensitivity of the area. The suspended sediment traps were installed away from FWPM areas but in the representative reaches.

Samples were freeze-dried, disaggregated and sieved to < 63 µm to ensure source and mixtures were like-for-like comparison. Elemental geochemistry was analysed via WD XRF using a PanAlytical Axios Max instrument (Sect. 7.4.2). Samples were also analysed for FRNs as described in Sect. 7.4.1. Geochemical unmixing for source apportionment was undertaken using the MixSIAR model as described by Blake et al. (2018).

7.5 Results and Discussion

Owing to the ubiquitous nature of the FRN signal across the catchment, these data (Fig. 7.6) provided a broad framework for interpretation of the geochemical unmixing results.

Fig. 7.6

FRN signatures for sources and mixtures samples from both tributaries and the high-order river channel

The FRN signatures of primary source material (i.e., cultivated soil, uncultivated soil, channel bank, and road-transported/derived material) conformed to established knowledge based on previous UK studies (Walling and Woodward 1992) wherein uncultivated soils carried higher activity concentrations of ¹³⁷Cs and ²¹⁰Pb_xs due to the shallow depth profiles of both radionuclides in undisturbed soil (Figs. 7.7 and 7.8). Cultivated soil was a mix of the surface and shallow subsurface soil materials which effectively lowers the activity concentrations of both FRNs. Channel bank material was deficient in both FRNs due to subsurface nature of material, i.e., much of the channel bank is not labelled by these FRNs. There was also a notable activity concentration of ¹³⁷Cs measured in road sweepings with enhanced ²¹⁰Pb activity concentrations  where interaction of pluvial, i.e., surface flow after rainfall, water with fine-grained material offered opportunity for enrichment. In this regard, it is important to also note the elevated ²¹⁰Pb_xs activity concentration of road material which was enhanced compared to surface soil. These observations along with the range of ¹³⁷Cs concentrations of road sweepings ranging from uncultivated to cultivated soil imply that the soil material sampled from roads originated from surface sources but was subsequently enhanced in ²¹⁰Pb_xs scavenged from pluvial road runoff water, i.e., tracer fingerprints developed as a consequence of the delivery pathway (Belmont et al. 2014).

Fig. 7.7

Illustrative examples of tracer boxplots comparing source term range (S1, S2…) to mixture range (M1, M2) where Zn demonstrates non-conservative tracer behaviour

Fig. 7.8

Unmixing model outputs showing most likely source contributions at the subcatchment (left) and catchment (FWPM beds) (right) scales, in summer and winter

Comparison of channel sediment FRN signatures to source materials offers insight into the different nature of main channel sediment compared to tributary sediment. The FRN signals of all material analysed from the main channel reach that hosted the FWMP beds (both upper and lower areas) were suppressed in terms of ¹³⁷Cs and ²¹⁰Pb_xs content and plotted in the range of channel bank material and cultivated material (Fig. 7.8). This suggests channel bank erosion was an important contributor to sediment sampled from the FWPM beds. The tributary sediment also showed an influence of channel bank erosion but the wider spread of FRN signals indicates that in the subcatchments, other sources were also important. Most tributary sediments had a ¹³⁷Cs concentration in the cultivated soils range with some elevated ²¹⁰Pb_xs concentrations although another possible permutation of sources that could lead to this is a combination of pasture and channel bank erosion. There was a general shift towards the lower corner of the FRN bi-plot space (Fig. 7.6) in the autumn/winter set indicating greater subsurface erosion (channel bank) in the more hydrologically active months.

Whichever the main control on surface soil inputs, the FRN data were unequivocal that channel bank erosion was a major sediment source in this system and that the FWPM beds in the main channel of the Clun contained a significant amount of sediment from this source.

Geochemical data (Fig. 7.7) were analysed in the context of broader FRN interpretations. Tracers for application in the mixing model were selected using exploratory data analysis tools (Blake et al. 2018) with elements (examples in Fig. 7.8) that appeared to show non-conservative behaviour (e.g., Zn concentration of suspended sediment falling outside the range of the source materials) being removed from analysis. The FRN inferences of a strong influence of cultivated soil in subcatchments, with a road-sediment signal indicating conveyance path and influence of bank erosion in main channel sediments, were supported by geochemical fingerprinting data (Fig. 7.7). These demonstrated cultivated soil transported by road as a key factor in the most intensively farmed subcatchments (Fig. 7.8 left). While the main channel also showed a strong cultivated source signal in spring–summer, winter signals were dominated by channel bank erosion inputs (Fig. 7.8 right) which is in line with visual evidence of bank collapse on the main channel reaches harbouring FWMPs. While catchment-wide diffuse water pollution from agriculture in the form of sediment input was indicated, and roads provided an efficient transport vector, the localised importance of bank erosion on FWPM habitat was identified as a key factor contributing to population decline. This process though was in turn driven by greater runoff and a flashier hydrological regime linked to runoff generation on both compacted pasture and cultivated land. The results informed management of fine sediment problems relating to both diffuse water pollution and riparian corridor management, e.g., targeted management of channel banks and road crossings.