Testing the Use of ²¹⁰Pb_ex to Study Sediment Connectivity in a Mediterranean Mountain Basin with Badlands

Badland Regolith ²¹⁰Pb_ex Activity Model

A simple mass-balance model defining the ²¹⁰Pb_ex areal activity density at a stable reference site (A_ref, Bq m⁻²) and at the end of a period of time (t, days) reflects the site activity density for the previous day (t − 1), the ²¹⁰Pb_ex activity decay rate (λ = 8·5 10⁻⁵ days⁻¹) and the ²¹⁰Pb_ex depositional input (or deposition flux). Assuming a steady (and essentially wet) deposition flux of fallout lead-210 (Appleby & Oldfield, 1978; Mabit et al., 2014), the ²¹⁰Pb_ex depositional input can be calculated by the product of daily precipitation (P, mm day⁻¹) and mean lead-210 activity per precipitation unit (a, Bq L⁻¹):

(1)

A_ref in equation 1 reaches equilibrium after 4–6 half-lives of lead-210 (about 90–140 years). However, ²¹⁰Pb_ex areal activity density for sites subjected to intense soil erosion, such as badlands, is far from equilibrium (Hancock et al., 2014; Wilkinson et al., 2015). On the assumption that each erosion event removes a layer of soil, and that depth distribution of ²¹⁰Pb_ex mass activity (Bq kg⁻¹) approaches a negative exponential form (Walling, 2003; Benmansour et al., 2013; Porto et al., 2013), the remaining ²¹⁰Pb_ex areal activity density of a site after the effects of an erosion event (A(Er), Bq m⁻²) can be calculated as (Walling et al., 2009):

(2)

Where: Er (kg m⁻²) is the erosion rate of the event, and h₀ is the relaxation mass depth (or ²¹⁰Pb_ex penetration depth) that represents the depth above which 63% (i.e. 1–1/e) of total ²¹⁰Pb_ex is distributed. By substituting A_ref in equation 2, we can represent a daily-step ²¹⁰Pb_ex areal activity density balance model for a badland (regolith) site (A, Bq m⁻²; Figure 2a):

(3)

The loss of ²¹⁰Pb_ex areal activity density by soil erosion in the badland regolith (Al, Bq m⁻²) can be determined by subtracting the activity before and after the erosion event (i.e. difference between equations 1 and 3):

(4)

Modelled ²¹⁰Pb_ex mass activity of the eroded sediments (Sa, Bq kg⁻¹; Figure 2a) can be calculated from the areal activity density loss at the badland regolith (equation 3) divided by the mass eroded per unit area:

(5)

Transit-Time Sediment ²¹⁰Pb_ex Activity Model

An equivalent formulation of the areal activity density ²¹⁰Pb_ex balance model presented in equation 1 can be used to represent the (daily-step) transit variation of the sediment ²¹⁰Pb_ex mass activity (TSa, Bq kg⁻¹; Figure 2b) for sediments deposited in atmospherically exposed sediment stores within the stream network:

(6)

Where: TSa_(t-1) (Bq kg⁻¹) is the mass activity of sediments deposited at time t − 1 (days), λ is the ²¹⁰Pb_ex activity decay rate (8·5 10⁻⁵ days⁻¹), F is the ²¹⁰Pb_ex depositional input (Bq m⁻² day⁻¹) and Bk is the blanket thickness of deposited sediments (kg m⁻²).

Site Simulations of ²¹⁰Pb_ex Activity Dynamics for Badland Surfaces and Transit Sediments

Site parameterization of the lead-210 atmospheric depositional input (or deposition flux) for the Vallcebre area is provided by the analysis of a reference ²¹⁰Pb_ex soil inventory (40 × 40 cm surface, 15-cm depth) in an undisturbed site (i.e. a natural meadow nearby the main badland areas) within the Cal Rodó basin (site location in Figure 1). ²¹⁰Pb_ex mass activity at the soil surface is 62 Bq kg⁻¹ and decreases exponentially with depth (Figure 2c). Depth integration of the soil ²¹⁰Pb_ex activity gives an areal density of 2,449 Bq m⁻² (Figure 2c). Assuming a steady-state balance between ²¹⁰Pb_ex input and decay, the reference soil inventory yields a depositional flux, F, of 0·21 Bq m⁻² day⁻¹ (76 Bq m⁻² y⁻¹ annual flux), which is similar to the mean atmospheric flux of fallout lead-210 for the northwest coast of the Mediterranean basin (71 ± 27 Bq m⁻² y⁻¹; García-Orellana et al., 2006). Taking into consideration that fallout lead-210 flux is essentially controlled by wet deposition (Mabit et al., 2014), ²¹⁰Pb activity per precipitation unit, a, for the Vallcebre area (with a mean annual precipitation of 860 mm) can be calculated as 0·09 Bq L⁻¹.The short fallout exposure duration of badland surface materials due to high erosion rates usually results in low ²¹⁰Pb_ex activity levels that are confined to the upper 1–2 cm of the regolith mantle (Hancock et al., 2014; Wilkinson et al., 2015). We, therefore, assume a small fallout lead-210 penetration for our regolith simulations, with a relaxation mass depth, h₀, of 8·3 kg m⁻², which is equivalent to a 1-cm depth for the average 1·2 g cm⁻³ regolith bulk density of the Vallcebre badland surface materials (Regüés et al., 1995). Detailed explorations of erosion dynamics for the Vallcebre badlands applying 3-year (1990–1993) experimental data from a small (1,240 m²) badland area (“el Carot macro-plot”; Castelltort, 1995) and extended 15-year (1994–2009) simulations using the event-oriented and physically based KINEROS2 model (Martínez-Carreras et al., 2007; Gallart et al., 2013) indicated that the event-scale erosion rate correlates closely (Pearson's R = 0·85; 91 events) with the product of 5-min maximum rainfall intensity and kinetic energy. Accordingly, the above-described regolith and sediment ²¹⁰Pb_ex activity models are fed by using a simple linear equation for calculating event badland erosion in the Vallcebre area (9% normalized root-mean-square error):

(7)

Where: Er (kg m⁻²) is the event-scale badland erosion, I₅ (mm h⁻¹) is the maximum rainfall intensity in 5 min and K_e (J m⁻²) is the storm kinetic energy.

We applied local daily precipitation records (i.e. precipitation amount, storm I₅ and K_e) to simulate ²¹⁰Pb_ex activity for the Vallcebre badland regoliths during 2013 and 2014 in mass-balance equations 3 and 5, with a preceding model warming-up period of 18 months, which resulted in a saw-tooth seasonal pattern of excess lead-210 associated with the temporal patterns of rainfall characteristics (Figure 2d). Badland surface ²¹⁰Pb_ex activity increased from October to May, accumulating the fallout radionuclide flux generated by low-intensity rainfall events from autumn to spring. Conversely, regolith ²¹⁰Pb_ex activity decreased from June to September, due to the erosive effect of high-intensity summer storms, which were particularly active in July and August 2013 (with I₅ and K_e reaching up to 94 mm h⁻¹ and 1,500 J m⁻², respectively). Summer storms caused a periodic reset of the badland regolith inventory, which resulted in maximum ²¹⁰Pb_ex activity levels below or near the lead-210 annual deposition flux for the study site (76 Bq m⁻², Figure 2d). Simulated badland erosion for the modelled period was 14·2 kg m⁻² y⁻¹, with an equivalent badland denudation rate of 1·1 cm y⁻¹ that broadly matches long-term estimations of badland erosion activity for the Vallcebre experimental area (Clotet et al., 1988; Gallart et al., 2013).

Simulation of transit-time influence on sediment ²¹⁰Pb_ex activity by using the mass-balance equation 6 along with the Vallcebre fallout flux calculation (F, 0·21 Bq m⁻² day⁻¹) and three plausible sediment blanket thickness values reported for first-order streams in Pyrenean catchments affected by badlands (Bk, 1, 5 and 10 kg m⁻²; López-Tarazón et al., 2011; Piqué et al., 2014) indicated a general increase in sediment ²¹⁰Pb_ex activity with increased in-stream residence time or age of the sediments (Figure 2e). Sediment ²¹⁰Pb_ex activity increases were particularly important for the first years of stream sediment storage, causing a sharp rise of modelled sediment activity after 2–5 years of in-stream sediment residence of up to an order of magnitude for an initial sediment ²¹⁰Pb_ex concentration of 10 Bq m⁻² and a blanket thickness of 1 kg m⁻². Similarly, empirical estimations of sediment residence time by Wallbrink et al. (2002) for the Brisbane and Logan rivers in Australia reflected a sharp increase in sediment ²¹⁰Pb_ex activity within 2–5 years of in-stream storage, which also suggests that, for simple channels, accumulation of fallout 210-lead in sediments may increase downstream with sediment age.

Formulation of Sediment Connectivity Hypotheses for the Vallcebre Research Catchments

Sediment Sampling and Laboratory Analyses

Sediment sampling took place during 2013 and 2014 at a variety of basin locations along a 1·5 km section of the Torrent del Purgatori, a first-order stream connecting the most significant badland areas of the Vallcebre catchments with the outlets of the Ca l'Isard sub-basin and the Cal Rodó basin (Figure 1a). The locations included the slope toe of a badland site at the head of the stream, downstream places at stream transit sites, and two automated gauging stations at the outlet of Ca l'Isard and Cal Rodó. Sediments were sampled by a range of methods: (i) a surface trap (permeable bag) of sediments at the toe of the badland site (Figure 1a, and c); (ii) direct collection of (2–3 cm depth) sediments from stream-bed deposits at three stream transit sites (at 100–300 m and 1·3 km downstream from the badland site) and the two gauging stations (Figure 1a and d); (iii) five time-integrated sediment samplers (following the design of Phillips et al., 2000) at the same five (stream transit and gauging station) locations (Figure 1a and e); and (iv) two automated (ISCO) sequential water samplers with high-frequency acquisition (as much as one sample every 2 min during high-discharge events) of (1 L) water and suspended sediment samples at the two gauging stations (Figure 1a and f). Because of the transient character of the flow in the streams studied, the time-integrated samplers were installed in nearly dry stream beds (Figure 1e). Consequently, their original design was slightly modified by piercing three (2 mm) air escape holes in the upper part of the samplers to facilitate the filling of the tubes at the beginning of the events. Complementary data for the calculation of catchment-scale sediment yield during the study period were provided by monitoring sediment load at the Ca l'Isard and Cal Rodó gauging stations in continuous by means of automated turbidity (D&A OBS3) and ultrasonic (Bestobell-Mobrey MSM40) suspended sediment sensors (Soler et al., 2012; Gallart et al., 2013).

A total of 50 sediment samples were collected during the study period, encompassing 11 badland sediment samples from the surface trap (taken from April to December 2014), 7 stream-bed sediment samples (taken from April to May 2014), 16 suspended sediment samples from the automated (ISCO) sequential samplers (taken from March 2013 to December 2014, and integrated over consecutive runoff events to reach a minimum size of 50–100 g of dry matter for laboratory analysis) and 17 suspended sediment samples from the time-integrated samplers (taken from June to December 2014). The samples were taken to the laboratory and oven-dried at 65°C. Both sequential and time-integrated sediment samples were recovered by means of sedimentation. Previous field observations of sediment dynamics in the Vallcebre catchments indicated that peak streamflow transports significant amounts of poorly weathered particles of mudstone in suspension, which can be strongly altered by sediment pre-processing in the laboratory (Soler et al., 2012). To preserve the original physical characteristics of the sediments, no sieving or sediment fractioning was applied. Absolute grain-size composition and specific surface area (SSA, m² g⁻¹) for all the samples was measured by a laser-diffraction (Malvern Mastersizer 2000) analyser.

Total ²¹⁰Pb and ²²⁶Ra (from ²¹⁴Bi) activity (Bq kg⁻¹) was measured by gamma spectrometry (gamma-ray emissions at 46·5 and 609·3 keV, respectively) using an iron-shielded, high-purity coaxial germanium detector (Canberra GR5023–7500 SL). Spectra were acquired and analysed by the Canberra GENIE 2000 software. Energy and efficiency calibrations were performed by certified gamma cocktail standards purchased from the Spanish Research Centre for Energy, Environment and Technology (CIEMAT). Counting time applied for gamma spectrometry was 86,400 s for all samples, obtaining a data precision range of 10–20% at the 95% confidence level. Prior to analysis, all samples were sealed in polyethylene bottles for 21 days to achieve equilibrium between ²²⁶Ra and its daughter ²²²Rn (1,622 years and 4 days half-life, respectively). ²¹⁰Pb_ex activity for each sample was determined by subtracting the supported lead-210 in equilibrium with ²²⁶Ra from the total ²¹⁰Pb activity. Measured ²²⁶Ra activity values were corrected (~0·8 factor) for the calculation of in situ ²²⁶Ra-supported ²¹⁰Pb concentrations, based on the ratio of total ²¹⁰Pb to (uncorrected) ²²⁶Ra observed in the lower part (i.e. depth > 10 cm) of our reference ²¹⁰Pb_ex soil inventory, where ²¹⁰Pb_ex was assumed to be absent (cf. Graunstein & Turekian 1986; Wallbrink & Murray 1996; Porto et al., 2013; Mabit et al., 2014). Uncertainty in the determination of ²¹⁰Pb_ex was highly variable, with analytical error ranging from 10% to over 100% of measured activity values. ²¹⁰Pb_ex activity may take negative values for sediment (or soil) samples where total ²¹⁰Pb and ²²⁶Ra-supported ²¹⁰Pb show similar concentrations and are affected by high measurement uncertainty levels (Mabit et al., 2010). Negative ²¹⁰Pb_ex activity measurements in this study were taken as zero for purposes of data analysis.

Data Analysis

We tested for the effects of the sediment sampling method and its relation with particle-size distribution (i.e. median grain size d50, and SSA) on experimental sediment ²¹⁰Pb_ex activity and its associated measurement uncertainties, by applying Kruskal–Wallis ANOVA to the full experimental set of 50 sediment samples. ²¹⁰Pb_ex activity data with analytical error above 100% of measured value were subsequently discarded for the evaluation of the sediment connectivity hypotheses.The temporal pattern of experimental ²¹⁰Pb_ex activity in sediments was evaluated against the simulated (saw-tooth) seasonal dynamics of fallout radionuclide activities for the badland surfaces to ascertain spatial coupling (by pattern conservation of fallout radionuclide concentrations) between the badland regoliths and the sediments in the basin streams. Comparison of radionuclide concentrations between source materials (e.g. soils, regoliths) and sediments generally reflects sediment enrichment ratios due to strong radionuclide binding to the most mobile, fine-grained particles, which requires the application of particle-size corrections for exploring source-stream sediment radionuclide distribution patterns (Collins et al., 1996; He & Walling, 1996; Walling, 2005; Smith & Blake, 2014). The simulated ²¹⁰Pb_ex activities at the badland surfaces were, therefore, particle-size corrected for comparison with the experimental ²¹⁰Pb_ex in sediments using the following conversion equation (He & Walling, 1996; Porto et al., 2003):

(8)

Where: C₀ is the original (simulated) ²¹⁰Pb_ex activity for sediments generated at the badland surface, C_C is the corrected ²¹⁰Pb_ex activity value, S_S is the mean SSA for the experimental sediments applied in the evaluation, S_R is the typical SSA for the badland regoliths and n is a scaling exponent. We used the minimum SSA measured for the 11 sediment samples collected at the badland toe location (surface sediment trap), as a conservative estimate of the particle size for the original regolith materials at the badland surface, S_R. A value of 0·75 was applied to the scaling exponent, n, which is within the range of empirical values reported by He & Walling (1996) for the correction of soil/sediment ²¹⁰Pb_ex measurements.

Temporal (month to year) (dis)connection between the sediment production at the badland surfaces and the mobilization of sediments in the basin streams was assessed by analysing downstream patterns of sediment ²¹⁰Pb_ex activity with a general linear mixed model. The general linear mixed model included (i) the downstream distance for the sediment sampling locations along the Torrent del Purgatori as a continuous fixed predictor of sediment ²¹⁰Pb_ex activity, and (ii) the sampling site as a random factor for analysis across monitored runoff/sediment transport events.

Effects of Sampling Method on Sediment Characteristics and ²¹⁰Pb_ex Activity Patterns

Our results show a strong effect of sampling method on the physical characteristics of sediments. Suspended sediments showed a smaller median (d50) grain size and larger SSA than deposited sediments collected at either the badland site (surface trap) or the stream bed (Figure 3a and b), which reflects general differences in the nature (i.e. suspended versus deposited) of sediments. In addition, suspended sediments collected by the sequential (ISCO) samplers were significantly finer than those collected by the time-integrated samplers (Figure 3a and b), which suggests physical interference of sediment sampling instruments with the sediment characteristics. In fact, at the scale of individual runoff events, sediments collected by the time-integrated samplers at the gauging stations showed d50 grain sizes 42–102% bigger than those collected by the sequential (ISCO) samplers at the same locations. Accordingly, Phillips et al. (2000) demonstrated that the application of time-integrated samplers for collecting suspended sediments may increase their median grain size by a 40–150% range, depending on flow velocity and original particle-size distribution. On the contrary, previous experimental comparisons of suspended sediment concentration and grain size by using automated and simultaneous control samples from the Vallcebre catchments indicated that the original grain-size distributions of suspended sediments were satisfactorily captured by the time-integrated (ISCO) samplers (Soler et al., 2006).

Mean sediment ²¹⁰Pb_ex activity was in general low, ranging from 0 to about 10 Bq kg⁻¹ for the variety of sampling methods (Figure 3c), which reflects the fresh rock origin of the sediments, but also evidences a very strong influence of sampling methods and sediment physical characteristics on radionuclide concentrations. In fact, the finer, suspended sediments that were retrieved by the sequential (ISCO) samplers showed higher radionuclide activities than the coarser suspended sediments retained by the time-integrated samplers. Furthermore, ²¹⁰Pb_ex activity for the coarse-grained, deposited sediments that were collected from the badland sediment bag and the stream-bed locations showed zero and near-zero values (Figure 3c). Measurement uncertainty severely affected the ²¹⁰Pb_ex data. ²¹⁰Pb_ex analytical error was above 100% of the measured value for all data except 16 measurements, all of which corresponded to suspended sediment samples of median grain size ranging from 5 to 20 μm (Figure 3d).

The impact of sediment particle-size distribution on ²¹⁰Pb_ex measurements can be explained by the selective adsorption of lead-210 by the finer, clay-sized mineral particles of soils and sediments (He & Walling, 1996; Mabit et al., 2014; Foucher et al., 2015). A common method designed to mitigate differences in particle-size distribution for radionuclide analysis consists of isolating the <63 μm fraction of the soils and sediments (Walling, 2005). Our results, however, indicate that ²¹⁰Pb_ex concentrations are only detectable for sediments with median grain size below 20 μm (Figure 3d). Other studies have suggested the application of wet fractioning of sediments by particle settling and flocculation to isolate <10 μm grains for radionuclide analysis, particularly in catchments where sediment transport is dominated by clay-sized particles (Olley & Caitcheon, 2000; Hancock et al., 2014; Wilkinson et al., 2015). Nonetheless, wet fractioning of sediments is undesirable in areas with badlands developed over friable and/or soft materials (e.g. marls and mudstone), where particle-size distribution may behave as a dynamic sediment property that can be strongly altered by sediment processing in the laboratory. For example, observations in the Vallcebre catchments and similar highly erosive mountain basins with badlands indicated that large floods typically transport significant proportions of poorly weathered, sand-grained particles of soft bedrock in suspension, which may be further weathered into new fine sediments in a variable time after deposition on the stream bed or by intense sediment manipulation in the laboratory (Mathys et al., 2003; Soler et al., 2008, 2012; Regüés & Nadal-Romero, 2012).

The dynamic nature of badland-originated sediments (i.e. radionuclide-unlabelled, coarse sediments may decay into fine particles by intensive sediment processing in the laboratory) and the strong influence of physical sediment characteristics (i.e. particle-size distribution) on both radionuclide activity and measurement uncertainty patterns stress the importance of selecting efficient methods for fine sediment sampling and investigation in highly erosive catchments with marly and mudstone badlands, where ²¹⁰Pb_ex concentrations are low and strongly influenced by high levels of analytical error. Our experimental results at the Vallcebre catchments suggest that the automatic sequential water samplers provide the best method for collecting representative samples (in terms of particle-size distribution) of fine, suspended sediments for implementing sediment tracing by radionuclide analysis in areas with very active badlands developed over soft bedrock. Additionally, the application of counting times longer than that used in this study (86,400 s) for deriving sediment ²¹⁰Pb_ex concentration estimates from gamma spectrometry or the use of alternative analytical methodologies with higher precision for the detection of low radionuclide activity levels (e.g. alpha spectrometry; Mabit et al., 2008) may help to minimize the uncertainty associated with the sediment ²¹⁰Pb_ex determinations.

Assessing Catchment Sediment Connectivity: Sediment Load and Spatiotemporal Patterns of Suspended Sediment ²¹⁰Pb_ex Activity

Suspended sediment load during the study period totalled 1,370 Mg for Ca l'Isard and 1,875 Mg for Cal Rodó, representing 0·52 and 0·23 kg m⁻² y⁻¹ sediment yield at the sub-basin and basin scales, respectively. These figures are about 50% below the long-term mean sediment yield records for the Vallcebre catchments (0·9 and 0·5 kg m⁻² y⁻¹ for Ca l'Isard and Cal Rodó, respectively; Nord et al., 2010; Gallart et al., 2013) and contrast considerably with the deep badland denudation estimate obtained in our simulation results for 2013 and 2014 (1·1 cm y⁻¹). This variation in erosive performance between the catchment and badland sites for the study period can be explained by the local dynamics of precipitation, which suggests low sediment transport capacity at the sub-basin and basin scales in 2013 and 2014. In fact, 2013–2014 precipitation (910 mm y⁻¹) was 6% above the historical average records, with regular activity of high-intensity summer storms, which strongly control badland erosion rates in the study area (Castelltort, 1995; Martinez-Carreras et al. 2007; Gallart et al., 2013). The long-lasting autumn precipitations that regulate suspended sediment load in the Vallcebre streams (Soler et al., 2008; Gallart et al., 2013) were, in turn, 40% below the mean historical rainfall amount, resulting in limited sediment transport at the broader catchment scale. Accordingly, downscaling of the 2013–2014 sediment yield records obtained at the Ca l'Isard and Cal Rodó gauging stations over the total badland extension for the Vallcebre catchments (11·7 ha) results in a figure 30–40% below the simulated badland erosion rate for the study period. These results suggest that significant amounts of fresh badland sediments remained stored within the stream network during the study period, which was also confirmed by direct observation of stream-bed deposits in the field.

The analysis of spatiotemporal patterns of sediment fallout radionuclide concentrations can facilitate the assessment of active connections and transit-time dynamics of sediments between parts of the catchment (Wallbrink et al., 2002; Walling, 2005; Belmont et al., 2014; Wethered et al., 2015). The group of experimental sediment samples unaffected by severe ²¹⁰Pb_ex measurement uncertainty (i.e. suspended sediments with homogeneous 5–20 μm median grain size) showed ²¹⁰Pb_ex activity levels ranging from 7 to about 25 Bq kg⁻¹ during the study period, which overall matches simulated radionuclide concentrations (Figure 4a). More importantly, the temporal variations in experimental ²¹⁰Pb_ex sediment activity reproduced the simulated temporal pattern of fallout lead-210 for sediments generated at the badland surfaces. In fact, ²¹⁰Pb_ex activity in the experimental sediments tracked the characteristic (saw-tooth) seasonal pattern of fallout radionuclide concentrations modelled for the badland regoliths, which predicted a build-up of ²¹⁰Pb_ex activity at the badland regoliths due to low-intensity precipitation events in autumn and spring, and a subsequent ²¹⁰Pb_ex depletion caused by high-intensity showers in summer (Figures 2d and 4a). As expected, the drop in measured suspended sediment ²¹⁰Pb_ex activity levels was particularly pronounced during July and August 2013, when strong summer storms with intensities up to 94 mm h⁻¹ in 5 min caused deep erosion at the badland sites. The close similarity of both the simulated and the experimental sediment radionuclide activity patterns discards any significant contributions of sediments from other surface cover types, such as the very stable pastures and forests of the Vallcebre catchments, where soil ²¹⁰Pb_ex patterns can be assumed to be in equilibrium at activity levels of about 60 Bq kg⁻¹ at the soil surface (Figure 2c). These results agree with previous sediment budget investigations in the Vallcebre catchments, which revealed the badland surfaces as the ruling source to basin sediment load (Gallart et al., 2002, 2013; Soler et al., 2008).

The observed conservation of the modelled seasonal arrangement of badland regolith ²¹⁰Pb_ex concentrations in the radionuclide signatures of collected suspended sediment samples (Figure 4a) indicates high fine sediment connectivity between the badland sites, the streams and the outlet of the Vallcebre catchments. Similarly, a broad range of sediment studies in small (i.e. less than 50 km²) Mediterranean mountain basins across the Pyrenees and Central Alps have highlighted a high connectivity and fast stream transmission of fine sediments originating in highly dynamic humid badlands, especially for areas where these erosive landforms are structurally integrated in the basin stream network (Descroix & Mathys, 2003; Francke et al., 2008; Nadal-Romero et al., 2008; Duvert et al., 2012; Gallart et al., 2013). At the larger regional scale, the high mobility of fine sediments generated in these badland landscapes can result in the delivery of large amounts of suspended sediments through the drainage networks. For example, sediment budget investigations and spectral fingerprinting analysis of suspended sediment transport for the Isabena basin, a central Pyrenean 445-km² basin affected by (about 1% of terrain) badlands, reflected a yearly average input of about 260,000 t y⁻¹ badland sediments and a 60–80% sediment delivery ratio (López-Tarazón et al., 2012; Brosinsky et al., 2014). Similarly, for the mesoscale regional context of the Vallcebre catchments (i.e. the upper Llobregat basin; 500 km²) where badlands represent ~0·5% of the terrain, integrated sediment budget and GIS analysis estimated badland sediment contribution to represent up to 30% of the regional flux of suspended sediments (Clotet et al. 1984; Moreno-de las Heras et al. 2016).

Unlike the downstream sediment ²¹⁰Pb_ex concentration increase with sediment age suggested by our transit-time sediment simulations (Figure 2e), fallout lead-210 activity levels for the experimental suspended sediments decreased downstream between the sampling locations of the Vallcebre catchments (Figure 4b). A variety of studies have explained downstream decreases in the ²¹⁰Pb_ex concentrations of suspended sediments by the contribution of unlabelled fresh materials from (nearly vertical) streambanks showing nil or little fallout lead-210 interception (Belmont et al., 2014; Hancock et al., 2014; Wethered et al., 2015). Channel entrenchment in the Vallcebre catchments is strongly conditioned by the occurrence of infrequent, strong autumn events with an approx. 5-year return period (Gallart et al., 2002, 2013). Nonetheless, the observed low sediment transport conditions for 2013 and 2014 did not support streambank erosive activity during the study period. More likely, flushing of the most dynamic, fine sediments from the stream bed by flooding may have prevented ²¹⁰Pb_ex accumulation in temporary sediment stores within the stream network. Decay of poorly weathered sand-grained particles of clayey bedrock stored in stream-bed deposits into fine particles may also add radionuclide-unlabelled suspended sediments to the streamflow (Soler et al., 2012), contributing to the observed downstream reduction in sediment ²¹⁰Pb_ex activity. These results must, however, be explored with caution due to the scarcity of events with paired observations between sediment sampling locations with reliable sediment ²¹⁰Pb_ex measurements in our data. More information is required, optimally for a wide variety of streamflow conditions (e.g. low, regular and extreme runoff events), to disentangle the complex interactions that control the downstream dynamics of sediment ²¹⁰Pb_ex concentration. In the light provided by our results, we hypothesize that high streamflow conditions that dramatically activate the mobilization of long-term sediment stores and general channel entrenchment in the drainage network (Gallart & Clotet, 1988; Gallart et al., 2013; Serrano-Muela et al., 2015) will exacerbate the downstream obliteration of the suspended sediment ²¹⁰Pb_ex signatures.

Process-based understanding of sediment connectivity (i.e. the continuity of sediment transfer between different zones of a catchment; Bracken et al., 2015) is required for effective basin management, particularly in Mediterranean mountain basins, where badland occurrence is a common phenomenon that can have important consequences for river dynamics, freshwater ecosystems and the security of water resources. Despite very intense discussion in a series of recent conceptual studies (e.g. Bracken & Croke, 2007; Lexartza-Artza & Wainwright, 2011; Wainwright et al., 2011; Bracken et al., 2015; Okin et al. 2015), there is little consensus about how to quantify sediment connectivity. A broad range of studies carried out in Mediterranean mountain basins with badlands have applied the sediment budget approach, which rely on the measurement, extrapolation and accumulation of sediment production/transport in different parts of the catchment to infer sediment transfer (Clotet, 1984; Llorens et al., 1997; Nadal-Romero et al., 2008; López-Tarazón et al., 2012; Gallart et al., 2013). In this context, the use of environmental tracers, such as the fallout component of the radioactive isotope lead-210, has provided alternative methodologies for inferring sediment transfer in a variety of environments (e.g. Walling, 2005; Porto et al., 2013; Wilkinson et al., 2015; Estrany et al., 2016). Radioisotope sediment studies are, however, notably lacking in badland environments, where their highly erosive nature is frequently perceived as a major limitation for ²¹⁰Pb_ex analysis (Mabit et al., 2008; Dercon et al., 2012). Differently, our study indicates that the highly dynamic nature of mountain Mediterranean badlands provides an opportunity for the assessment of sediment connectivity on the basis of fallout lead-210 activity patterns. In fact, badland surface ²¹⁰Pb_ex activity is characterized by a seasonal pattern (i.e. progressive accumulation in autumn/winter followed by a sharp summer reduction of fallout levels) that can be used to track the continuity of sediment transfer between different parts of the catchment by spatiotemporal analysis of sediment ²¹⁰Pb_ex concentration. Overall, our study reveals sediment ²¹⁰Pb_ex analysis as a powerful tool that can complement more traditional sediment load monitoring programs for the study of fine sediment connectivity in Mediterranean basins affected by the activity of highly erosive landforms.