A comparison of 210Pbxs, 137Cs, and Pu isotopes as proxies of soil redistribution in South Spain under severe erosion conditions

The use of fallout radionuclides as proxies for the study of soil redistribution processes in semiarid environments of Southern Spain has been restricted to 137Cs. The potential and limitations of alternative proxies such as 239+240Pu and 210Pbxs should be explored given the expected constrains imposed by the features of the study area. Four soil and one sediment cores were collected in a highly eroded area to test the feasibility of these proxies under demanding conditions. The use of gamma spectrometry and ICP-MS did allow exploring the potential of 210Pbxs and Pu isotopes against the well-established tracer, 137Cs. The activity ratios 239+240Pu/137Cs were explored to evaluate the previous evolution of the sampling sites. Soil redistribution rates were estimated using the model MODERN. Despite all the profiles showed intense perturbation, Pu isotopes showed the highest potential thanks to higher sensitivity and sample throughput. The deviations of 239+240Pu/137Cs inventory ratios (0.012–0.158) from the global fallout average (0.026 ± 0.003) suggest that the sediment core was a deposition site involving alternate episodes of topsoil removal and incorporation from different sources. The calculated erosion rates ranged 34–43 t ha−1 year−1, being in good agreement for 137Cs and 239+240Pu. 210Pbxs was not used due to low-quality data. The use of Pu as a tracer of soil redistribution processes in semiarid areas seems to be promising even under severe erosion conditions. The use of 210Pbxs is not recommended in this area due to its low concentration.


Introduction
In the EU countries of the Mediterranean basin, more than 200,000 km 2 of land is under high risk of soil erosion (Griesbach et al. 1997). Andalusia (South Spain) is one of the most vulnerable areas to water erosion (De La Rosa et al. 1996;García-Ruiz 2010) due to a combination of particularly low vegetation cover, land use types, and high rainfall erosivity (Kosmas et al. 1997;Koulouri and Chr 2007;Panagos et al. 2015) and soil erodibility (Panagos et al. 2014) together with a varied history of land management (García- Ruiz et al. 2013). These factors together with an increasing agriculture intensification produce important both on-site, e.g., reduction of fertile soil, and off-site impacts, e.g., eutrophication of water bodies (Montgomery 2007).
Determining past soil erosion rates over time scales that encompass at least several decades can provide a better insight into long-term trends in soil erosion impacts. Moreover, reconstructed soil erosion rates at appropriate temporal and spatial scales can be used to validate/evaluate soil erosion models (Lang and Bork 2006) which can then be used to predict future impacts and to define and evaluate Responsible editor: Shiming Ding mitigation strategies. In Andalusia, several experimental studies based on yearly monitoring have determined soil loss rates at the plot scale (Francia Martínez et al. 2006;Gómez et al. 2009aGómez et al. , b, 2018 and catchment scale (Taguas et al. 2009(Taguas et al. , 2013Gómez et al. 2014). However, these records do not exceed a decade in duration, being a notable exception the study carried out by Mabit et al. (2012) who used fallout radionuclides (FRN), in particular 137 Cs, to estimate the average soil loss rates since the 1950s.
FRN provide spatially distributed information on soil and sediment redistribution (Dercon et al. 2012). This proxy assumes that these radionuclides are supplied either by dry or wet supply via local or global fallout. Then they strongly bind to surface soil. In this way, their redistribution across a landscape is mainly driven by the same processes responsible for the soil redistribution. The proxy is applied in relative terms; hence, the FRN inventories at the sampled site are compared to those of reference sites. This allows estimating redistribution rates in terms of eroded or deposited mass per surface unit and time unit. One of the main advantages of this approach is that longterm monitoring of experimental fields is not required, resulting in relatively high benefit/cost ratios (Dercon et al. 2012;Allewell et al. 2017).
Among the different FRN, 210 Pb (T 1/2 = 22.3 years) is a naturally occurring nuclide. Its parent nuclide, 222 Rn, exhales into the atmosphere, where it decays to 210 Pb, and it is then deposited in the soil as fallout in excess, 210 Pb xs . 210 Pb xs can provide information about processes occurring within a time-lapse up to 150 years ). On the other hand, 137 Cs (T 1/2 = 30.1 years), 239 Pu (T 1/2 = 24,110 years), and 240 Pu (T 1/2 = 6563 years) are artificial nuclides that were introduced into the environment during the Cold War era, so they may provide information regarding processes lasting up to 55-60 years .
The use of FRN in Andalusia has been extremely limited and mainly based on the use of 137 Cs Ballais et al. 2013). However, the use of 137 Cs as a proxy for soil redistribution is hampered by biogeochemical constrains, rapidly decreasing environmental concentrations due to its short half-life and the accidents of Chernobyl and Fukushima nuclear power plants that injected additional 137 Cs signals (see, e.g., Foster 2011, 2013;Mabit et al. 2013). Additionally, 137 Cs and 210 Pb xs are usually measured by gamma spectrometry, which is characterized by having a relative high detection limit and low sample throughput (for example, in the conditions provided in Sect. 2.4, 0.5-1 sample per day and detector).
In last years, Pu isotopes have been subject of increasing interest because of their long half-life and the lack of tropospheric sources other than global fallout. The use of mass spectrometric techniques for the measurement of Pu isotopes is a good alternative to radiometric techniques as they can provide high sensitivity and high sample throughput. To this end, techniques such as accelerator mass spectrometry (AMS) and sector focusing inductively coupled plasma mass spectrometry (SF-ICP-MS) have been used (Ketterer and Szechenyi 2008;Tims et al. 2010;Sanders et al. 2016). However, AMS and SF-ICP-MS have very high operational costs, and consequently, they are not an affordable alternative for many researchers. To make it more affordable, the analysis of Pu isotopes has been extended to conventional (quadrupole based) ICP-MS instruments coupled with high efficiency sample introduction systems (Ketterer et al. 2002;Xu et al. 2015).
As far as we know, neither 210 Pb xs nor Pu isotopes have been evaluated in Andalusia for the analysis of soil erosion/ redistribution processes. There are several risks associated to such use from a practical point of view. 238 U concentration in natural soils of this region is relatively low in comparison with other regions of Europe (López-Coto et al. 2013); thus 210 Pb concentrations are consequently low. Furthermore, the high porosity, scarce soil moisture, and the low frequency of precipitation events suggest that the degree of disequilibrium between 210 Pb and its grandparent 226 Ra could be small in soils, leading to extremely small concentrations of 210 Pb xs . It also suffers from high uncertainty of measurement results at low activity concentrations (Iurian et al. 2016). In the case of Pu isotopes, as the fallout supply of Pu decreases with latitude, it makes its quantification more difficult in this region than in Northern Europe. Moreover, Pu seems to show a higher binding preference for organic matter in soil (Meusburger et al. 2018), which contents are relatively small in Andalusia. According to Allewell et al. (2017), Pu isotopes have a high potential to become one of the main soil radiotracers; however, more studies are needed to evaluate its suitability under different environmental conditions.
The main goal of this work is to provide background information on the potential suitability of 210 Pb xs and 239+240 Pu for the study of soil and sediment redistribution in regions whose characteristics establish a priori concerns about their feasibility. The suitability of these techniques was tested by comparing them to other classical indicators of redistribution of soil and sediments such as 137 Cs. The main research questions are (1) how detectable are the selected nuclides, (2) what is the order of magnitude of their vertical migration, and (3) how suitable are these nuclides to study soil and sediment redistribution. For this purpose, we have selected a study area with particularly high erosion rates, so the limitations of the techniques can be established under especially demanding conditions. This work would be the first published attempt to use inventories of 210 Pb xs and Pu isotopes to calculate redistribution rates in this region.

Study area
The sampling zone was located inside and in the surroundings of the Hornachuelos Natural Park (Fig. 1). The park is located southeast of the Iberian Pyrite Belt and covers an extension of ~ 60,000 ha. The geological settings are mainly characterized by sedimentary and resistant metamorphic rocks with a high degree of metamorphism, mainly consistent in meridional gray-brown soils and rankers developed upon slates, shales, quartzite, and, in general, Paleozoic metamorphic rocks (Moreira 1995). This includes bluish gray loams and slates typical of the South of Sierra Morena mountain range. Dominant soil taxonomy has been characterized as Lythic Xerochrepts (Rosa et al. 1984). Climate has been classified as subhumid Mediterranean, with average temperatures of 19 °C and precipitations within the range of 500-800 mm year −1 , mainly concentrated between September and May (Ministerio de Agricultura, Pesca y Alimentación 2008).
Soils are characterized by a scarce development of the edaphic horizons, a low concentration of fertilizer elements, and a strong acidity (Moreira 1995). Farmland is mainly extensive with orchard and cereal production. Vegetation is dominated by typical species of Mediterranean sclerophyllous forest such as holm oak (Quercus Ilex), cork oak (Q. suber), gall oak (Q. faginea), and wild olive tree (Olea europaea) (Hidalgo-Fernández et al. 2014). In flatter areas, meadows such as shrubs are dominant. Trees and cereal production is mainly used to feeding cattle and, more recently, animal species destined for big game hunting. At present, more than half the surface of the park is comprised of hunting grounds, leading to certain works that have altered the original ground (installation of fences, reforestation, etc.) (Moreira 1995).

Soil sampling
Four soil cores and a reservoir sediment core were collected in the study area ( Fig. 1, Table 1). Soil core S1 was collected at the base of a hill close to the county road. The collecting Fig. 1 Location of sampling sites point showed signs of soil accumulation from the upper part of the hill; however it was located near a sharp step, suggesting that both soil erosion and accumulation were possible. S2 was collected at the top of the same hill with no evidences of water stream intrusions, on an a priori basis. Core S3 was collected in a mid-elevation plain of a park near the town of Posadas. The sampling area did show signs of past soil public works such as flattening and removal of the vegetable cover; hence, intense soil mixing should be expected. On the other hand, its location suggested mid-moderate soil erosion. Core S4 was collected in the highest flat top peak of a recreational park close to La Breña reservoir. According to the orography and the presence of vegetation, scarce or null soil removal should be expected. Finally, sediment core Sed-1 was collected in an emerged plain inside the reservoir. The reservoir collects water from Guadiato and Guadalquivir rivers that are being used for irrigation during the dry season. This fact leads to large variations of the accumulated water volume during a year. The steep hills surrounding the main water mass, the presence of 20-30-cm-thick gullies crossing the land hill down to the water body, and the presence of emerged soil masses suggested the potential accumulation of material driven by the runoff during the rain events; hence, this sample site could be considered a sedimentation emplacement on an a priori basis. Sediments act as sinks of radionuclides; hence, what we could expect for this sample is a higher radionuclide inventory than those found in the soil samples. It is worth noting that during the sampling and the previous exploratory visit to the area, many sites having steeped hills and elevations were surrounded by fences keeping private hunting grounds. Under these conditions, access to many potential sampling sites was impossible, meaning that the sampled sites were not the best possible sampling site candidates but a commitment solution. Al/Fe hollow tubes, 3.5 cm i.d., and 35-45 cm long were used for sampling. The tubes were slowly unscrewed out with a corner pipe wrench in order to avoid loses from the deeper side, and then they were capped and prepared for transportation to the laboratory. The exploratory nature of this work prevents a thorough application of these techniques on a selected catchment and an adequate application of well-established techniques to select and properly validate reference sites, such as the CHeSs approach (Arata et al. 2017).

Sample pretreatment and texture analysis
The tubes were sliced into 3-4 cm thickness layers with a reciprocal saw, and the individual samples were oven-dried at 65 °C until constant weight. Thereafter, metallic pieces were removed using a magnet, the stones roots (if any) and man-made tools were removed by hand, and the samples were crushed using a ball mill, with subsequent sifting by a 0.7-mm mesh sieve. Depending on the soil moisture and compaction state of the soil layer, this method allowed us to collect 10-45 g per slice. It is worth indicating that during cores handling at lab, several anomalies appeared such as large holes in the core S2 (apparently animal tunneling, 6-11 cm depth) or finding plastic tools (S3, 5-11 cm depth). That led to a minimum soil recovery for certain soil slices. For those slices, no sample enough was available for gamma spectrometry or Pu isotope analyses.
The sand, silt, and clay contents of soil samples were determined using the Bouyoucos hydrometer method (Gee and Or 2002). First, the soil was oven-dried at 105 °C for 1 day and passed through a 2-mm sieve. Then, < 40 g (or a suitable mass) of soil is taken and mixed with 100 mL of sodium hexametaphosphate solution, for soil aggregate dispersion. After 10 min of soaking the mixture, it is mixed in the dispersion vessel for 5 min. Afterward, the mixture is put into a 1000 mL sedimentation cylinder, and distilled water is added to the 1000 mL line. The suspension is stirred thoroughly by moving it up and down with a brass mixing rod. The timer is started when the mixing rod is removed from the cylinder. Simultaneously, the hydrometer is gently immersed in the suspension, and 30 s later, the hydrometer reading is recorded. After 60 s, a second reading is made. Subsequent readings are taken after 5, 30, and 90 min, and the last reading is taken after more than 8 h. Before the hydrometer was used, a blank reading, including temperature, was performed to correct hydrometer readings. It consists of a cylinder with 100 ml of sodium hexametaphosphate solution added to the 1000 mL line with distilled water. Calculations for sand, silt, and clay percentages, based on the readings at each time interval, were carried out according to Gee and Or (2002). 40 K, 137 Cs, 210 Pb, 212 Pb ( 228 Th), 214 Pb ( 226 Ra), and 234 Th ( 238 U) were analyzed at CITIUS (Center for Research, Technology and Innovation of the University of Sevilla, Centro de Investigación, Tecnología e Innovación de la Universidad de Sevilla) by gamma spectrometry using their gamma emissions of 1461, 662, 46.5, 238, 295/352, and 63.3 keV, respectively (Hurtado-Bermúdez et al. 2019). Firstly, the sieved samples were sealed into 80-mL cylindrical beakers and were stored for at least 4 weeks; in this way, 226 Ra and their daughters 214 Pb and 214 Bi reach the radioactive equilibrium condition. Thereafter, the samples were measured in a low-background Canberra HPGe GR-6022 detector (60% relative efficiency) inside a Canberra 777A ultra-low-background lead shield of 15-cm-thick high-purity lead passive shield and a graded liner consisting of low-background tin with a thickness of 1 mm and high-purity copper with a thickness of 1.5 mm. The spectra were analyzed using Canberra Genie 2000 gamma software v3.2. Uncertainties are provided using k = 1. Full-energy peak efficiencies (FEPE) were calculated for different energies and an empty beaker using Canberra LabSOCS (Laboratory Sourceless Calibration Software) software. Afterwards, selfattenuation correction factors were calculated using a transmission experiment with non-collimated point sources. The combination of both quantities gave us corrected FEPE and consequently sample activity for any isotope (Hurtado and Villa 2010;Hurtado et al. 2007). The analyses were validated through successful participation in several Monte Carlo intercomparison exercises (Vidmar et al. 2008;Lépy et al. 2012) and also internally through the measurement of several reference materials, IAEA-RGU-1, IAEA-RGTh-1, IAEA-RGK-1, IAEA-444, and IAEA-447. The limits of detection are in the range of 25 Bq kg −1 ( 40 K), 1.2 Bq kg −1 ( 137 Cs), 8.5 Bq kg −1 ( 210 Pb), 3.1 ( 212 Pb), 6.4 ( 234 Th), and 3.5 Bq kg −1 ( 226 Ra) (using sample at 22.5-25.5 cm of S1 core). 210 Pb xs activity concentration was calculated by subtracting to overall 210 Pb activity concentration that of 226 Ra (hence, it was assumed that 226 Ra and supported 210 Pb remain in radioactive equilibrium). Being this approach is quite conventional, certain concerns have been put on it as the radon emanation coefficient is not taken into account. However, the experimental determination of that coefficient is not easy, and the correction is expected to be minimum in soils ; hence, the correction by 226 Ra activity concentration has been used as a commitment approach. When required, statistical tests for homogeneity and variance have been performed using Statgraphics Centurion 18. All the tests have been performed at the 95% confidence level (α = 0.05).

240 Pu analyses
Pu isotopes were analyzed by ICP-MS after applying a methodology described by Ketterer et al. (2002). Briefly, 10-40 g of each sample was calcined at 550 °C for 12-24 h, and Pu was subsequently leached from the ashes using 16 M HNO 3 (J.T. Beaker™) for 16 h at 80 °C after the addition of ~ 7 pg of 242 Pu for the sake of the application of the isotope dilution technique. After filtration and dilution of the filtrate to 8 M HNO 3 , Pu was reduced to + IV oxidation state by sequentially adding FeSO 4 , NaNO 2 , shaking, and warming at 75 °C for 2 h. Thereafter, 150 mg of the Triskem TEVA extraction chromatography was added to each sample, and shaking was applied for 1 h. The resins were collected in 23-mL LDPE pipettes fitted with a glass wool plug. The resin was sequentially rinsed with 25 mL 2 M HNO 3 , 15 mL 8 M HCl, and 10 mL 2 M HNO 3 to remove mass interfering elements such as uranium. Pu was extracted using three sequential elution steps with deionized water, 0.05 M ammonia oxalate, and deionized water again (0.4 mL each one). Under these conditions, the limits of detection for 239+240 Pu were in the range of 5-10 mBq kg −1 .
One out of each five samples was used as quality control samples consisting of replicate sample preparation and measurements, blank samples (sandstone samples isolated from radioactive fallout), and the analysis of several aliquots of CRM samples (IAEA-384). Pu contents in this material are quite much higher than the Pu concentrations we expected in the fallout-level samples; hence, every IAEA-384 aliquot was diluted into sandstone (1:350 m:m). The samples were measured at Northern Arizona University (NAU) using a quadrupole ICP-MS (Thermo X2) coupled to an ESI Apex-HF sample introduction system and a 10-roller peristaltic pump (Gilson). Additionally, several of the quality control replicates were prepared and measured at CITIUS using an Agilent 8800 ICP-MS/MS coupled to a CETAC ultrasonic nebulizer. All the quality control samples resulted in values within the expected ones excepting one replicate of the IAEA-384 CRM. This fact was associated to a low chemical yield of plutonium in a sub-batch of samples, meaning that samples with very low chemical yields have not been reported but repeated when possible.

Use of the 239+240 Pu/ 137 Cs activity ratios and soil mobilization rates
For the cases where both 137 Cs and plutonium isotopes were detectable, we have tested the information that could be provided by 239+240 Pu/ 137 Cs activity ratios about the soil evolution. Experimental data (Legarda et al. 2011) showed that the contribution of Chernobyl accident to the 137 Cs activity concentrations in soil columns is detectable, but its contribution to the inventories at southern Spanish mainland soils can be neglected against the contribution from global fallout. Regarding plutonium, different evidences showed that atmospheric input due to the Chernobyl accident was minimum and at a very local scale . Consequently, the main sources of these nuclides at this latitude are Globall Fallout, and thus, it can be assumed that both radionuclides were supplied at a same pace. A value of activities ratio of 0.026 ± 0.003 has been proposed for global fallout (Hodge et al. 1996) as corrected to July 1998 for the sake of comparison to available data , although the different biogeochemistries of these elements can result in deviations from that value as explained below.
The total inventories of Pu and Cs isotopes should be similar to that global fallout value (on what follows, GFV) once provided the soil were undisturbed. We argue, however, that the activity ratios can deviate dramatically from that reference value taking into account that these elements show different biogeochemistry, different migration behaviors, and given the effect of partial soil redistribution. Pu is preferentially associated with colloidal and higher molecular weight materials . The deeper migration ability of Pu with respect to Cs has been already shown for Mediterranean soils (Guillén et al. 2015). Therefore, we propose that the activity ratios 239+240 Pu/ 137 Cs cannot be homogeneous through the core profile, although the activity inventory ratios could agree with the GFV for an hypothetical unperturbed soil core.
On the contrary, we would expect that activity ratios show different values deviating from the GFV as a function of depth: (1) the deeper migration ability of Pu regarding Cs would deviate the activity ratios in the deeper part of a core to values higher than the GFV. (2) For upper parts of the core, we could expect by contrast an activity ratio less than GFV. That is exactly what can be derived, for example, from the data of (Guillén et al. 2015).
Subsequently, erosion would remove soil masses (topsoil) with activity ratios less than GFV, producing values higher than GFV in the remaining inventory. Once the eroded soil mass accumulates on top of a previously unperturbed soil core, that location would result in inventory ratios below the GFV (see Graphical abstract, a). Should the same eroded soil mass accumulate on top of weakly eroded soils, the resulting inventories should result in values similar to the GFV (Graphical abstract, b). The studied park is characterized by dramatic changes of the soil uses, meaning that obstacles between a soil source (erosion) and a soil sink (deposition) can appear and disappear with time. And under these conditions, soil sites suffering first weak erosion (resulting inventory ratios ≥ GFV) and then soil deposition supplied from strongly eroded sites (inventory ratios > GFV) would result in inventory ratios > GFV; i.e., the results at that site should be the result of erosion and deposition episodes.
We will compare this argument with the results provided from erosion and deposition rates as shown below. To do that, MODERN model was used. This model and its potential have been widely described in the literature, as can be seen, for example, in Arata et al. (2016a, b). This model estimates erosion or deposition rates based on the comparison (in the original reference, "alignment") of the total FRN inventory at the sampling site and its depth profile at reference site; in this way, the model returns a solution as a thickness of the soil layer affected by erosion or deposition. The main assumption is that the depth distribution of the selected FRN is the same at the reference and the sampling sites, as it could be expected for any situation where FRNbased models could be applied. Among the main features of MODERN are that its application does not require a transect sampling approach, and, additionally, it does not make any assumption on the shape of the radionuclides profile. The model is available under request to University of Bäsel.

Soil cores
The activity concentrations obtained in the soil samples for 40 K, 226 Ra, and 210 Pb are shown in Figs. 2, 3, and 4, respectively. The data about the erosion proxies 210 Pb xs , 137 Cs, and 239+240 Pu are shown in Table 2. Additional information (in detail values of 210 Pb, 226 Ra, 212 Pb, and 234 Th activity concentrations) are provided as supplementary material (Table S1).
The analyzed radionuclides felt below their corresponding limits of detection in a large proportion of the samples: 137 Cs, 210 Pb xs, and 239+240 Pu were undetectable in 69%, 62%, and 54% of the soil and sediment analyzed samples, respectively. This finding is not surprising as the  (Gascó et al. 2006) or Mediterranean forest soils (Guillén et al. 2015). These facts suggest that the lack of detection is related to intense soil mixing and removal processes.

Natural radionuclides
40 K has a relatively constant isotope abundance of potassium in nature (~ 0.012%), and it forms part of the rock matrix as feldspar and muscovite. However, this element is also supplied with fertilizers; it can be dissolved in environmental water and can be taken up by plant roots through ion channels and specific carriers (Schlesinger 2021). What we could expect is that 40 K activity concentrations were nearly constant below the O-horizon and that its concentration varies between sampling points according to the soil composition.
In general, the order of magnitude of the ranges of activity concentration found in the four soil cores (142-1142 Bq kg −1 ) was in good agreement with data collected at similar latitudes in Spain (Vaca et al. 2001), with the median and ranges proposed for Spain (UNSCEAR 2000) and with the values that could be calculated from the potassium element concentrations published in FOREGS geochemical database for this region (Salminen 2006). These ranges are, furthermore, below the European median. However, several interesting features appeared. The range of variations with depth found in cores S2-S4 (20-48%) was noticeably higher than of S1 (12%) and much higher than those found in other studies in Southern Spain . The 40 K depletion found at the top of the hill could indicate that upper soil layers could have been removed, or, alternatively, that the base of the hill accumulated a part of the potassium supplied either by dead leafs from vegetation or by soil masses supplied from some nearby locations. The fact that 40 K activity concentrations in the mid-lower part of core S1 were similar to those of the top of the hill suggests that this last hypothesis could be a plausible one. The fact that the maximum activity concentrations found in core S3 was more than twice those collected in cores S1 and S2 is not surprising as it is reflecting the heterogeneities of the sampling area. On the other hand, the small activity concentration found in the middle of the S4 core cannot be explained by an experimental artifact, so we argue that this layer contained a different soil matrix. This hypothesis is supported by the lack of detection of other natural radionuclides ( 212 Pb and 234 Th) and by certain changes in the sand:silt:clay ratio of the soil (Table 1).
Supplementary data provided in Table S1 shows that the concentrations of 212 Pb ( 228 Th) and 234 Th ( 238 U) are within similar order of magnitudes to median background values calculated for Spanish soils (UNSCEAR 2000) and below world-scale medians. For every soil core, the activity concentrations are within the same order of magnitude inside the uncertainty intervals, although certain heterogeneities of the representative values lead to ranges of variation of 14-50% for 212 Pb and 21-82% for 234 Th. This internal variability suggests that the soil cores reflect certain mixing between different soil masses. This fact could be associated to natural processes such as runoff and soil deposition as expected for core S1, but it is also compatible with the addition of successive additions of soil layers from different sources with intense compaction as could be expected for core S3. 210 Pb was detectable in almost all samples, being activity concentrations within the range 8-77 Bq kg −1 (Fig. 4). Excepting a few values in core S2, this range agrees with values previously published for surface unpolluted soils at similar latitudes of Southern Spain (Bolívar et al. 1995;Gascó et al. 2006), in the first case assuming that both 210 Pb and 210 Po are in equilibrium in the soil system. 226 Ra (Fig. 3) was also detectable in almost all the samples, and For every soil core, the range of variation of 226 Ra activity concentrations was 8, 19, 28, and 33% for cores S1 to S4, respectively. The internal variability found for cores S2-S4 seems to be incompatible with unperturbed profiles showing nearly constant activity concentration with depth, like those shown in other parts of Spain; see, e.g., Mabit et al. (2012) and Navas et al. (2017). The variability found for 210 Pb was much higher than that of 226 Ra: 36, 67, 42, and 60% for cores S1 to S4, respectively. This finding was expected given that the overall activity concentration results from the combination of two sources (in situ and in excess). However, the most remarkable feature of all the 210 Pb profiles was the fact that none of them did follow the typical unperturbed core profile. This one typically shows a maximum near the top associated with what should be a continuous atmospheric supply of this radionuclide. On the contrary, several layers had relatively homogeneous activity concentrations, and sometimes detectable activity concentrations were found below layers with undetectable activity concentrations. Furthermore, the activity concentrations of 226 Ra were, sample per sample, quite similar to those of 210 Pb. Consequently, as shown in Table 2, the number of samples having 210 Pb xs activity concentration higher than zero was extremely small.  The lack of 210 Pb xs at the upper part of S1 and S3 and its presence in certain deeper layers such as those detected in cores S1 and S2 can be explained by a combination of intensive removal of the upper part of the soil profiles removing 210 Pb xs -rich soil layers and the deposition of soil from neighbor sources. Average values of atmospheric fluxes of 210 Pb in Southern Spain coastal regions have been estimated in ranges of 50-60 Bq m −2 year −1 , while in Northern Spain, they reached 133 Bq/m 2 (Lozano et al. 2011(Lozano et al. , 2013Sánchez-Cabeza et al. 2007). Additionally, low contents in organic matter, low pH, or small precipitation could lead to a decreased retention of the atmospherically supplied 210 Pb in the soil particles (Meusburger et al. 2018). Combining these facts to low mountain climatology and the subsequent air mass transport in the west-to-east direction, it could be expected that the local supplies of 210 Pb xs to the topsoil were relatively scarce, even below the 15-25 Bq kg −1 level that has been characterized as typical of unsaturated Mediterranean soils (Abril et al. 2018).
In any case, certain doubts arise regarding the suitability of 210 Pb xs as a potential soil erosion proxy in this region. Similar limitations for the use of this proxy in certain scenarios have been argued by Mabit et al. (2014).

137 Cs
As for 210 Pb xs , 137 Cs was detected just in a few samples of every core (0, 4, 2, and 3 layers for cores S1 to S4, respectively). It is worth mentioning that it was not detected in any layer of the core S1, similar to 210 Pb xs , a fact suggesting the removal of the upper soil layers in this ground. The core S2 (top of the hill) shows, opposite to 210 Pb xs , a certain accumulation of 137 Cs in the top part of the core, being undetectable below 5.5 cm. Cores S3 and S4, on the other hand, were characterized by a small proportion of samples where 137 Cs could be detected including the top of the cores. Furthermore, it was also detected in deep layers despite the fact it could not be detected in many intermediate layers. These facts are again incompatible with an unperturbed profile.
The detected activity concentrations that have been found in this work cover the range 1.2-17 Bq kg −1 . This range is comparable to values found in other points of Southern Spain (Vaca et al. 2001;Gascó et al. 2006;Mabit et al. 2012;Guillén et al. 2015). The dependence of the range of concentrations with depth and with the different pluviometry regimes (Legarda et al. 2011) suggests comparing inventories rather than concentrations, as shown in Sect. 3.3. An immediate question arises regarding the distribution of 137 Cs along the different soil cores. In the case of unperturbed soil profiles, it would be expected finding an exponential decrease of activity concentration with a surface or sub-surface maximum. For perturbed soils (for example, after plowing), homogeneous activity concentration with depth may be expected. None of the soil cores fit to any of these two categories. 137 Cs was not detected in core S1, while in the case of core S2, as previously mentioned, the core perturbation avoids confirming a possible exponential decrease. However, it could be argued that, assuming an exponential decrease with depth of the 137 Cs activity concentration below 5-11 cm depth, its activity concentration could be below the limit of detection of the technique. Indeed, another interesting feature is the fact that no 137 Cs could be detected in core S2 below 10 cm depth, a fact that is in strong contrast with previous samplings in the South of Spain (Vaca et al. 2001;Mabit et al. 2012). This difference could be either a consequence of the soil recent evolution or, alternatively, an effect related to the properties of the soils of the sampling sites, such as the contents of clays and karstic leaching (Meusburger et al. 2016). The differences found for 137 Cs in cores S1 and S2 could be explained by an accumulation at the hill base of soil masses with a low 137 Cs concentration, such as subsoil eroded and transported from gullies (Tims et al. 2010). Indeed, this observation of higher inventories at higher grounds (S2, S4) is consistent with the behavior found for 210 Pb xs .
Soil core S3 showed small activity concentrations in the topsoil and a subsequent decrease below the detection limit below 5 cm depth. A possible explanation of this feature is that the upper layers of the soil have been removed. Under these conditions, this emplacement could be classified a priori as an eroded site. On the contrary, soil profile S4 shows quasi-homogeneous distribution in the upper part with a sharp increase in the middle-lower part of the profile. This profile suggests that the top flat peak where this core was taken was previously submitted to some kind of work involving soil mixing from different origins. The maximum value was quite similar to that found in the top layers of core S2.

Plutonium isotopes
In this work, Pu was detected in 2, 2, 4, and 6 layers of soil cores S1 to S4 (Table 2). In the case of the S1 core, Pu isotopes were the only artificial fallout radionuclides that could be detected, with very small activity concentrations of 239+240 Pu (42-65 mBq kg −1 ). In our case, no Pu was detected below the top 3 cm of core S1, and the small concentrations found here in the topsoil are, once again, consistent with the removal processes of the top layers of the ground. The maximum activity concentrations were found at core S4 (up to 350 mBq kg −1 ). Excepting core S4, Pu isotopes were detected up to a maximum depth in the range of 20 cm. These results agree with previous measurements in Southern Spain salt marshes soils (Gascó et al. 2006) or Mediterranean forest soils (Guillén et al. 2015) but in contrast to results from similar latitudes in Italy (39°N) (Raab et al. 2018). Hence, this finding supports the previous claim for intense upper soil removal.
Fallout Pu concentrations in soils are commonly below activity concentrations of 1 Bq kg −1 . For example, concentrations within the range of 50-700 mBq kg −1 have been found in the Sila Masif Upland (Italy) (Raab et al. 2018). Similar values have been reported for soils from northern areas in Europe such as the Swiss Alps (50-1000 mBq kg −1 . Similar higher activity concentrations (~ 10-900 mBq kg −1 ) have been found in soil cores collected from relatively upper latitudes of Spain (Guillén et al. 2015).
Core S2 showed for Pu isotopes a sharp peak at the top of the core with activity concentrations higher than core S1 (145-295 mBq kg −1 ). This fact is in agreement with the results found for 137 Cs, but then Pu was undetectable below the top of the core in the samples from layers immediately above and below the bioturbation layer (5-11 cm). In core S3, Pu was detectable in the upper layers (similar to 137 Cs), but also in several deeper layers, at low to intermediate activity concentrations (less than 150 mBq kg −1 ). Finally, core S4 showed a relatively homogenous profile within the upper 20 cm. The peak detected at 17.5 cm (same than for 137 Cs) cannot be considered significant given the magnitude of the relative uncertainty. Then a drastic drop of activity concentrations appeared in the lower part of the core, being quantification possible at concentrations slightly less than 15 mBq kg −1 . Again, these results suggest that Pu and Cs concentrations follow similar but not identical behaviors. Using the Pu + ICP-MS approach allows a better magnitude of the relative uncertainties and a higher sample throughput than using 137 Cs + radiometric measurement. This fact emphasizes the advantages of the use of Pu isotopes for the application of the FRN proxy. Indeed, the fact that their quantification was possible in several layers where 137 Cs was undetectable seems to suggest that the analysis of Pu could provide more information than 137 Cs does. It is also possible that this fact could be coupled to a possible higher vertical migration of Pu relative to Cs , as explained below.

Sediment core
The results obtained from the reservoir sediment core for 40 K, 210 Pb xs , 137 Cs, and Pu isotopes are shown in Fig. 5. Detailed data of activity concentrations of 210 Pb, 212 Pb, 226 Ra, and 234 Th are provided as supplementary material  (Table S2). It can be seen that the activity concentrations of 40 K, 210 Pb, and 212 Pb drop to about half below the 15 cm depth level. This fact suggests that the material deposition evolution in this core has changed through time and involved more than a single source, as explained below. Interestingly, the 210 Pb xs activity concentrations here reported for the upper part of the core seem to be among the highest values reported in this work (41-65 Bq kg −1 ). This observation was verified by comparing the populations of 210 Pb xs activity concentrations for soil samples (n = 10) and for the upper part of the sediment samples (n = 5). Medians instead averages were used because the first batch of samples did not correspond to Gaussian distributions. Mann-Whitney W-test at a significance level of 5% was applied, revealing that the median in the second batch of data was higher than for the first one, by a factor > 4 (P-value = 0.013). 210 Pb xs on the contrary remained nearly undetectable below the 15 cm depth. 210 Pb xs activity concentration is nearly homogeneous for the upper part of the core, and this feature is neither compatible with unperturbed soils (where exponential decline should be expected) nor with floodplain soils being developed under a non-episodic accumulation of 210 Pb (Abril et al. 2018).
On the other hand, 137 Cs was detected in relatively large concentrations (1-7 Bq kg −1 ) in the upper part of the core until 17.5 cm depth. Something similar was found for Pu isotopes, although detection was possible even at the lowest part of the core (< 150 mBq kg −1 ) despite they were not detected in intermediate sediment slices.
According to this, our initial hypothesis is that this sediment accumulates materials from different sources after uplands soil removal, downhill transport, and subsequent deposition on the sediment top. Assuming that the isotopic ratio 234 Th/ 212 Pb could be representative of the isotopic ratio 238 U/ 232 Th, which is a well-known indicator of the origin of particles in sediments (Ivanovich and Harmon 1992), the upper part of the column shows regular values (1.1 ± 0.4-1.8 ± 0.6) but raises to much higher values (3.3 ± 0.6-10.8 ± 4.3) below the 15 cm depth. Hence, the lowest part of the core should contain, possibly, a different substratum from that of the upper part of the core. This hypothesis is supported by the changes in activity concentrations of 40 K found below the 15 cm depth. Consequently, this sediment could belong to a depositional and mixing zone. The profile here described seems compatible with two possible scenarios: saturated sediments in the top while the sediment remains submerged (leading to an extremely low 222 Rn exhalation) with a scarce capacity for 210 Pb redistribution along the sediment column or floodplain soil developed by episodic supplies of material. The lack of 210 Pb xs below the 15 cm depth layers seems compatible, in turn, with unsaturated soil, where the lack of pore water allows radon exhalation through connected pores and a subsequent depletion of atmospherically supplied 210 Pb in the inner layers (Abril et al. 2018). Therefore, we argue that the sediment profile is reflecting an unsaturated soil substrate being covered with surface soil from another source, highly saturated, acting as a natural 210 Pb xs accumulator.
Summarizing, natural radionuclides such as 40 K, 212 Pb, and 234 Th suggest at least two different sources supplying material to this core with very different signatures, with a clear contrast above and below the 15 cm depth layer and even some intermediate mixing between them. This fact agrees with detectable activity concentrations of 137 Cs, 210 Pb xs , and 239+240 Pu above that depth and undetectable levels below that depth. The scenario suggested by the natural radionuclides in the sediment core resulted compatible, on the other hand, with the results found for artificial radionuclides. The retention of 137 Cs and 239+240 Pu in a top soil is modulated by erosion processes. As the soil column loses surface masses, where Pu and 137 Cs activity concentrations are higher, the inventory of these nuclides decreases. These isotopes subsequently accumulate in another location, acting as a sink. When a soil already depleted in these nuclides suffers additional erosion, this Pu and 137 Cs depleted material accumulates on top of an accumulation point. These processes of erosion, runoff transport, and deposition can be extended through many years. It can be the same location than before or not depending upon the existence of abrupt processes such as human-driven soil redistribution. Under these conditions, a same sink position can receive soil mass supplied by two different sources having different levels of retention of FRNs. The lack of detection below the 15 cm depth layer may be explained following two alternative possibilities: (1) the underlying soil could be supplied by an already eroded ground, before the incorporation of additional soil masses. That would define a complex scenario involving alternate, non-continuous episodes of upper soil removal and incorporation (i.e., top soil replacement instead of accretion). (2) This sediment location could act as a sink of soils from different locations having different previous histories. The lower part in the core would correspond to the deposition of soil supplied by gullies, which are depleted in fallout reactive particle nuclides (Tims et al. 2010). The top part of the sediment location should be supplied from topsoil keeping a part of their radionuclide inventories. This last possibility is coherent with the hypothesis depicted above based on the results obtained for 210 Pb xs . The random detection of Pu in several lower layers would be a combined consequence of two different effects: the better performances of ICP-MS for Pu isotope detection over the gamma analysis for 137 Cs and the ability of plutonium to be transferred deeper inside the soil/sediment column (Meusburger et al. 2016).

Inventories
As shown above, the interpretation of data can be complicated by the combination of several factors involving element's biogeochemistry and soil redistribution. From this point of view, a comparison of inventories could help to clarify the net role (source vs. sink) of every location. Figures 6, 7, and 8 summarize the 210 Pb xs , 137 Cs, and 239+240 Pu inventories calculated for the soil and sediment cores. 210 Pb xs inventories in soil ranged from 0 to 3.33 ± 0.72 kBq m −2 with an average of 1.7 ± 0.2 kBq m −2 for soils. Gascó et al. (2006) reported 0.1-9.4 kBq m −2 in soils of Doñana National Park, being the average 3.9 ± 1.8 kBq m −2 . Values of 1.94 ± 0.08 kBq m −2 were reported in reference sites at higher latitudes in Spain, NE Spanish Pre-Pyrenees (Gaspar et al. 2013), and different authors have reported inventories of up to 5.4-10.9 kBq m −2 at reference sites in different locations around the region of Calabria, Italy (Porto et al. 2012). The inventories found in the soil samples indicate extensive removal of the upper layers of soils at the sampled points. This last possibility suggests in turns that the use of 210 Pb xs could be compromised despite its successful application in other emplacements having similar climate conditions. On the contrary, the inventory found in the sediment core was noticeably high (9810 ± 1008 Bq m −2 ). From a qualitative point of view, the different magnitudes of the inventories in soil suggest that the use of 210 Pb xs could allow discriminating eroded sites (S1 and S3) and depositional sites (Sed-1).
The inventories of 137 Cs in soil cores ranged between 0 and 1049 ± 237 Bq m −2 (Fig. 9). The maximum value corresponded to core S4, and it overlapped the inventory calculated for the sediment core, 1087 ± 79 Bq m −2 . This range of inventories is slightly less than previous data published for reference sites sampled at similar latitudes (Montefrío, Granada, Spain, 37°19ʹ N, 4°0ʹ W), having a value of 1.9 ± 0.2 kBq m −2 , and it is also the same order of magnitude than values predicted for 137 Cs deposition with the above-mentioned mean yearly rainfall (1400-1800 Bq m −2 ) (Legarda et al. 2011). As expected, the values of this work inventories are usually much less than inventories of reference sites belonging to Northern latitudes in Spain; see, for example, Menéndez-Duarte et al. (2009) andSánchez-Cabeza et al. (2007). On the contrary, just 1570 ± 80 Bq m −2 were reported in Pre-Pyrenees lands .
The general trends for 137 Cs inventories (Sed-1 ≈ S4 > S2 > S3 > > S1) clearly show that the locations of S3 and S1 suffered an intense removal of the upper soil layers, as suggested in the previous section. In this case the conclusion agrees with that provided by the 210 Pb xs inventories. On the contrary, S4 inventory was clearly higher than that of S2, while in the case of 210 Pb xs , they were quite similar taking into account the corresponding uncertainty intervals. Furthermore, the use of 137 Cs inventories do not allow a Finally, in the case of Pu isotopes, the inventories ranged from 1.9 to 60.6 Bq m −2 . Similar to that found for 137 Cs, the inventories calculated for S4 and Sed-1 were within the same order of magnitude (60.6 ± 3.9 and 70.0 ± 2.3 Bq m −2 , respectively). Using the linear relationship between Pu deposition and average annual rainfall proposed in (Gascó et al. 2006), the expected inventories lie between 40 and 54 Bq m −2 , a range that is in good agreement with our maximum values for soils. The availability of data regarding fallout-level Pu inventories of the South of Spain is very scarce. Values from 16.4 Bq m −2 (eroded site) to 101.1 Bq m −2 (depositional site) have been published for salt marshes soils in Doñana National Park (Gascó et al. 2006). On the other hand, 256.2 Bq m −2 was reported at a dam sediment column from SW Spain (Abril et al. 2018), while Chamizo (2009) reported 45-137 Bq m −2 in estuarine sediment cores. In this case, the relationship between inventories shows that Sed-1≈S4 > > S3 > S2 > S1.
Location S4 is, in relative terms, a maximum of inventories among the soil cores for both natural and artificial radionuclides. S1 and S3 are clearly revealed as erosion sites. Sed-1 reveals as a potential slight accumulation site, which is much more pronounced in the case of 210 Pb xs , possibly as a consequence of the reasons explained in Sect. 3.2. The main differences appear for emplacement S2, which is clearly an eroded site according to 137 Cs and 239+240 Pu inventories, but not according to 210 Pb xs when relative uncertainties are taken into account.
Another derivative of the data here shown relates to the different sensitivity offered by the three approaches used in this work. Such sensitivity should be considered in terms of (1) the ratio of typical order of magnitude of the activity concentration to limit of detection, (2) the relative magnitude of uncertainties for both activity concentrations and inventories, and also (3) the ability to quantify the presence of the nuclides in deeper layers. From these points of view, Pu isotopes seem to be a most remarkable proxy for soil redistributions than 210 Pb xs or 137 Cs. Indeed, the combined use of 137 Cs and Pu isotopes can reinforce the discussion of the results, especially when the calculation of erosion rates requires corrections of the estimative values by the soil particle size, as suggested in Meusburger et al. (2016). Otherwise, a systematic overestimation of the soil redistribution rates could be derived due to the preferential transport of small size particles, which are assumed to be enriched in 137 Cs regarding Pu isotopes.
Interesting issues appear if 239+240 Pu/ 137 Cs activity ratios are calculated for the full inventories instead individual samples. S4 reflects an inventories ratio of 0.058 ± 0.014, which agrees quite well with the inventories ratios calculated for soils in Mediterranean forests, 0.063-0.083 (Guillén et al. 2015). We argue that the deeper penetration of Pu regarding Cs can be detected in eroded soils. Removal of surface soil layers leads to inventory ratios above the reference value in the eroded sampling site. This is the situation that we have found for core S3, where the inventory ratio was 0.184 ± 0.055 (z-score = 1.42). If erosion continues deeper than Pu maximum, the resulting inventory ratio decreases, as found for core S2 (0.012 ± 0.002; z-score = −0.92).
The profile of soil S4 did not correspond neither to unperturbed nor plowing profile. On a strict sense, it should not be the best possible option for a reference profile for the sake of calculation of erosion rates. However, the good agreement of 137 Cs and 239+240 Pu inventories at this core with the values found at similar latitudes suggests the possibility to use it in order to evaluate erosion rates on a relative basis. To do that, MODERN was used as explained above. In this case, we did transform the data collected in the profile S4 for Cs and Pu isotopes, with variable slice thickness, into constant 3 cm thickness. The other supplied data were the radionuclide inventories at the sampling locations (S1, S2, S3 and Sed-1). The simulations were not applied to 210 Pb xs data due to their low quality, in relative terms.
The simulations for 137 Cs did reveal integrated soil losses of − 16, − 16, and − 22 cm for soil cores S1, S2, and S3 for the period 1963-2017. For Pu isotopes, the results were in reasonably good agreement with the former results: − 20, − 18, and − 16 cm, respectively. The results obtained for 137 C correspond to soil erosion rates in the range of 34, 34, and 41 t·ha −1 ·year −1 , respectively (43, 38 and 34 t·ha −1 ·year −1 , respectively, using Pu isotopes). These values are consistent with RUSLE predictions in the study area (12-65 t·ha −1 ·year −1 ); details of the application of RUSLE at this scale (1:400,000) can be found in (Ministerio de Agricultura, Pesca y Alimentación 2008). However, the scale of this approach does not allow obtaining accurate data for the specific coordinates of the sampling points. Similarly, the average erosion rates in the range of 10-40 t·ha −1 ·year −1 for a set of farms located between Bembézar and Retortillo dams were calculated by Alcántara Jurado et al. (2006). In any case, the calculations show erosion rates that can be described as severe, as expected. On the other hand, MODERN predicts two different scenarios for the location of Sed-1 depending upon the used radionuclide. For 137 Cs, a slight soil loss is predicted, − 1.2 cm, while for Pu isotopes, a slight deposition rate is predicted (2.8 cm). These differences are certainly the result of the fact that the inventories at S4 and Sed-1 are barely the same for Cs isotopes when the uncertainty bars are considered, although higher resolution was achieved using Pu isotopes.

Summary and conclusions
Different soil redistribution markers have been analyzed in soil and sediment cores collected in a region where severe to very severe soil erosion was expected. The limitations of the use of 210 Pb xs as a proxy for the calculation of soil erosion rates in semiarid environments with high erosion rates have been shown. The exploratory use of Pu isotopes, compared to 137 Cs, revealed certain advantages for the former in terms of analytical sensitivity and sample throughput. The collected data suggests intense soil erosion since 1954 to 2012 (34-43 t ha −1 ·year −1 ), which are within the order of magnitude proposed through the application of RUSLE. For the collected reservoir sediment core, discrepancies between the data provided by 210 Pb xs on one hand, and the artificial radionuclides on other hand, can be explained by arguing that the sediment profile was reflecting an unsaturated soil substrate being covered with highly saturated surface soil from another source and acting therefore as a natural 210 Pb xs accumulator. The use of Pu as a tracer of erosion processes in semiarid areas needs to be further evaluated in future studies at different sites around the world in order to demonstrate how affordable is its use as an universal alternative to the use of 137 Cs and 210 Pb xs .