Genesis of magnetic anomalies and magnetic properties of archaeological sediments in floodplain wetlands of the Fossa Carolina

Floodplain wetlands are complex systems influenced by many natural and anthropogenic operators. Due to the influence of high and varying groundwater table and high organic contents, geophysical prospection in wetland floodplains quickly reaches the limits of its effectiveness. At the Early Medieval canal Fossa Carolina in southwest Germany, a study design employing magnetometry, drillings, sampling, and in situ rock magnetic measurements was used for environmental magnetic interpretation of magnetic anomalies in magnetograms and sediment layers. This approach offers reliable archaeological interpretation of magnetic anomalies and magnetic properties under the site specific sedimentological conditions of a floodplain wetland. It was also found that man‐made magnetic anomalies in the floodplain are due to the genesis of different remanent magnetizations – specifically, greigite (Fe3S4) can cause distinct magnetic anomalies in floodplains that can be recognized readily in surface magnetic data.


| INTRODUCTION
Alluvial and semi-terrestrial systems can contain remarkable archaeological record (Brown, 1997). Numerous archaeological and geoarchaeological studies have been carried out in alluvial systems during the last decades (Brown, 1997;Howard, 2003;Needham & Macklin, 1992). Every alluvial archaeological record represents a complex interaction between cultural and geomorphic operators (Needham & Macklin, 1992). To better understand the underlying material structure, the floodplain topography and its genesis must be considered, including all biological, chemical, and physical aspects of specific floodplain wetlands (Brown, 1997;Howard & Macklin, 1999).
Site-scale magnetic prospecting is one of the most powerful and effective tools in archaeological prospection (Aspinall, Gaffney, & Schmidt, 2008;Fassbinder, 2015;Kvamme, 2009). However, alluvial environments are often considered problematic for magnetometry.
Due to the limited depth range of magnetic prospecting the overlying sedimentary cover can offer difficult conditions for signal penetration (Clark, 1992). Also, the choice of instruments can produce different results (Linford, Linford, Martin, & Payne, 2007). Nevertheless, sitedependent variables are the primary factors for the detection and interpretation of archaeological structures by magnetic prospection (Cuenca-Garcia et al., 2018;English Heritage, 2008). In this case the most important site-dependent variables are rock, sediment, and soil magnetism (Dunlop & Özdemir, 1997;Evans & Heller, 2003;Jordanova, 2017;Liu et al., 2012;Thompson & Oldfield, 1986). A broad understanding of these site-specific variables is necessary for detailed interpretation of significant archaeological structures in magnetograms (Fassbinder, 2015).
Man-made magnetic anomalies have been found at several sites in river floodplains (Weston, 2001), including Fossa Carolina . So far, the genesis of these anomalies has only been discussed in general terms without considering the physical properties of floodplain sediments. Therefore, the aim of this article is to provide a detailed magnetic investigation of the genesis of magnetic anomalies in the floodplains of Fossa Carolina.

| STUDY AREA
In AD 792/793, Charlemagne started to build a navigable connection between the Rhine-Main catchment and the Danube catchment. This canal is called Fossa Carolina and was one of the major hydroengineering projects in Early Medieval Europe. Bridging of the Central European watershed was of high geostrategic relevance, due to the expanding fluvial communication and transportation network of the Franconian Empire at that time (McCormick, 2002;. Nevertheless, the canal was never finished  and only remains of it are visible today (Schmidt et al., 2019;Schmidt, Werther, & Zielhofer, 2018). The post-Carolingian canal filling was reconstructed using excavation, drillings, and sediment analysis (Völlmer et al., 2018;.
The Fossa Carolina is located on a valley watershed between the Altmühl and Swabian Rezat Rivers in Franconia, Bavaria, southern Germany. The valley fills consist of Pleistocene sandy to loamy fluvial sediments. The canal has a length of 2.9 km, as proven by drillings , archaeological excavation (Werther et al., 2015), and direct-push sensing Völlmer et al., 2018). Furthermore, extensive multi-disciplinary research has been carried out at the Early Medieval canal over the last decade (Ettel, Daim, Berg-Hobohm, Werther, & Zielhofer, 2014;Hausmann et al., 2018;Kirchner et al., 2018;Linzen et al., 2017;Schmidt et al., 2018;Schmidt et al., 2019;Stele, 2017;. Therefore, the site offers an excellent case study. Two sections of the Fossa Carolina floodplain were selected for our study: the West-East Section and the Northern Section foothills of the Weissenburger Alb of the southern Franconian Jura (Jätzold, 1962) at an elevation of 413 to 420 m above sea level.
They are situated in water sensitive areas and covered by bogs, floodplain gley soils, and colluvial deposits (LfU-WMS, 2019; . The sections are divided by the current watershed. Due to construction of the canal the original watershed was relocated 850 m toward the northeast and the West-East Section . At present, the areas differ slightly in composition and elevation: few remains of the original excavation ramparts of Fossa Carolina have been detected in the Northern Section by LiDAR (light detection and ranging) digital elevation models (DEMs), and the entire structure is clearly visible in the East-West Section (Figure 1).

| METHODS
The study design is illustrated schematically in Figure 2. It is based on a sequence and combination of geophysical procedures and on-site methods. Esri geographic information system (GIS) ArcGIS Desktop was used for geodata handling, management, spatial analysis, and mapping (Esri, 2019). For geodetic surveying of spatial data (e.g. magnetic surveys and drillings) a Topcon HiPer II global navigation satellite system (GNSS receiver) was used (Topcon, 2019). The entire workflow comprises magnetometry, identification of drilling sites, drillings, sampling, and in situ rock magnetic measurements. Environmental magnetic interpretation was carried out based on the results of the magnetometry and in situ magnetic measurements ( Figure 2).

| Magnetometry
The magnetometry was carried out with a Bartington Instruments

| Drilling campaigns
Based on the results of the magnetometry, drilling locations were defined using a GNSS receiver. The magnetic anomalies that indicate canal structures were sampled with two core transects of three cores each using a closed 50 mm polyethylene inliner ( Figure 3).

| Magnetic susceptibility measurements
Volume magnetic susceptibility measurements were carried out with a Bartington Instruments MS3-Meter-MS2C sensor configuration to localize magnetically conspicuous layers in the core profile (Dearing, 1999). The closed inliners were passed through the loop of a MS2Csensor and measurements were carried out with a spatial resolution of 2 cm and a susceptibility accuracy of ±2 × 10 -6 SI (Bartington Instruments, 2019b). After mass anomalies (κ peaks) were identified in the cores, anomalies were sampled and their stratigraphic location determined by comparing the set of cores in the transects. Prior to their further analysis with a variable field translation balance (VTFB), the original samples were analysed for mass specific susceptibility (χ) and frequency dependent susceptibility (χ fd% ) with a Bartington Instruments MS3-Meter-MS2B sensor configuration (Bartington Instruments, 2019b;Dearing, 1999;Dearing et al., 1996).

| Variable field translation balance (VFTB)
Field dependent and temperature dependent magnetization measurements (Krása, Petersen, & Petersen, 2007) were carried out on all samples with apparent mass anomalies. Hysteresis loops and backfield curves with the key points saturation magnetization (M S ), saturation remanent magnetization (M RS ), coercivity force (B C ), and coercivity of remanence (B CR ) were estimated which allows a differentiation of F I G U R E 2 Work flow of the approach to magnetic anomaly interpretation presented in this study magnetic properties of mass anomaly producers (Day, Fuller, & Schmidt, 1977;Dunlop, 2002aDunlop, , 2002bEvans & Heller, 2003;Roberts, Tauxe, Heslop, Zhao, & Jiang, 2018). In addition, the shape of hysteresis loops and backfield curves allows a first magnetomineralogic typification (Fabian, 2003;Tauxe, Bertram, & Seberino, 2002;Tauxe, Mullender, & Pick, 1996). From temperature dependent magnetization measurements (M/T) Curie temperatures and crystallographic phase transitions of magnetic minerals can be inferred which allows identification of remanence carriers in the sample (Dunlop & Özdemir, 1997;Hanesch, Stanjek, & Petersen, 2006;Moskowitz, 1981). Visualization, analysis and interpretation of VFTB measurements were supported by the RockMagAnalyzer 1.0 software designed by Leonhardt (2006). 2.30 m below ground, this strongly ferrimagnetic layer is duplicated by strata with maximum κ values of 15 000 × 10 -6 SI. Interbedded are sandy alluvial sediments with moderate κ values between 80 and 150 × 10 -6 SI. The upper colluvial sediments also have moderate maximum volume magnetic susceptibility of 200 × 10 -6 SI. For further analysis of the mass anomaly remanence carriers, samples for VFTB measurements and mass specific and frequency dependent susceptibilities were taken from the layers with (extremely) high κ peaks in QP2 and QP1 (Figures 4 and 5).

| Magnetic properties of mass anomaly samples
Hysteresis loops for all mass anomaly samples in Figure 6 Table 1). The χ fd% value and its ratio with χ indicate superparamagnetic (SP) enhancement as a result of burning (Dearing, 1999). Hysteresis measurements point to the magnetite type with only sample 2 having a potbellied hysteresis loop shape parameter (Fabian, 2003). The values of B C and B CR are relatively low, which indicates soft magnetic material and, thus, high magnetic viscosity of the remanence carrier.
Hysteresis and coercivity ratios of both samples indicate vortex state, a mixture of different magnetic grains (Day et al., 1977;Dunlop, 2002a;Roberts et al., 2017). The thermomagnetic measurements in Figure 8(Q-X) and the calculated dominant magnetic phase transitions have marked differences (see also Table 1). The heating cycle of sample 2 (Figure 8(R)) has a magnetization increase between 150 and 250 C. This range demarcates the Curie temperature which could point to titanomagnetite (Fe 2.4 Ti 0.6 O 4 ) (Dunlop & Özdemir, 1997;McElhinny & McFadden, 2000;Soffel, 2002). The heating curve for sample 2 then decreases to zero magnetization at 600 C.
The heating curve for sample 4 (Figure 8(V)) has a convex shape until it rapidly reaches zero magnetization at 560-630 C (see second derivative in Figure 8(W)). The calculated dominant phase transition (Table 1) (Table 1; (Dearing et al., 1996)). They reach M S at 230 and 280 mT (Figure 6(A, C)). In contrast to samples from group 1, they have relatively low M S and much higher B C and B CR . The hysteresis loop shape for both samples is distinctly potbellied (Fabian, 2003). The ratios M RS /M S and B CR /B C indicate the dominance of high coercivity single domain (SD) particles. The heating cycles for both samples (Figure 8(B, J)) have three temperature dependent magnetic phase transitions. The first transition is marked by a bend in the heating curve with an increase of magnetization at 100 C, demarcation the Curie point of goethite (αFeOOH) (Dunlop & Özdemir, 1997). The then following steady magnetization increase ends abruptly at 350 C, which is indicative for a maximum unblocking or alteration temperature of greigite (Fe 3 S 4 ) (Skinner, Erd, & Grimaldi, 1964) as the dominant solid magnetic phase of samples 6 and 10 (Chang et al., 2014;Reynolds et al., 1994;Roberts, Chang, Rowan, Horng, & Florindo, 2011). After the general magnetization decrease, there is still a small peak at 580 C demarcating the partial transformation of greigite to magnetite.
Stratigraphically, these samples were situated below the samples from group 2 and are, thus, also part of the canal fillings (Figures 4   and 5). Although they have different hysteresis and backfield parameters, they have a similar thermomagnetic behaviour ( Figure 8(E-H, M-P)). Both heating curves decrease rapidly at 300 C to 400 C, which demarcates greigite as the dominant remanence carrier (Table 1). During the heating cycle greigite transforms entirely to magnetite. However, the greigites in samples 8 and 12 have different particle size distributions, as indicated by different hysteresis and coercivity ratios (Figure 6(B, D): Sample 8 is already saturated at 250 mT, while sample 12 reaches saturation at 300 mT). Also, B CR /B C for sample 8 is high (Figure 6(D) and Table 1), which indicates probably SP enrichment (Rowan & Roberts, 2006;Rowan, Roberts, & Broadbent, 2009).

| Magnetic anomaly interpretation: an environmental magnetic perspective
The magnetic properties of the various mass anomaly samples responsible for the linear magnetic anomalies in magnetograms at Fossa Carolina originate not only from different natural remanent magnetization (NRM), but also from different post-depositional genetic processes (Evans & Heller, 2003;Liu et al., 2012). The linear, intensive, and dipolar anomaly originates in a 40 cm thick layer of black, reddish-brown and red layers at a depth of 1.35 to 1.55 m. The highest κ values occur in a brown to reddish-brown, layer with a high allochthonous, ferrimagnetic titanomagnetite fraction. The high viscosity of the soft magnetic titanomagnetite and magnetite triggered the orientation of the remanence carriers in the soft water-soaked sediments toward Earth's magnetic field vector, thus causing intensive dipolar anomalies. In areas with simple positive magnetic anomalies, we assume that these layers are not under the influence of ground water (south of the Northern Section in Figure 3).
In natural environments titanomagnetite formation can occur in two forms: i. Primary formation of titanium rich magnetic iron compounds can only take place in magmatic rocks (Schmincke, 2010;Schön, 2011).
There are no basaltic rocks or their weathering products in the river basin of the Swabian Rezat (ArcTron 3D GmbH, 2009;Schmidt-Kaler, 1976). Therefore, the magnetic mass anomalies of the group 1 T A B L E 1 Selected environmental magnetic parameters and dominant magnetic phases (lower row)  Figure 8).
samples must have been produced by an allochthonous/anthropogenic influence. This is supported by the fact that sediment colours and magnetic properties of these layers show clear signs of fire exposure. The findings of , that the titanomagnetite/magnetite layer, which produces the intensive dipolar anomaly, correlates stratigraphically with the construction of Fossa Carolina, and the fact that these intensive anomalies in the SQUID data of Linzen and Schneider (2014), which show that the same remanence carriers are found in the entire northern part of the canal, support the interpretation of the mass anomalies as anthropogenically introduced/influenced and burned materials. Additional (geo) archaeological research is necessary to determine the source and specific characteristics of this material.
Greigite-generated mass anomalies (groups 2 and 3), which partially follow the course of the canal and of the canal fills, is supported by the fact that samples 6 and 10 have similar magnetic behaviour as the SD greigites found by Faßbinder and Stanjek (1994), Hall, Cisowski, and King (1997), and Roberts et al. (1998). The characteristics of sample 12 are comparable to those reported by Jelinowska et al. (1998). The magnetic properties of sample 8 relate to genetic processes, which we interpret to be similar to those described by Rowan et al. (2009). All layers with greigite-generated mass anomalies may have undergone post-sedimentary remagnetization: These highly coercive SD greigites were generated as secondary/authigenic formations under highly organic, anaerobic conditions after deposition of the organic rich canal fillings. Thus, the organic canal fills offered the necessary conditions for producing mass anomalies due sulphidic remanence carriers and the sediments acquired their stable chemical remanence magnetization (CRM) (Snowball, 1997). These findings are similar to those of Stanjek, Fassbinder, Vali, Wägele, and Graf (1994) in gley soils, where the environmental conditions were similar to those in the canal fills. Also, the hysteresis and coercivity ratios of the greigite in the canal fills point at differences in the domain state of the dominant solid FeS magnetic phases. For a correct domain state diagnosis, further analyses such as FORC diagrams or scanning electron microscopy/transmission electron microscopy (SEM/TEM) observations are necessary (Chang et al., 2014;Roberts et al., 2019). Should these methods deliver uncertain results, methods such as X-ray diffraction could also be used (Linford, Linford, & Platzman, 2005).
In order to make such magnetic and non-magnetic analyses possible in the future, we recommend the use of closed polyethylene inliners when drilling/sampling, because they allow longer storage of the samples without any changes in the magnetic mineralogy during the storage (Stele, 2017). Samples for in situ magnetic measurements should be then taken from the inliner immediately before analysis.
This sample preparation approach prevents the oxidation of possible ferrimagnetic iron sulphides in the samples.

| CONCLUSIONS
At the Fossa Carolina in southern Germany a workflow comprising magnetometry, identification of drilling sites, drillings, sampling and rock magnetic measurements was used to interpret magnetic anomalies. The approach offers a reliable estimate of: • the depth of magnetic mass anomalies; • differentiation between natural and anthropogenically induced structures; and • genetic processes leading to the production of mass anomalies.
Our results indicate that secondary/post-sedimentary processes in the canal fillings produced magnetic anomalies in the Fossa Carolina. Greigite-generated mass anomalies in deeper sediment layers can cause detectable anomalies in near-surface magnetic surveys with a 1 m fluxgate at a depth of up to 3.5 m. There is also strong evidence that titanomagnetite/magnetite-generated mass anomalies at Fossa Carolina were anthropogenically introduced. Evidence of greigite in organic-rich canal fills in a European Watershed means that these processes can be expected in palaeochannels or active channels in middle European floodplains with similar characteristics.