Biostratigraphic Evidence Relating to the Age-Old Question of Hannibal's Invasion of Italy, I: History and Geological Reconstruction

Controversy over the alpine route that Hannibal of Carthage followed from the Rhone Basin into Italia has raged amongst classicists and ancient historians for over two millennia. The motivation for identifying the route taken by the Punic Army through the Alps lies in its potential for identifying sites of historical archaeological significance and for the resolution of one of history's most enduring quandaries. Here, we present stratigraphic, geochemical and microbiological evidence recovered from an alluvial floodplain mire located below the Col de la Traversette (~3000 m asl—above sea level) on the French/Italian border that potentially identifies the invasion route as the one originally proposed by Sir Gavin de Beer (de Beer 1974). The dated layer is termed the MAD bed (mass animal deposition) based on disrupted bedding, greatly increased organic carbon and key/specialized biological components/compounds, the latter reported in Part II of this paper. We propose that the highly abnormal churned up (bioturbated) bed was contaminated by the passage of Hannibal's animals, possibly thousands, feeding and watering at the site, during the early stage of Hannibal's invasion of Italia (218 bc).


B I O S T R AT I G R A P H I C E V I D E N C E R E L AT I N G T O T H E A G E -O L D Q U E S T I O N O F H A N N I B A L' S I N VA S I O N O F I TA LY, I : H I S T O R Y A N D G E O L O G I C A L R E C O N S T R U C T I O N *
Controversy over the alpine route that Hannibal of Carthage followed from the Rhône Basin into Italia has raged amongst classicists and ancient historians for over two millennia. The motivation for identifying the route taken by the Punic Army through the Alps lies in its potential for identifying sites of historical archaeological significance and for the resolution of one of history's most enduring quandaries. Here, we present stratigraphic, geochemical and microbiological evidence recovered from an alluvial floodplain mire located below the Col de la Traversette (~3000 m asl-above sea level) on the French/Italian border that potentially identifies the invasion route as the one originally proposed by Sir Gavin de Beer (de Beer 1974). The dated layer is termed the MAD bed (mass animal deposition) based on disrupted bedding, greatly increased organic carbon and key/specialized biological components/compounds, the latter reported in Part II of this paper. We propose that the highly abnormal churned up (bioturbated) bed was contaminated by the passage of Hannibal's animals, possibly thousands, feeding and watering at the site, during the early stage of Hannibal's invasion of Italia (218 BC).

INTRODUCTION
The main approach routes (Livy, 1965;Proctor 1971;de Beer 1974;Polybius 1979) and associated mountain passes (cols) identified as likely transit points of Hannibal's army into Cisalpine Gaul (northern Italy) have been debated and discussed by historians for over two millennia (Freshfield 1886(Freshfield , 1899Dodge 1891;Wilkinson 1911;de Beer 1956de Beer , 1967de Beer , 1969de Beer , 1974Connolly 1981;Seibert 1993;Lazenby 1998;Bagnall 1999;Lancel 1999;Mahaney 2004;Mahaney et al. 2007bMahaney et al. , 2008aMahaney et al. , 2008bMahaney et al. , 2008cMahaney et al. , 2010aMahaney et al. , 2010bMahaney et al. , 2010cMahaney et al. , 2014. These invasion passages are referred to here as the southern, northern and intermediate routes, as shown in Figure  1 (Hart 1967;Proctor 1971;de Beer 1974). The southernmost follows the lower Rhône River valley northwards to the Drôme River, thence across the Col de Grîmone (1318 m; Alpes du Dauphiné) to the Durance Basin, through the Queyras (Guil River) to the Col de la Traversette, finally exiting through the two-tier rockfall similar to that described by Polybius (1979) into the upper Po River catchment. The northern route parallels the Rhône River to the Isère River, past present-day Grenoble along the Arc River to either the Col du Mont Cenis (2083 m) or the Col du Clapier (2497 m), finally exiting into the Dora Riparia west of Torino (Turin in Fig. 1). The intermediate route follows the same initial course as the northern route, then west of the Pelvoux Massif to join the Durance River continuing towards the Col de Mont Genèvre (1830 m), exiting into the Dora Riparia. After two millennia, the question remains unsettled of the specific path over the Alps taken by Hannibal or by Hasdrubal, his brother, who followed him 11 years later (Proctor 1971;Lazenby 1998;de Beer 1974). However, time-motion analysis of these three potential routes, and investigations of each pass into Italia carried out by Mahaney (2008) revealed the southern route to be the most direct and highest track into northern Italy from the Rhône Basin, and all authorities agree that Hannibal had to march with speed to cross the Alps before winter set in. Furthermore, Polybius (1979), a primary ancient authority who interviewed survivors of the invasion and retraced the invasion route some 60 years after the fact, is definite that Hannibal descended upon the Po River plains rather than the Dora Riparia approximately 50 km to the north.
Uncertainty regarding Hannibal's route stems from contested interpretations of ancient place names, preferential use of either Livy (1965) or Polybius (1979) as a primary ancient source (the two are not compatible) and competing views by other historians of travel times between points.
Traditional arguments have not relied on physical or environmental parameters (Mahaney 2008), because no physical artefacts (nails, coins, buckles etc.) have ever been found on any of the potential alpine routes. However, organic compounds associated with people and animals can easily persist for several millennia and are easy to detect, leading us to hypothesize that Hannibal's army and his horses/mules/elephants may have left organic evidence of their passage. To investigate this possibility, we located a site at~2580 m asl along the southern route in the Alps that could have been used as a campsite for a large group of people and animals (~30 000 infantry,~7000 cavalry, unknown number of handlers and 37 elephants; Lancel 1999) (Fig. 2). Then, we used multidisciplinary techniques to search for pedological, micropalaeontological, biochemical and lipid biomarker evidence that might help determine if the site was along Hannibal's invasion route.

ENVIRONMENTAL FACTORS
There are distinctive topographical landmarks for the Hannibalic route that are identified in the ancient literature (Proctor 1971;de Beer 1974;Polybius 1979;Mahaney 2008;Mahaney et al. 2014). These include a gorge on the approach to the summit, a 'certain bare rock', a bivouac near the col, a view into Italia, fired rock and a regrouping area, and all of these elements are common to most potential invasion routes. However, the only route that meets two key historical descriptions-the rockfall and frozen ground-is the southern route across the Col de la Traversette into the Po River plains of northern Italy.
Of all environmental factors used to identify the invasion route (Mahaney 2008), there are two paramount clues. The first is the superposed two-tier rockfall described by Polybius (1979). It is interesting that of all the repeat histories published over the years, and quotes from both Livy and Polybius, few, if any, have focused on Polybius' mention of a two-tier deposit, one mass produced by two mass wasting events. Following a survey of all likely passes used by the Punic Army (Mahaney 2008), the only such rockfall debris below all the cols that could have blocked the infiltrating army is located on the lee flank of the Alps in the upper Po Valley, below the Col de la Traversette (Mahaney et al. 2010c(Mahaney et al. , 2014 at 2600 m asl. Perhaps relying on a first-hand account by Silenus (Hannibal's historiographer; Lancel 1999), Livy maintained (Livy 1965, XXI, 37) that the rockfall provided such an impediment to passage of the army that Hannibal ordered its firing to allow the passage of the horses and elephants. According to this account, wood was brought to the site at 2600 m asl, then set alight, producing sufficient heat to physically disrupt and crack boulders that could not be moved by hand. Bagnall (1999) questioned Livy's interpretation, arguing that green wood would not be expected to burn. Wood might have been available up to perhaps 2400 m asl (Mahaney et al. 2010b), provided that one might expect starving soldiers to be capable of harvesting/transporting it to the rockfall, also a doubtful outcome. Not even Kuhle and Kuhle (2012), who mention the firing event, were able to substantiate how it might apply to the northern route across the Col du Clapier, their preferred route of passage. Previous work by  showed conclusively that while the rockfall described by Polybius exists below the Col de la Traversette, there is no evidence that it was fired, as no carbonized/split boulders are resident in the rockfall mass (Sodhi et al. 2006). Moreover, Polybius is mute on the firing event, and since he was the prime authority who followed the invasion route, the obvious conclusion is that, despite its perpetuation, the firing never happened.
The second vital clue is the presence of frozen ground. Sporadic permafrost, where present today, is deeply buried in the Western Alps (Mahaney et al. 2007a), and that was probably true in Hannibal's time, because today's climate is similar (Neumann 1992). Thus, it is likely that the major terrain obstacle of frozen ground referred to by Polybius (1979 was not permafrost but, rather, discontinuous frozen soil or firnpack (dense snow) covered with fresh snow. The Col de la Traversette, compared to the other mountain passes, is the only pass that reaches to within the snowline (~2800 m asl), with discontinuous permanent snow (firnpack). At lower elevations, the Col Agnel and the Col la Croix, adjacent to Traversette (Fig. 1), provide somewhat shorter routes into Italy, but lack firnpack, which is a prerequisite to identifying the correct col. Moreover, none of the other major cols-Clapier, Genèvre and Cenis-have permanent snow/firnpack or blocking rockfall, which is why de Beer (1974) and other researchers (Mahaney 2008) favoured the Traversette as the route Hannibal used to cross the Alps. Even with elevations imprecisely known at the time, Polybius's (1979) description of the invasion route strongly suggests the Traversette as the crossing point into the peninsula, an inference supported by Varro's mention in De re rustica (quoted by Proctor 1971) of the passes in geographical order from north to south-'Hannibal's Pass' lying south of the Col de Genèvre (Fig. 1). Moreover, reference to Varro's mention of the five cols, including Hannibal's Col as the highest, is quoted in Servius' commentaries, translated in Savage (1934).
In addition to the two-tier rockfall and the presence of frozen ground, a third requirement for a successful passage into Italy is the presence of forage plants (grasses and sedges) and sufficient watering holes for animals and humans alike. If the Col de la Traversette is the route taken by Hannibal, as all topographical data indicate (Mahaney 2008), the mire sites investigated here (G5 and G5a-described below) could have provided an area of forage and fresh water for use by Hannibal's troops and animals. The mire evidence presented here strongly supports previous estimates that the southern route through the Traversette was the route taken by Hannibal.

SITES
Recently, Mahaney (2008) used time-motion analysis, elevation modelling and field observations on sites along the southern route originally favoured by de Beer (1974) to identify potential hearths and watering hole/feeding grounds that might have been used by the Punic Army during transit. The army would have had access to stream water in many places along the Guil catchment (west flank), but there are only a few grassy and level areas above 2000 m asl upon which to graze horses and provide feed. After searching those areas for soils composed mainly of organic materials (Histosols; Soil Survey Staff 1999), we located a candidate site~425 m below the Col de la Traversette. The~60 ×~40 m wide location (latitude/longitude: 44°42' 587N; 07°0 32 74E; 2580 m asl), designated G5 and G5a, is the only organics-rich peat bed (mire) above 2000 m asl in the Guil Valley along the southern route to the Col de la Traversette. It lies 30 m higher than nested Late Glacial (~13-15 ka) and Younger Dryas (~11-13 ka) moraines in the lower Guil Valley (Mahaney and Keiser 2012;) and below the level of nearby talus, all of which are Little Ice Age (LIA) or post-LIA in age. Therefore, sediments of Hannibalic age should be present.
The site is located on a gently sloping bedrock bar composed of metabasalt (Tricart et al. 2003), lying adjacent to the headwaters of the Guil River and, therefore, is subject to periodic overbank flooding. The stratum overlying the flood-derived sediments consists of~1 m of organic-rich, alluvial mire (Fig. 2). In summary, the mire comprises one of the few near-level, forage-covered watering holes in the upper Guil catchment, located astride the southern route approaching the Col de la Traversette. Thus, it would have provided a prime watering hole, if Hannibal's cavalry and pack animals passed this way en route to the mountain pass.

MATERIALS AND METHODS: THE GEO-RECORD
Two cores [G5(1) and G5(2) ( Fig. 2 (b));~68 cm depth] were recovered from the mire in 2011. The cores provided radiocarbon ages, along with preliminary stratigraphic and geochemical data that, in turn, led to the excavation of an adjacent trench (G5a) to a depth of~75 cm in 2013 ( Fig. 3 (a)). A photograph of the excavated section from 0 to 38 cm ( Fig. 3 (a)) shows that the horizontal bedding of the trench matches that in the macrophotograph of the combined core ( Fig. 3 (b)).
Cores were recovered using 5-cm diameter PVC tubes fitted with a core catcher and were stored cool until opened in the laboratory. Colours were determined using the colour chips of Oyama and Takehara (1970); G5 under dry conditions in the laboratory and G5a in the field. In the laboratory, cores were split and allowed to dry slightly to improve handling prior to sedimentological description, surface penetration, photography and subsampling. Since the two cores were stratigraphically correlated, they were combined into one single core to increase sediment volume for analysis subsection by subsection ( Fig. 3 (b)).
Sediment samples acquired from the trench and the two cores ( Fig. 3 (b)), provided measurements of organic carbon, bulk density and moisture. Subsamples (~1 cm 3 ) were removed with a stainless steel knife at 5-cm intervals. One set of samples was oven dried at 60°C for 24 h, and the moisture content and the wet and dry bulk density calculated based on weight change and subsample dimensions. The total organic matter content was estimated from the basic sequential LOI (550°C/1 h) using a muffle furnace. Elemental analysis was also performed in triplicate on the 2013 G5a samples, using a Fisons NCS 1500 NA elemental analyser.
The mineralogy of the G5a section was investigated by SEM/EDS (for more information on SEM imagery interpretation, see Mahaney 2002) and XRD at various depths to determine composition and variations. The quantitative mineralogical composition was interpreted and modelled by the Rietveld algorithm-based program Siroquant-3 (Taylor 1991).
Pollen was analysed from 10 samples in the GUIL5a sequence (depths 5, 15, 20, 25, 30, 35, 40, 45, 50 and 55 cm). Samples were prepared using standard methods, including digestions in 10 % HCl, 10 % KOH, concentrated HF and acetolysis, and mechanical separation using 10-μ m and 150-μm sieves. Dehydrated residues were stored in silicone oil and mounted for light microscopy. Pollen grains were enumerated under brightfield optics at ×400 magnification, and identified using standard references (Moore et al. 1991). A main pollen sum consisting of trees, shrubs and herbaceous taxa was used to calculate percentages of all taxa. Aquatic pollen types were excluded from the sum; percentages of aquatic taxa were determined on the basis of the main sum.
All 14 C dates are from peat samples in two cores and one trench. These were prepared by accepted standards, stored cold/frozen until dating by conventional or AMS means at the Queen's University Belfast CHRONO dating centre, at Beta Analytic and at the Radiocarbon Laboratory at the University of Tartu, Tartu, Estonia.
Radiocarbon dates were adjusted by 13 C and calibrated using IntCal-13 in OxCal v.4.2.4 (Bronk Ramsey 2009;Reimer et al. 2013), and then further modelled using Bayesian statistical analyses. The first four columns in Table 1 show laboratory numbers, uncalibrated 14 C dates with uncertainties and depths below surface. The two dates highlighted with lighter shading bracket the age of Hannibal's invasion in darker shading. The four columns show the calibrated ages of the radiocarbon dates at one and two standard deviations. The next four columns under 'Modelled' show the Bayesian-derived ages (Bronk Ramsey 2009) at one and two standard deviations. The last column shows the calculated agreement index values for the model (A model ) and overall (A overall ), each of which is at >95% probability. For simplicity, the rows are grouped into pre-invasion beds, post-invasion beds and MAD bed (mass animal deposition).

Stratigraphy
Four stratigraphic subsections were identified. The uppermost and youngest (post-invasion subsection II) extends from 0 to 26 cm depth. We term this the post-invasion section. In the excavated trench, the exposed soil ( Fig. 3 (a)) contains an organic-rich surface horizon that exhibits a variable black colour (10YR 2/1 to 10YR 2/2), with intact fibre and roots present near the surface that exceed the 20 cm depth of peat requirement for Histosols. This unit displays increasing overbank sedimentation and decreasing organic input with depth, culminating in a 5-cm-thick, unweathered, silty sand bed ( Fig. 3 (a)), showing increased bulk density, but with little organic content or moisture ( Fig. 3 (b)). This sand bed is composed mostly of overbank sedimentation that began at AD 1350, or~600 calendar years before present (cal BP), prior to the Little Ice Age (LIA) climatic deterioration starting c. AD 1500. The increase of organic matter above the sand bed upwards to the~15 cm depth records plant growth during the Medieval Warm Period (MWP), prior to the onset of the LIA, and is paralleled by elevated faecal biomarker signals described in Mahaney et al. (2016). Above this, the low increases of organic matter document a slow but punctuated rise in organic carbon throughout the top 10 cm of the profile. Moving downwards, the next post-invasion subsection (I) extends from 26 to 38 cm depth, and is comprised of more highly decomposed soil (Hemist, Soil Survey Staff 1999; Fig. 3 (a)), most probably caused by stream migration away from the site of deposition that caused increased organic/peat accumulation with moderate but variable decomposition. All organic material in this interval exists as finely disseminated peat lacking any fibrous character, except for a few narrow roots. At 26-27 cm, a calibrated radiocarbon age of 1035 ± 48 cal BP (Beta-305008) dates the upper part of this section to within the beginning of the MWP at~1200 cal BP (c. AD 950). This interval is organic-rich with parallel beds and is presumably a product of increased humidity and temperature.
The next subsection (III, or MAD) extends from 38 to 50 cm, and near the top at 42 cm, the bulk density drops to its lowest level in the core, except for surface samples (Fig. 3 (b)). At the same time, the organic content rises to its highest level below surface values, and the sediment colour becomes darker black, with a solid colour expression of 10YR 1/1. This interval consists of an open structure of loosely matted organic fragments mixed with finer decomposed material, suggesting that it has been mixed or churned (bioturbated). This churning is rarely, if ever, seen in bog cores elsewhere in alpine or other areas (for examples, see Mahaney 1990, figs 7.4 and 7.5).
For the lower part of subsection III, there are three calibrated radiocarbon dates-3507 ± 118 cal BP (Ta-3022), 3511 ± 36 cal BP at 50 cm (UBA-24876) and 3811 ± 52 cal BP at 45 cm (UBA-23647)-all of which predate Hannibal's invasion. However, at 40 cm in the upper part of this subsection, there are two calibrated radiocarbon dates, of 2042 ± 46 cal BP (UBA-24875) and 2586 ± 117 cal BP (Ta-3021), that bracket the invasion year of 2168 cal BP (218 BC). Using Bayesian analysis in OxCal v.4.2.4 (IntCal-13 calibration curve), we calculated an age model for the entire stratigraphic sequence (Fig. 4), yielding a modelled age that firmly places the age of the upper portion of subsection III at the time of Hannibal's invasion of Italy in 218 BC (2168 cal BP).
Based on the evidence, we interpret the upper part of this churned-up (bioturbated) layer as most closely representing Hannibal's invasion, which produced the MAD bed. The older and deeper part of this subsection is interpreted to result from human/animal traffic, possibly from activity of Hannibal's army that disturbed and churned pre-existing, stratified organic beds into one massive, non-laminated bed of finely disseminated peat. The increase of organic carbon and moisture, originating from trampled vegetation and dung from animals, provided an organic base supporting the ingress of Clostridia, a prominent endospore-producing bacterial component of the mammalian gut, as discussed among biological contingents in Part II. For the pre-invasion subsection below 50 cm (subsection IV), the stratified base of the core at~68 cm depth yields a calibrated age of 3400 ± 35 cal BP (Beta-305009) that is slightly younger than the three dates above it, thus supporting the interpretation of sediment reworking due to a mass invasion of animals and humans in the overlying MAD bed. The bottom of the core to 80 cm crumbled upon recovery and provided little usable data. The total core depth of 68 cm, comprising variable organic matter content, yields an estimated sedimentation rate of 5 mm yr -1 .

Mineralogy and geochemical analysis
The G5-G5a section data recorded here provide palaeoecological and geochemical data adding to the corpus of information previously reported for the Traversette area (Mahaney et al. 2010b).
Primary minerals follow from the local metabasalt bedrock, with detrital species of Caplagioclase, albite, amphibole, antigorite, opal phytolith and micron/submicron organic materials mostly adhering to primary mineral grains. Opal phytoliths are derived from plants that incorporate silica from the soil, depositing it within the intracellular and extracellular plant structure. As plants decay, silica remains as a weathering-resistant, almost elliptical microscopic mineral (Alexandre et al. 1997).
Because leaching might affect the distribution of organic compounds within the cores, ratios of soluble to insoluble macroelements such as Ca/Ti, Na/Ti, Ca/Al, Na/Al, P/Ti and P/Al were studied to test for translocation. These chemical element ratios show no evidence of leaching through the core or trenched section, although minor fluctuations occurred with organic dilution of the mineral component or high mineral content as in the sandy bed at 26-30 cm depth.
The 20-40 % organic matter staggered through the G5 section ( Fig. 3 (b)), coupled with the 60-80 % mineral content by weight, affords considerable physical stability to the peat beds. This is further indicated by the out-of-phase organic matter to bulk density measurements in Figure 3 (b) that support field observations depicting a stable, semi-dry peatland, wet at the base and becoming progressively drier upwards in the section (see the macrofossil evidence below).

Palaeoecological analysis
Macrofossil analyses of the entire G5 core showed a dominance of herbaceous assemblages, corresponding well with a reconstructed climate record (Neumann 1992) for the Western Alps with near-unvarying temperature/precipitation straddling the time of the Hannibalic invasion. At the base (65-50 cm), the abundance of Typha seeds confirms an open freshwater marsh condition. From 15 cm to the surface, conditions became drier, as suggested by the introduction of Sphagnum spp. followed by ligneous species at the surface of the site.
While palaeoecology studies for Early to Late Holocene time periods are rare in the Southern Alps, the Swiss (Valsecchi and Tinner 2010) document sections, some at high elevations, similar to sections reported here. Coincidentally, these authors report 14 C dates in peat, at 35 cm depth, to 890 ± 60 yr BP (700-900 cal yr), correlating close to MWP dates reported herein. Pollen (Fig. 5) was analysed from 10 samples in the GUIL5a sequence. Pollen grains were abundant and well preserved throughout. The mean pollen sum was 195 grains (min.-max. = 119-380); the mean total number of grains counted was 333 (min.-max. = 218-457). The pollen assemblages confirm the persistence of the alpine mire through the period of record. The assemblages are dominated by pollen of Cyperaceae, Alnus, upland herbaceous taxa (including Poaceae) and needle-leaved trees. The percentage of tree pollen is low through the sequence. Needle-leaved taxa dominate the tree pollen assemblages, and reflect transport from local and regional montane forests at treeline or below. Broad-leaved trees are recorded at low abundance throughout (generally <3%, with a maximum of 8% at depth of 15 cm), reflecting the high-elevation setting of the site and the low local abundance of broad-leaved trees. Shrub pollen is present at moderate abundance, dominated by Alnus types, and probably reflects local mire dynamics. The pollen assemblages broadly support the four zones presented above. The upper section (0-25 cm) is characterized by a minor up-core decrease in needle-leaved tree pollen, which may reflect a shift in the position of the treeline or in forest composition with climatic changes associated with the MWP-to-LIA transition. A modest increase in herbaceous taxa, bryophytes, ferns, fern allies and shrubs also occurs from 25 cm and above, which supports the interpretation from macrofossils and other proxies of a drier mire surface. These changes could also, in part, reflect land-use change and forest clearance, supported by increases in pollen of Chenopodiaceae/Amaranthaceae in the upper samples. The middle portion of the pollen sequence is characterized by moderate abundances of Cyperaceae pollen, corresponding to the lower post-invasion zone. The MAD section (38-50 cm) corresponds to a large peak in Cyperaceae pollen. Below the MAD section, Cyperaceae pollen drops significantly, and the assemblages are dominated again by needle-leaved trees and upland herbs, including Poaceae.
The strongest signals in the pollen record are in the local wetland pollen types, particularly Cyperaceae. The Cyperaceae (sedges) are a large family of plants found in a variety of aquatic habitats, and are common dominants in alpine meadows (Cronk and Fennessy 2001). Cyperaceae pollen may be locally very abundant in sediment records, and thus Cyperaceae pollen Figure 5 A summary of the pollen data for GUIL5b. Pollen types are expressed as percentages of the main pollen sum, which excludes aquatic taxa (Cyperaceae). Needle-leaved trees include Abies, Larix, Picea and Pinus; herbaceous taxa include Poaceae, among other types. The raw data can be accessed by contacting Sarah Finkelstein (finkelstein@es. utoronto.ca) and/or Magda Sobol (magdalena.sobol@mail.utoronto.ca).
abundances are generally interpreted as indicative of local mire processes (Miola et al. 2006). Cyperaceae pollen is not reliably identified to genus or species (Moore et al. 1991); thus, the pollen percentages in the GUIL5a record probably include several genera and species, each of which may vary in terms of ecological preferences. Because of the strong affinity for aquatic habitats, fluctuations in Cyperaceae pollen tend to be interpreted as signals for fluctuations in water level. However, some Cyperaceae are emergent plants in standing water while others are associated with mesic but aerated forest soils (Cronk and Fennessy 2001), making inference of absolute changes in the position of the water table based solely on Cyperaceae pollen difficult. The increase in Cyperaceae pollen in the upper section of the GUIL5a record, coincident with other indicators of drying, probably reflects a shift in species composition of Cyperaceae in response to a drying mire surface.
The spike in Cyperaceae in the MAD bed (38-50 cm) is less likely to be a response to a change in the position of the water table, because of the rapidity with which this section was deposited, based on the radiocarbon age model. Thus, the spike in the Cyperaceae pollen may be interpreted as further support for human passage through this section. Cyperaceae frequently dominate in alpine meadows. While animals tend to prefer grasses to sedges for grazing, sedges will be grazed upon when grasses are not available (Elliott and Henry 2011). Given the evidence for faecal matter in the MAD section (see Part II), it is possible that the Cyperaceae peak in the MAD layer reflects the deposition of pollen grains through animal guts following the passage of large numbers of grazing animals. Pollen grains survive gut passage; coprolites have been used as sources of palynological information in a variety of contexts (Carrion et al. 2001;Kropf et al. 2007).

Transhumance
Transhumance has been practiced for millennia in the Western Alps (Schwörer et al. 2014) and, as Walsh et al. (2007) have shown, human activity (including transhumance) accelerated during the last millennium in the Parc National des Ecrins, Hautes-Alpes, sites located near Mt Viso. Archaeologists are divided over the extent to which environment determines human activity, with some pro (Leveau 2007) and others con (Walsh 2005). Transhumance may be read into the slight increase of faecal biosignatures recorded during the MWP part of the G5-G5a sections, at depths < 30 cm. However, these sediments lack evidence of the kind of major bioturbation observed in the MAD bed. There is nothing to indicate a high level of animal (including human) disturbance in the two sections other than what exists in the MAD bed (~45 cm). Furthermore, the abundance of bacteria-endospores for the most part-sourced specifically from mammals and located primarily within the churned-up beds (see Part II), could only result from the passage of hundreds, if not thousands, of animals, indicating movement of something more than animals involved in tribal trade. CONCLUSIONS Previous analysis of environmental factors has strongly favoured the southern route for Hannibal's passage through the Alps. In addition to features indicated by ancient authors, including the presence of frozen ground and the two-tiered rockfall, there must have been sufficient grounds for grazing large numbers of army animals, including horses and elephants.
The highly anomalous MAD sediments recovered in cores from site G5 and the G5a section in the upper Guil Valley reveal intensive convolution of organic sediment, indicating major disturbance to the beds. Radiocarbon-dating indicates that this disturbance occurred within the Hannibal window. This extensive churning observed in the MAD layer suggests trampling/disaggregation of organic layers by animals, and possibly humans as well, that cannot be accounted for by small-scale transhumance. The MAD layer contains a spike in Cyperaceae pollen, which may be indicative of the passage of large numbers of grazing animals. The beds also feature strong faecal biosignatures and the presence of bacteria from mammals, discussed in Part II.
This may be the first tangible, if unusual, evidence of human/animal activity at the time of Hannibal's invasion of Italia. Although we cannot determine conclusively that the evidence pertains to Hannibal, the results are consistent with the passage of large numbers of animals and people, and because of the high elevation, are inconsistent with normal transhumance in the area. If confirmed, the findings presented here have far-reaching implications for solving the Hannibalic route question and, more importantly, for the identification of a site that might be expected to yield significant historical archaeological data and artefacts related to the Punic invasion. If the site was affected by human-animal traffic, as the evidence indicates, there is every possibility that artefacts such as coins, belt buckles, daggers, equestrian fasteners and so on might have been buried in the mire. If such archaeological evidence can be found and definitively linked to Hannibal, it would answer the question of which route Hannibal and his army took into Italia.