# Geoarchaeology dossier: the Col de la Traversette physical-science case Question: which pass did Hannibal use to cross the Alps in 218 BC — here restricted to the physical evidence (mire deposit, radiocarbon, biomarkers, microbes, rockfall) advanced for the Col de la Traversette by W. C. Mahaney and colleagues, and the published criticism of it. Every quotation below was read in a saved text (section 9 lists the files). Where a statement rests on memory or on a summary seen only through a search tool, it is marked "(unverified)". Page numbers for the two Archaeometry papers are inferred: the copies obtained are the QUB Research Portal early-view PDFs (footer "Archaeometry (2016) --"), and the final pagination (Part I 59(1): 164-178; Part II 59(1): 179-190, from the Crossref records saved as `texts/meta_crossref_arcm1223*.json`) has been mapped onto them as journal page = PDF page + 162 (Part I) and + 177 (Part II). Each such reference is given as "p. NNN (PDF p. N)". Governing rule applied throughout, in both directions: a coarse measurement cannot refute a fine-scale claim, and a radiocarbon date with a range of centuries cannot single out one autumn. Wherever the evidence has a resolution of centuries or millennia, the claim it is said to support or refute is assessed at that resolution and no finer. Bottom line in one paragraph. The Traversette work has found a real, unusual thing: a disturbed, organic-rich, dung-bearing bed in a small floodplain mire at 2580 m on the French side of the col, containing peat of many ages. It has not established that the bed dates to 218 BC (the precision comes from putting 218 BC into the age model as a fixed event; every individual date that the 2017 paper says brackets 218 BC in fact excludes it at 2 sigma, and the one later date that includes it spreads over 379-152 BC), that the dung is from horses rather than cattle, sheep, goats, mules or wild ungulates (the authors' own biomarker criterion points to ruminants; the DNA is class-level and unauthenticated), that the scale was an army rather than herds (the medieval transhumance layer in the same section carries an equal or larger faecal signal), or that the animals were Hannibal's rather than Hasdrubal's (207 BC), a Gallic host's (231/225 BC), a Roman army's, or a mule train's. No artefact has been found. The rockfall on the Italian side exists and predates 218 BC, but the methods used (lichen cover at saturation, soil development) resolve millennia, not the "recent" (prosphatos) break-away Polybius describes, and whether Polybius describes a rockfall at all is disputed. The evidence is therefore consistent with the Traversette hypothesis and does not refute it; it does not confirm it, and it cannot at its present resolution. --- ## 1. Summary table | # | Claim (Mahaney et al.) | Evidence offered | What it establishes | What it does not establish | Assessment | |---|---|---|---|---|---| | 1 | A "mass animal deposition" (MAD) bed exists in the G5 mire below the col | Cores G5, section G5a (2011-2013), sections G5B/C/D (2015): a non-laminated, "churned", low-bulk-density, organic-rich bed at ~38-50 cm (Part I p. 172, PDF p. 10; MAA 2017 pp. 23-35) | A disturbed, mixed bed containing peat of many ages exists in this mire; it is unlike the horizontally bedded peat above and below | The agent of disturbance. The mire sits on the Guil headwater floodplain and "is subject to periodic overbank flooding" (Part I p. 169, PDF p. 7); the alternatives (flood scour and channel fill, frost churning, slumping, burrowing, stream migration) are dismissed in a paragraph each, qualitatively (MAA 2017 p. 32) | Real feature; trampling is one plausible cause among several; no diagnostic test of process is reported | | 2 | The MAD bed dates to 218 BC (2168 cal BP) | Two dates at 40 cm said to "bracket the invasion year" (Part I p. 172); an OxCal model containing the row "Hannibal: 218 BC 2168 0" (Table 1, p. 171, PDF p. 9); later sections G5B/C/D "based on Bayesian statistical calculations" (MAA 2017 p. 29) | In G5 core: the top of the churned bed contains peat formed no earlier than 169 BC (UBA-24875, 2070 +/- 31, 2 sigma 169 BC-AD 8) mixed with peat of 814-407 BC (Ta-3021) and 1950-1400 BC. In G5C: peat at 51 cm of 379-152 BC (UBA-30330). In G5D: peat at 45 cm of 545-204 BC (UBA-30315) | 218 BC. The two "bracketing" dates both exclude 218 BC at 2 sigma in the authors' own IntCal13 table and in IntCal20, and are mutually inconsistent (Ward-Wilson T = 23.4, p = 1e-6). The only measurement whose range includes 218 BC (UBA-30330) puts 17 % of its probability in 230-200 BC and 42 % in 355-282 BC. The modelled 218 BC was read in the first version as a prior returned as a posterior; that reading is withdrawn, see the correction under 2.4 | Not established. The dates are compatible with a late-3rd-century disturbance and equally with a 4th-century, a 2nd-century or a multi-event history; a mixed bed cannot be dated to a year by any number of dates | | 3 | The bed records mass deposition of horse dung | 5-beta-stigmastanol and deoxycholic acid (DCA) concentrations; stanol ratio > 0.7; off-site controls negative (Part II Table 1 p. 183, PDF p. 6; pp. 184-186) | Herbivore faecal input at every depth from 15 to 60 cm, with maxima at 15-25 cm and 35-45 cm; none in the controls | Species: the authors' own criterion, "the presence of DCA in the absence of lithocholic acid (LCA) indicates a ruminant animal source" (p. 183), points away from horses (hindgut fermenters, not ruminants); 5-beta-stigmastanol is a general herbivore marker. Number: concentrations per gram organic carbon are not a census; the MAD-bed values (35-45 cm: 22.7/16.9 stigmastanol, 31.6/15.8 DCA ug/gOC) are of the same order as the 15-25 cm values (26.1/30.6 and 40.4/28.1) which the authors assign to medieval transhumance | Herbivore dung: yes. Horses: not shown; the reported lipid pattern favours ruminants. Army scale: not shown | | 4 | Clostridia 16S rRNA genes = horse gut bacteria | 454 pyrosequencing of nine samples; Clostridia > 12 % at 40-45 cm vs < 2 % above and below; Veillonellaceae 6 %; "horse gut bacterial flora is comprised of +70 % Clostridia" (Part II pp. 187-188, PDF pp. 10-11) | A Clostridia-rich bacterial assemblage at 40-45 cm in one section (G5a) | Gut origin ("we cannot presently be certain that the Clostridia detected are associated with endospore-forming gut bacteria", p. 188); species (Clostridia dominate the gut of every large herbivore and are also anaerobic soil and peat bacteria; the analysis is at class/family level); antiquity (no ancient-DNA authentication: no damage patterns, fragment-length distributions, extraction blanks or contamination controls are reported; class-level 16S cannot tell 2,200-year-old spores from living ones) | Consistent with faecal input; not species-specific, not age-specific. Allen (Conversation, 2016): "we cannot yet be absolutely certain that these bacteria do actually come from horses or humans" | | 5 | Parasite eggs "only found in horses" were recovered | Press statements only: Mahaney in EARTH (2016): "We even found a tapeworm egg that's only found in horses"; Allen in Smithsonian (2016): "what may turn out to be a horse tapeworm egg"; Mahaney 2020 p. 14: "roundworm eggs"; p. 19: "helminth parasite eggs" that DNA has yet to link to any animal | Nothing yet: no peer-reviewed identification, image, taxon, count or depth was found in any paper read | Species, age, or even existence of the eggs as a published result | Unpublished claim. In his own 2016 article Allen framed it as a hope ("We may also be able to find parasite eggs") | | 6 | The mire is the army's camp or watering place below the col | "~425 m below the Col de la Traversette", "~60 x ~40 m", "the only organics-rich peat bed (mire) above 2000 m asl in the Guil Valley" (Part I p. 169, PDF p. 7) | A small (about 0.24 ha) wet meadow at 2580 m on the French side, about 0.9 km west of the col and about 370 m below it (SRTM check, section 3d) | That an army of tens of thousands with 10,000+ animals camped there; that it is Polybius' summit camp (3.53.9-10, two days at the top; 3.54.2, the view of Italy). Mahaney 2020 p. 15 concedes "the Col de la Traversette itself could barely hold 100 soldiers" | A plausible watering stop on the way up; not the Polybian camp. The physical site and the textual site do not coincide | | 7 | The two-tier rockfall below the col on the Italian side is Polybius' landslide | Mahaney et al. 2010 (abstract only); MAA 2014: older Late-Glacial lobe with a Cryochrept palaeosol, younger "middle Neoglacial" lobe with a thin Cryorthent; lichens on both; apron "~240 m" vs Polybius' "300 yards" (p. 82); volume up to 75,000 m3 (EARTH 2012) | A rockfall of two ages exists at ~2600 m below the col; both lobes are older than ~2 ka (lichen cover at saturation; soil development), so it was there in 218 BC | That it is what Polybius describes: 3.54.7 speaks of the path "broken away" (aporrox) for about 1.5 stades, "previously, and then even more, having recently broken away" (prosphatos aperrogyias), and 3.55.6 of Hannibal "building up the cliff" (ton kremnon exoikodomei). Lichenometry saturates beyond ~2 ka and soil ages have +/- 1 ka precision, so neither can address a "recent" event in either direction. Kuhle and Kuhle (2012, abstract) hold that no landslide is needed to read the text; Hunt (EARTH 2016) says such rockfalls are common in the Alps | Existence established; identification is an interpretation; the dating resolution is millennial and cannot test the Polybian claim | | 8 | The "white rock" (leukopetron ochyron, 3.53.5) | No physical evidence offered for the Traversette route in the papers read (Mahaney 2008 book not obtained). Hunt locates it at a gypsum-dolomite anticline near the Bramans gorge on the Clapier route (EARTH 2012) | Nothing physical | Anything: a "strong/defensible white rock" is a textual landmark that many valleys can supply | Not addressed by the Traversette evidence; unverified for every route | | 9 | The Combe du Queyras is the second ambush defile | Mahaney et al. 2010b (abstract): the Isere gorges were forested, making an ambush there "impossible"; the Queyras gorge is narrow | That the Guil has a gorge and that the authors prefer it | Which gorge Hannibal was in; Polybius' distances (the pass reached in a day from the leukopetron, 3.53.9) are not tested against the Queyras in the papers read; de Galbert (2016) reports that General Guillaume (1967) could not place a 1.5-stade break-away on the Traversette descent (unverified beyond de Galbert) | A geographical-plausibility argument, not physical evidence | | 10 | The deposit is Hannibal's | Date "218 BC"; scale ("hundreds, if not thousands, of animals", Part I p. 175); elevation excluding "normal transhumance" (p. 176) | Nothing about identity | Hannibal 218 vs Hasdrubal 207 (Livy 27.39: he led his army "per munita pleraque transitu fratris", through places mostly made passable by his brother's crossing), the Gallic hosts of 231 and 225 BC, Roman armies of the 2nd century, or transhumant herds (which reach 2500 m+; Part I itself reads transhumance into the 15-30 cm peak). "The only artifacts obtained are in the form of microbes and faecal matter" (Mahaney 2020 p. 21) | Not established; identity is asserted, not tested | | 11 | Elephants crossed the Traversette | None physical. Mahaney 2024 p. 61 claims elephant remains on the Arno "have been AMS C14 dated to within the Hannibal time window", citing Romano and Palombo 2017 | Nothing: Romano and Palombo 2017 (abstract read) is a history-of-science study of 16th-19th-century interpretations of fossil elephant bones, not a report of dated Punic elephant remains | Elephants on any route. The only radiocarbon-dated Punic-period elephant bone reported in Europe (Cordoba, Spain; 4th-3rd c. BC) is, in its authors' words, not one of the specimens Hannibal took across the Alps (Smithsonian smart-news, 2025/26) | The 2024 claim is contradicted by its own source; elephants remain physically undocumented on every candidate pass | --- ## 2. Radiocarbon: extraction, calibration, and what the dates can and cannot resolve ### 2.1 The measurements Material and laboratories (Part I p. 170, PDF p. 8): "All 14C 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 13C and calibrated using IntCal-13 in OxCal v.4.2.4". No pretreatment protocol, fraction dated (bulk peat, humic acid, humin, or plant macrofossils) or delta-13C values are given in either Archaeometry paper or in MAA 2017. "Peat samples" read literally means bulk peat. In a bed the authors themselves describe as churned and mixed, a bulk sample is a mixture of carbon of different ages and its date is a weighted average that need not correspond to any event. Fourteen dates were found. Seven are in Part I Table 1 (p. 171, PDF p. 9; G5 core and G5a trench); seven are in Mahaney et al. 2017, Mediterranean Archaeology and Archaeometry 17(1), p. 29 (sections G5B, G5C, G5D, excavated 2015). One laboratory number (UBA-30330) is printed for two different measurements in MAA 2017; the second is given here as "UBA-30330b (sic)". | Lab code | 14C BP | Depth (cm) | Section / bed (authors) | Source | |---|---|---|---|---| | Beta-305008 | 1130 +/- 30 | 26 | G5, post-invasion subsection I top | Part I Table 1 | | UBA-24875 | 2070 +/- 31 | 40 | G5, MAD bed upper part | Part I Table 1 | | Ta-3021 | 2530 +/- 90 | 40 | G5, MAD bed upper part | Part I Table 1 | | UBA-23647 | 3537 +/- 28 | 45 | G5, MAD bed lower part | Part I Table 1 | | Ta-3022 | 3270 +/- 100 | 50 | G5, MAD bed lower part | Part I Table 1 | | UBA-24876 | 3282 +/- 28 | 50 | G5, MAD bed lower part | Part I Table 1 | | Beta-305009 | 3170 +/- 30 | 65 | G5, pre-invasion beds | Part I Table 1 | | UBA-30324 | 1393 +/- 49 | 22 | G5B, above churned zone | MAA 2017 p. 29 | | UBA-30321 | 2678 +/- 41 | 35-37 | G5B, churned zone | MAA 2017 p. 29 | | UBA-30331 | 1842 +/- 41 | 45 | G5C | MAA 2017 p. 29 | | UBA-30330 | 2190 +/- 35 | 51 | G5C ("218 BC bed" placed at 50 cm) | MAA 2017 p. 29 | | UBA-30330b (sic) | 2523 +/- 36 | 58 | G5C | MAA 2017 p. 29 | | UBA-30318 | 1724 +/- 38 | 35 | G5D | MAA 2017 p. 29 | | UBA-30315 | 2329 +/- 55 | 45 | G5D ("218 BC bed" placed at 45 cm) | MAA 2017 p. 29 | ### 2.2 Calibration (IntCal20, own code) Code: `code/calibrate.py` (pure Python/numpy; normal likelihood with combined measurement and curve variance; flat prior; no sequence model; HPD ranges; Ward and Wilson 1978 homogeneity test). Curve: `data/intcal20.14c` (Reimer et al. 2020, downloaded from intcal.org). Full output with every HPD interval: `data/radiocarbon_calibration.json` and `data/radiocarbon_calibration.md`; curve plot for the 40 cm pair: `data/radiocarbon_curve_40cm.svg`. Convention: 218 BC = astronomical year -217 = 2167 cal BP (the authors use 2168; the one-year difference is immaterial). "Window" = 230-200 BC. | Lab code | 14C BP | 68.3 % (IntCal20) | 95.4 % (IntCal20) | Median | P(230-200 BC) | Authors' 2 sigma (IntCal13, cal BP) | |---|---|---|---|---|---|---| | Beta-305008 | 1130 +/- 30 | AD 913-975 (56 %); AD 889-903 (13 %) | AD 875-992 (88 %); AD 830-850 (4 %); AD 776-785 (3 %) | AD 930 | 0 | 1173-962 | | UBA-24875 | 2070 +/- 31 | 149-136 BC (8 %); 112-41 BC (56 %); 9-1 BC (5 %) | 169 BC-AD 8 (96 %) | 81 BC | 0 | 2125-1949 | | Ta-3021 | 2530 +/- 90 | 796-717 BC (22 %); 709-661 BC (14 %); 654-542 BC (33 %) | 814-407 BC (96 %) | 635 BC | 0 | 2769-2357 | | UBA-23647 | 3537 +/- 28 | 1926-1876 BC (38 %); 1843-1822 BC (18 %); 1796-1778 BC (13 %) | 1952-1863 BC (49 %); 1854-1768 BC (46 %) | 1871 BC | 0 | 3895-3720 | | Ta-3022 | 3270 +/- 100 | 1670-1655 BC (3 %); 1637-1430 BC (65 %) | 1774-1372 BC (90 %); 1353-1299 BC (3 %) | 1552 BC | 0 | 3820-3249 | | UBA-24876 | 3282 +/- 28 | 1608-1581 BC (23 %); 1559-1507 BC (45 %) | 1619-1499 BC (95 %) | 1546 BC | 0 | 3573-3450 | | Beta-305009 | 3170 +/- 30 | 1495-1477 BC (21 %); 1456-1419 BC (47 %) | 1503-1398 BC (95 %) | 1447 BC | 0 | 3454-3345 | | UBA-30324 | 1393 +/- 49 | AD 605-666 (69 %) | AD 571-689 (87 %); AD 742-772 (8 %) | AD 643 | 0 | not given | | UBA-30321 | 2678 +/- 41 | 896-872 BC (22 %); 839-802 BC (46 %) | 908-792 BC (96 %) | 837 BC | 0 | not given | | UBA-30331 | 1842 +/- 41 | AD 130-144 (9 %); AD 154-194 (25 %); AD 198-241 (35 %) | AD 117-256 (83 %); AD 283-327 (11 %) | AD 200 | 0 | not given | | UBA-30330 | 2190 +/- 35 | 355-282 BC (42 %); 231-195 BC (20 %); 188-176 BC (6 %) | 379-152 BC (95 %) | 274 BC | 17.2 % (220-215 BC: 3.1 %) | not given | | UBA-30330b | 2523 +/- 36 | 776-748 BC (17 %); 688-666 BC (13 %); 643-566 BC (39 %) | 793-716 BC (28 %); 710-661 BC (18 %); 655-541 BC (49 %) | 643 BC | 0 | not given | | UBA-30318 | 1724 +/- 38 | AD 256-283 (20 %); AD 327-402 (49 %) | AD 247-299 (30 %); AD 304-412 (65 %) | AD 338 | 0 | not given | | UBA-30315 | 2329 +/- 55 | 481-357 BC (57 %); 278-258 BC (6 %); 245-233 BC (4 %) | 545-347 BC (70 %); 316-204 BC (21 %); 732-697 BC (3 %) | 400 BC | 3.6 % (220-215 BC: 0.9 %) | not given | Result in one sentence: of the fourteen dates, twelve exclude 230-200 BC entirely at 2 sigma; one (UBA-30315, G5D 45 cm) touches it with 3.6 % of its probability; one (UBA-30330, G5C 51 cm) includes it with 17 %, while placing 42 % of its probability in 355-282 BC. All seven of the authors' own IntCal13 2-sigma ranges in Part I Table 1 exclude 2168 cal BP. ### 2.3 What the calibration curve allows here IntCal20 values (14C BP): 450 BC 2408; 400 BC 2346; 350 BC 2207; 300 BC 2209; 250 BC 2227; 231 BC 2225; 218 BC 2216 +/- 14; 207 BC 2206; 200 BC 2174; 150 BC 2111; 100 BC 2097; 50 BC 2066; AD 1 2030. Three features matter. (i) The Hallstatt plateau (c. 800-400 BC) ends around 400 BC; below it the curve is usable. (ii) Between about 350 and 250 BC the curve is nearly flat again (2207-2227 BP), so a sample from anywhere in that century, or from the 230s-200s, calibrates to a spread reaching back to c. 380 BC. (iii) After 250 BC the curve falls steeply (about 1.2 14C years per calendar year to 150 BC), which is why UBA-24875 (2070 BP) gets a tight range that ends in the 2nd-1st century BC. Direct test of resolution: a sample that formed in exactly 218 BC and was measured with a good AMS error (+/- 30) would return 2216 +/- 30 BP and calibrate to 380-197 BC at 2 sigma (68 %: 358-345, 315-273, 262-242, 233-203 BC). A sample from 207 BC (Hasdrubal) gives 377-195 BC; one from 231 BC (the Gallic host cited by de Galbert from Polybius 2.22) gives 386-340 and 322-200 BC. The three are indistinguishable. The method's resolution at this point of the curve is roughly 180 years, and no radiocarbon result from this site could ever single out the autumn of 218 BC, or 218 from 207. ### 2.4 The "bracketing" pair and the age model (Part I) Part I p. 172 (PDF p. 10): "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)." Four problems. 1. "Bracket" is not a dating operation. Two dates from the same depth, one too young and one too old, do not date the level between them; they show the level contains carbon of two ages. Ward and Wilson's test on the pair gives a weighted mean of 2119 +/- 29 BP with T = 23.35 against a 5 % critical value of 3.84 (p = 1.3e-6): the two measurements cannot be two estimates of one true age. Either the peat at 40 cm is a mixture of at least two ages (which is what "churned" means) or one measurement is wrong. Note also that the weighted mean 2119 BP calibrates to c. 200-50 BC, still excluding 218 BC. 2. The precision is imported, not measured. Table 1 (p. 171) lists, between UBA-24875 and Ta-3021, a row "Hannibal: 218 BC 2168 0 2168 0": the invasion is entered into the OxCal sequence as an event of known date with zero uncertainty. The "modelled" column then returns that event at 2168 cal BP. The agreement indices quoted (A_model = A_overall = 97.6 %) test whether the measured dates are compatible with the model; they cannot test the fixed event, which has no likelihood to disagree with. The model's own posterior for UBA-24875 (2464-1597 cal BP at 2 sigma) and for the boundaries ("Transition" 2356-997 and 3684-2610 cal BP) shows how little the sequence constrains anything. **Correction.** The first sentence of item 2 and its reading of the "Hannibal: 218 BC" row are withdrawn. A reference date printed in a results table does not show that the date was imposed on the model. Fig. 4 of Part I (p. 172, PDF p. 10), the OxCal v4.2.4 plot of the model, shows the model's elements: eight radiocarbon dates in three groups labelled "Pre-invasion Beds", "MAD bed; Hannibal invasion" and "Post-invasion bed", between four boundaries (three labelled "Transition"), inside a sequence named "Hannibal: Alps G5 and G5A sites". There is no row for a 218 BC element; the 218 BC appears as a dashed vertical line and a dashed box drawn round the two 40 cm dates, with the label "218 BC or 2168 cal BP" in a larger typeface. OxCal's multiple plot draws a row for each element of a model as run unless the user removes it, so the simplest reading is that the table row is a reference entry and the line an annotation. This cannot be confirmed either way: the OxCal input is not printed in Part I, Part II, MAA 2017, Scientific Culture 2020 or 2024, or the 2026 paper, and no supplementary file is deposited with Wiley, QUB Pure, UCL Discovery or Zenodo (attempts logged in `texts/ACCESS_LOG_revision.tsv`). What stands: nothing in the model's printed output places the bed in 218 BC (the modelled ranges quoted above); "firmly places" (p. 172) and the Fig. 4 caption's "high probability of >95%" rest on the agreement index A_model = 97.6 read as a probability, which it is not; and the "Modelled" columns of Table 1 cannot be the posteriors of the dates in the model plotted, since for UBA-23647 the unmodelled 2-sigma range (3895-3720 cal BP) and the modelled range (3509-2109) do not overlap while the agreement index printed beside them is 98.3, and an index near 100 requires the posterior to sit on the likelihood. The plot's row labels carry calibrated means ("2040 ± 45" for UBA-24875, radiocarbon age 2070 ± 31), so plot and table cannot be matched row by row without the input. Conclusion: the dating evidence does not independently establish a disturbance specifically in 218 BC, much less identify the animals responsible as Hannibal's. 3. Both dates individually exclude 218 BC, in the authors' table as well as in IntCal20: UBA-24875 2125-1949 cal BP (176 BC-AD 1), Ta-3021 2769-2357 cal BP (820-408 BC). A conclusion of "firmly ... 218 BC" cannot be drawn from two measurements neither of which includes 218 BC. 4. The stratigraphy is inverted. At 45 cm the peat is older (3537 +/- 28) than at 50 cm (3282 +/- 28, 3270 +/- 100), and 40 cm holds both 2070 and 2530. Every date inside the bed between 40 and 50 cm except one is Bronze Age or Early Iron Age. Whatever churned the bed churned in material 1,200-1,700 years older than the Hannibalic horizon; on the authors' account the deposit is a mixture, and a mixture yields, at best, a terminus post quem for the mixing from its youngest component. The terminus post quem reading. If UBA-24875 is a valid date on peat inside the churned bed, the churning ended no earlier than 169 BC (2 sigma), which is after 218 BC. The 2026 paper sees this and reverses the 2017 reading: the 2070 +/- 31 date is now "one lone AMSC14 age ... [that] marks the first recovered organic material following the invasion at a depth ca. 43 cm", so that "it still took nearly a century for the degraded Hannibal landscape (at ~40-45 cm depth) to recover enough flora to produce carbon and yield a reasonable date", while "AMS-C14 dates older than 2244 yr are reworked" (Mahaney 2026, section on age controls). Both readings cannot hold: in 2017 the 2070 date was in the MAD bed and bracketed 218 BC; in 2026 it is post-event recovery peat a century later, and every older date is reworked. On the 2026 reading there is no date from the event at all, only a floor (Bronze Age reworked material) and a ceiling (recovery after c. 170 BC), and the 218 BC assignment rests on nothing but the historical prior. ### 2.5 The 2015 sections (MAA 2017) G5B: the only dates are 22 cm (AD 571-689) and 35-37 cm (908-792 BC). "The 218 BC bed was determined using Bayesian statistical calculations of the most likely probability of the Hannibal event that occurred between the two dated beds" (p. 29). There is no measurement within 500 years of 218 BC in this section; the "218 BC bed" at ~33 cm is an interpolation between a 7th-century-AD date and a 9th-century-BC date. G5C: 45 cm AD 117-256; 51 cm 379-152 BC; 58 cm 793-541 BC. This is the one section with a date that includes 218 BC (UBA-30330), and its "218 BC bed" is placed at 50 cm, one centimetre above the sample. It is equally a "300 BC bed", a "207 BC bed" or a "170 BC bed". G5D: 35 cm AD 247-412; 45 cm 545-204 BC (70 % of it 545-347 BC). The "218 BC bed" is placed at the 45 cm sample, whose median is 400 BC. The authors note that "the 14C dates within the churned-up bed are age inconsistent as expected within the bioturbated area" (p. 31): an admission that dates from inside the bed cannot date it. Magnetic susceptibility (the paper's subject) shows higher and more variable values in the churned beds; the interpretation offered, weathered iron from discarded implements or dung-enriched organic sediment, is speculative and no metal was found. ### 2.6 What the dates can and cannot resolve Can: they show that the mire has accumulated peat since at least c. 1500 BC; that horizontally bedded, undisturbed peat has accumulated since roughly the 2nd century BC to the 1st century AD (the youngest dates above the bed in each section are AD 117-256, AD 247-412, AD 571-689, AD 875-992); and that the churned bed contains carbon of every age from the Bronze Age to the 2nd century BC. The disturbance therefore ended somewhere between about 380 BC (oldest date that could be event-age) and the 2nd century BC (youngest carbon inside the bed), and different sections need not record the same event. Cannot: distinguish 218 from 207 BC; distinguish either from any year between c. 380 and c. 150 BC; distinguish one event from several; distinguish an army from a century of herds; or date the event at all if the bed is a mixture rather than a single deposit. The claim "dated to 218 BC" has a resolution of one year; the measurement has a resolution of about two centuries; the two do not match, and the mismatch runs in the authors' favour only because the prior was inserted. In the other direction, the same rule protects the hypothesis: nothing in the dates refutes a 218 BC disturbance. A late-3rd-century event is fully compatible with UBA-30330 and UBA-30315 and, on the recovery reading, with UBA-24875. The correct statement is that the dates are consistent with, and do not test, the Hannibalic attribution. --- ## 3. The chain of inference, link by link ### (a) What the MAD bed physically is, and what produces such deposits Description (Part I p. 172, PDF p. 10): "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 ... the organic content rises to its highest level below surface values, and the sediment colour becomes darker black ... 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". Interpretation (p. 173, PDF p. 11): "activity of Hannibal's army that disturbed and churned pre-existing, stratified organic beds into one massive, non-laminated bed of finely disseminated peat." Setting (p. 169, PDF p. 7): "The site is located on a gently sloping bedrock bar composed of metabasalt ... 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". A 5-cm silty sand bed of overbank origin is present at ~15-20 cm (p. 171), and the base of the section is alluvium over bedrock. Processes that produce non-laminated, mixed, organic-rich beds in a high-alpine floodplain mire include: trampling and dunging by any large herbivores (domestic or wild, once or over decades); scour-and-fill during overbank floods, which rip up and redeposit peat clasts of mixed age (a reworked-peat bed dates exactly as this one does: young matrix, old clasts); cryoturbation and needle-ice heave in a site the authors themselves describe as reaching "within the snowline" with "discontinuous frozen soil or firnpack" (p. 167, PDF p. 5); slumping on the "gently sloping" bar; and burrowing (marmots are resident, Part II p. 186). The authors dismiss frost churning because "organic beds act as insulators" and overbank filling because channels are "of restricted depth" (MAA 2017 p. 32) and note that frost churning "is not known to produce convoluted horizons" in the area (Part II p. 188). No thin-section micromorphology, no grain-size or peat-humification profile through the bed, no test of the peat fragments' individual ages (which would separate reworked clasts from in-situ growth) is reported. Trampling is a reasonable hypothesis; it is not the demonstrated one. The bed is not unique to one place in the mire: it appears in G5, G5a, G5B, G5C and G5D, at depths whose upper surface ranges "from 30 cm in G5B, 47 cm in G5C and 37 cm in G5D" (MAA 2017 p. 31), with dates inside it ranging from the 10th century BC to the 2nd century BC. Two more sites to the south "produced inconclusive results indicating the ecologically disturbed beds are likely deeper than ~1 m" (Mahaney 2020 p. 18, PDF p. 12). ### (b) Biomarkers and microbial evidence: species and number Lipids (Part II pp. 183-186). Compounds: 5-beta-stigmastanol (24-ethylcoprostanol) and deoxycholic acid; stanol ratio 5b-stigmastanol/(5b-stigmastanol + 5a-stigmastanol) above 0.7 at all depths but two (25-30 and 50-55 cm: 0.68 and 0.65). Table 1 (p. 183) values (ug per g organic carbon), depth: stigmastanol / DCA: 15-20 cm 26.11 / 40.44; 20-25 cm 30.59 / 28.05; 25-30 cm 10.45 / 11.92; 30-35 cm 10.62 / 6.47; 35-40 cm 22.66 / 31.55; 40-45 cm 16.96 / 15.83; 45-50 cm 13.12 / 1.95; 50-55 cm 4.62 / 4.23; 55-60 cm 4.29 / 3.62. The authors' reading (p. 184): the 20-30 cm maximum "could be due to increased animal faecal matter input coinciding with the MWP", making the site "more accessible for transhumance", while at 35-40 cm, "the churned-up layer correlating to the Hannibalic event, both of these faecal biomarkers observe a surge in concentration". The "surge" is relative to the 25-35 cm trough; in absolute terms the MAD-bed values equal or fall below the medieval-transhumance values. Whatever the medieval peak represents (herds over centuries, on the authors' reading), the Hannibalic peak is not larger, so the biomarker magnitude does not by itself separate "army in one autumn" from "flocks over generations". Species. The authors write (p. 183): "5b-stigmastanol is a 5b-stanol commonly used as a biomarker for faecal material produced by ruminant mammals (Evershed et al. 1997; Bull et al. 2002). It was also the most abundant stanol observed in horse faeces by Leeming et al. (1996). DCA is a major secondary bile acid in the faeces of humans and some higher animals ... The presence of DCA in the absence of lithocholic acid (LCA) indicates a ruminant animal source for the faecal material (Bull et al. 2002)." The paper then reports DCA without LCA. By the criterion the authors themselves cite, the bile-acid pattern indicates ruminants (cattle, sheep, goats, ibex, chamois), not horses. The same page (p. 186) says: "The Capra ibex, Rupicapra rupicapra and Marmota marmota that are native to the region are ruminants, and therefore would contribute similar faecal signatures to those identified" (marmots are rodents, not ruminants; the error is the authors'). The published lipid data thus do not discriminate a horse signal from a ruminant signal, and if anything point to ruminants. Fiedler et al. 2025 (review of faecal lipid markers, `texts/fiedler2025_...`) describe 5-beta-stanol profiles as serving "for distinguishing between omnivorous and herbivorous" sources; discrimination among herbivore species requires bile-acid profiles (as the authors cite) or DNA. Microbes (Part II pp. 187-188). Nine samples, 454 pyrosequencing of 16S rRNA amplicons, 1,876-4,357 sequences per sample (Table 1). "Of particular note is a maximal abundance of 16S rRNA genes from the Clostridia class in the 40-45 cm sample of the MAD bed ... comprising more than 12 % of the total bacterial flora ... Previous studies suggest that Clostridia are rarely found in soil above 2-3 % of the total eubacterial population (Jansen 2006)" (p. 187). "Typically, horse gut bacterial flora is comprised of +70 % Clostridia (Costa and Weese 2012). The Veillonellaceae family also constitutes 6 % of all 16S rRNA gene sequences in this sample ... While we cannot presently be certain that the Clostridia detected are associated with endospore-forming gut bacteria, the detection of increased levels of bile acids and faecal lipid biomarkers in the same layer provides compelling evidence in favour of this hypothesis" (p. 188). What is missing for a species claim: (1) Clostridia dominate the gut microbiota of horses but also of cattle, sheep, goats, deer, pigs and humans (the paper's own references, Ley et al. 2006, are about mammals generally); a class-level abundance identifies no host. (2) Clostridia are obligate anaerobes and are a normal component of waterlogged, anoxic peat; the 2-3 % benchmark cited (Jansen 2006) is for aerated soil, not mire. (3) No ancient-DNA authentication is reported: no extraction blanks, no damage-pattern or fragment-length analysis, no replication in a second laboratory, so nothing shows that the sequences are 2,200 years old rather than from the living anaerobic community of that horizon. (4) One section only was sequenced (G5a); the 2020 and 2026 papers state that the Clostridia spike recurs in other sections ("common depths of 35-60 cm in all samples running up to 2017 (Young, 2018)", Mahaney 2026), but no data have been published. (5) Host-specific identification was feasible: lake-sediment sedaDNA metabarcoding with mammal primers (Giguet-Covex et al. 2019, `texts/giguetcovex2019_srep_sedDNA.txt`) identifies Bos, Ovis and Canis at genus level in alpine sediments of the last millennia; no such analysis, and no host-specific Bacteroidales or mitochondrial marker, has been reported for the Traversette mire. Allen announced in April 2016 "an extensive microbiology programme to try and assemble either complete or partial Clostridia genomes from the samples"; no publication from it was found (section 6). Number. Neither lipids per gram of carbon nor 16S proportions are counts of animals. The 2017 conclusion that the bacteria "could only result from the passage of hundreds, if not thousands, of animals" (Part I p. 175, PDF p. 13) is an assertion; the equal medieval signal in the same section (which the authors attribute to transhumance) shows the signal can be produced without an army. ### (c) Tapeworm and parasite claims No parasitological result has been published in any of the papers read (Archaeometry Parts I and II, MAA 2014, 2017, Scientific Culture 2020, 2024, Journals Press 2026). The claims exist only in press and in passing: Mahaney to EARTH (2016): "We even found a tapeworm egg that's only found in horses"; Allen to Smithsonian (April 2016): "Already the team has uncovered what may turn out to be a horse tapeworm egg from the manure ... 'There is even the possibility of finding an elephant tapeworm egg,' he says. 'This would really be the pot of gold at the end of the rainbow.'"; Allen in The Conversation (April 2016): "We may also be able to find parasite eggs - associated with gut tapeworms - still preserved in the site"; Mahaney 2020 p. 14: biomarkers "related to horse and mule manure with roundworm eggs and bacteria", and p. 19: "It remains for DNA analysis to link firmly recovered microbes and helminth parasite eggs with Spanish and/or North African animals." Tapeworm in 2016 has become roundworm by 2020; no taxon, image, count, depth, or method appears anywhere. The horse tapeworms (Anoplocephala spp.) shed eggs that are distinctive, but so do the tapeworms of cattle and sheep (Moniezia), and identification to host requires the egg to be seen and described, which has not been published. Assessment: unpublished; treat as no evidence. ### (d) The mire's location relative to the col, and its size against an army Coordinates from the papers: G5/G5a 44 deg 42.587' N, 7 deg 03.274' E, 2580 m (Part I p. 169); G5B 44 deg 42.591' N, 7 deg 03.281' E, 2579 m; G5C 44 deg 42.590' N, 7 deg 03.256' E, 2579 m (MAA 2017 pp. 28-29). Col de la Traversette: 44.71089 N, 7.06635 E (OSM node 293467280, in the project's `data/places.json`); an SRTM30 check (opentopodata) returns 2943 m there, matching the col's map height of 2947 m, and 2571 m at the mire. The mire is therefore about 0.94 km west of the col and about 370 m below it, on the French (Guil) side; the authors say "~425 m below" (Part I p. 169), presumably against a rounded 3000 m for the col. Size: "~60 x ~40 m" (p. 169), about 0.24 ha, "one of the few near-level, forage-covered watering holes in the upper Guil catchment". Against Polybius: the army spent two days at the summit of the pass (3.53.9-10) and Hannibal showed the men Italy from the summit at dawn (3.54.1-3). A 0.24 ha meadow cannot hold the army (Polybius 3.56.4 gives 20,000 foot and 6,000 horse arriving in Italy; the number that reached the pass was larger) and it is not at the summit. Mahaney's own concession (2020 p. 15, PDF p. 9): "he could not have amassed the entire army on any of the cols, and the Col de la Traversette itself could barely hold 100 soldiers, so he must have addressed his key lieutenants not the entire army." Hunt (EARTH 2016): "the Traversette has no [area that could accommodate a] summit encampment. It's essentially a knife-edge up there. The camp [Mahaney] proposes is more than 300 meters below the summit." Mahaney's reply in the same piece, that "none of these passes have open areas where one might camp, not even the Clapier", is a point about all candidates, and it means the mire is at best a watering stop on the approach, not the camp Polybius describes. Elephants and horses on the approach: the mire is on a route that the authors describe as passable; its existence says nothing about the descent. ### (e) The rockfall against Polybius 3.54.5-7 and Livy 21.36-37; lichenometry and soil-age precision The text. Polybius 3.54.7 (Greek, `texts/polybius_3_greek.txt`): "... oute tois hypozygiois dynaton en parelthein dia ten stenoteta, schedon epi tri' hemistadia tes aporrogos kai pro tou men ouses, tote de kai mallon eti prosphatos aperrogyias" -- "it was impossible for either the elephants or the pack-animals to pass owing to the extreme narrowness of the path, a previous landslip having carried away about one and a half stades of the face of the mountain and a further landslip having recently occurred" (translation in `texts/polybius_3_english.txt`). 3.55.6: "ton kremnon exoikodomei meta polles talaiporias" -- he "began constructing a road on the face of the precipice" (Shuckburgh) / built up the cliff with great labour. 3.55.7-8: one day's work made a path for horses and pack animals; three days for the elephants. Livy 21.36.1-37.4 has the rupes, the fired rock and the vinegar. What Polybius describes is a stretch of path along a precipice, roughly 280 m long (1.5 stades), which had broken away before and had recently broken away further, so that the path had to be rebuilt (exoikodomein: to build up, build out). This is a narrow-ledge failure, not necessarily a debris sheet lying across a slope; Kuhle and Kuhle 2012 (abstract, full text not obtained) argue exactly that "a critical reassessment of the ancient texts and a closer look at the geomorphological situation clearly show that this interpretation is unwarranted", and in 2015 (abstract) that "the random selection and investigation of topographical or sedimentological features ... informed only by free translations of the Greek originals, is a misguided enterprise". Mahaney's 2013 reply (abstract) does not engage the Greek; it describes the critique as "a tirade". The deposit. Mahaney et al. 2010 (abstract): "the only such two-tier landform lies below the Col de la Traversette, at ~2600 m above sea level ... A reconstruction of the initial deposit, likely Late Glacial ... Further specifications on the geometry of the Neoglacial cover sediment are based on weathering characteristics, lichen cover and soil development. The 'myth' that Hannibal fired the rockfall ... is unsubstantiated by the lack of any carbonized rock." MAA 2014 (pp. 78-84): "the older mass of Late Glacial age lying astride the younger mass originating during the Neoglacial" (p. 78); "The rockfall (cf. landslide described by Polybius) was said to measure 300 yards, approximately equivalent to 275 m, which is close to the actual size near the source bedrock where the broad apron of debris measures ~240 m" (p. 82); the younger lobe carries "a thinner Cryorthent which is largely lacking in significant vegetative cover", boulders "with similar lichen cover" in both lobes (p. 82), weathering-rind populations at sites V8 (Neoglacial) and V9 (Late Glacial), and a "middle Neoglacial age" for the younger soil (p. 84). EARTH 2012 (Mahaney speaking): "a volume of up to 75,000 cubic meters spread out across 250 meters". The "300 yards" is Polybius' 1.5 stades read as the width of the debris apron. Precision. Rhizocarpon-type lichen cover at "longevity" (the authors' term for lichens that have reached their maximum size and stopped growing) can only say "older than the lichen's lifespan", i.e. more than roughly 2,000 years in this setting: it is a floor, not a date, and it applies equally to a deposit of 3000 BC and one of 300 BC. Soil-development ages (Cryorthent vs Cryochrept, rind thickness) are relative and calibrated against regional chronosequences with precisions of the order of +/- 1,000 years. "Middle Neoglacial" (c. 2.5-3.5 ka) is compatible with anything from c. 1500 BC to c. 500 BC. These methods therefore cannot show that the younger lobe fell "recently" before 218 BC, and equally cannot show it did not: the Polybian claim is at the scale of years, the measurement at the scale of millennia. The rule cuts both ways: the coarse date does not refute the Traversette rockfall as the Polybian site, and it cannot confirm it. Uniqueness. Hunt (EARTH 2016): "such rockfalls are common throughout the Alps and found at several other potential crossings, including the more ...[Clapier-Savine Coche]". Burned rock exists on the Clapier descent, which Mahaney's own Raman work put at brush-fire temperatures and, on weathering and lichen grounds, at "100 or 200 years ago" (EARTH 2012) -- the same lichen argument, used there to make a deposit young. Hoyte's 1959 attempt to take the elephant Jumbo over the Clapier was turned back by a rockfall on the Italian side and finished over Mont Cenis (unverified; secondary accounts only). ### (f) The "white rock" (leukopetron ochyron, Polybius 3.53.5) Greek: "... anagkasthenai ton Annibam meta tes hemiseias dynameos nyktereusai peri ti leukopetron ochyron choris ton hippon kai ton hypozygion" -- "Hannibal was compelled to pass the night with half his force at a certain place defended by bare rocks" (`texts/polybius_3_english.txt`; Shuckburgh: "near a certain white rock"). The word is a compound of leukos (white/bright) and petra (rock), qualified as ochyron (strong, defensible); whether it is a proper name (Leukopetron) or a description is a philological question, not a geological one. None of the Traversette papers read offers a candidate; Mahaney's 2008 book (not obtained) may. Hunt's candidate for the Clapier route is "a massive, 600-meter-tall exfoliating anticline made of gypsum and dolomite, which rises up just east of the Bramans Gorge ... 'It stands out in this area as brilliant white'" (EARTH 2012). Gypsum outcrops are common in the Brianconnais and Vanoise zones (unverified); so is bare pale rock on most approach routes. Assessment: not a discriminating physical criterion, and the Traversette evidence does not address it. ### (g) The Guil valley / Combe du Queyras as the second ambush The claim rests on Mahaney et al. 2010b (abstract only): "New data from palaeobotanical reconstruction of the northern approach route show the Gorges de la Bourne and the Gorge du Breda, astride the Isere River, to have been forest covered during the invasion, which would have made the Gallic assault impossible. The existing evidence argues for a southern route, the approach through the narrow defile of the Combe de Queyras". Two objections apply. First, forest cover does not make an ambush from cliffs impossible (Polybius 3.53 describes the attack on the column in the gorge from the heights); the argument is asserted, not tested. Second, the Queyras gorge is one narrow defile among many; Hunt names "the Gateway to the Alps, just outside of Grenoble, where two great massifs squeeze the Isere" and the Bramans gorge (EARTH 2012). De Galbert (2016) reports that General Guillaume, who examined the Traversette route in 1967, could not locate on its descent the break-away of three half-stades that Polybius describes (unverified beyond de Galbert's summary). Nothing physical from the mire bears on the ambush site. ### (h) Even if genuine, the deposit is evidence only if it is Hannibal's Candidates that the dating cannot separate from 218 BC: Hasdrubal's crossing in 207 BC, which Livy 27.39 (`texts/livy_27_latin_latinlibrary.txt`; sub-section numbers not marked in that text) says followed his brother's route -- "cum per munita pleraque transitu fratris quae antea inuia fuerant ducebat, tum etiam duodecim annorum adsuetudine peruiis Alpibus factis inter mitiora iam transibant hominum ingenia" (he led through places mostly made passable by his brother's crossing, that had before been impassable; and the Alps had been made passable by twelve years of use); the Gallic host of 231 BC (Polybius 2.22 as cited by de Galbert; Book 2 is not among the saved texts, so unverified here) and the Gaesatae of 225 BC (unverified); Roman armies in Gaul from 125 BC (de Galbert lists Fulvius Flaccus 125 and Marius, with Domitius Ahenobarbus and Fabius Maximus 122-121 BC implied; unverified list); and, over the whole 380-150 BC envelope, any number of trading and transhumant movements. Livy's remark that the Alps were now "only a road" (uiam tantum Alpes esse, 27.39) within twelve years is itself evidence that the passes carried regular traffic in the late 3rd century. On artefacts, Mahaney is explicit: "While permits were obtained to carry out the geological and biostratigraphic work, no attempt was made to obtain archaeological input beyond what is outlined in various papers. The only artifacts obtained are in the form of microbes and faecal matter obtained from sections dug in sediment piles in various places and from cores taken in similar settings" (2020 p. 21, PDF p. 15). Part I (p. 176, PDF p. 14) had promised that "artefacts such as coins, belt buckles, daggers, equestrian fasteners and so on might have been buried in the mire"; the 2017 season with National Geographic funding produced none that has been reported. Known finds at the candidate cols, as far as could be established (all from secondary sources; to be verified against the scholarship dossier): Col du Petit-Saint-Bernard: Roman road station and temple/fanum, coins of the imperial period, and an Iron Age stone circle at the summit (Gallia; seen only as a search summary, unverified). Col de Montgenevre: the Via Domitia from c. 118 BC; the Cottian kingdom's road (unverified). Col du Mont-Cenis: Roman road of the imperial period; Hoyte's 1959 elephant crossing (unverified). Col du Clapier and Col de la Traversette: no reported Iron Age or Punic finds; de Galbert (2016) states that no Neolithic or Iron Age remains are known between the Durance and the Traversette (his claim; unverified). The Buco di Viso tunnel (1479-80) below the Traversette was cut because mules could not cross the col with loads (de Galbert; the tunnel is a documented fact, its motive as stated is unverified here). None of the candidate cols has any artefact that ties it to the Punic army; the Traversette is not worse off than the others in this respect, and no better. ### (i) Elephants over the Traversette; exposure at 2947 m in early November Polybius gives one day to make the path passable for horses and pack animals and three days for the elephants (3.55.7-8), the elephants "in a wretched condition from hunger" (3.55.8), and says the summit region is treeless and snow-covered all year (3.55.9), which fits any col above the treeline (c. 2,300-2,500 m in the Cottian Alps; unverified figure). At 2947 m in early November (the setting of the Pleiades, 3.54.1, i.e. late October / early November by the pre-Julian calendar; computed in `dossiers/terrain.md`) fresh snow on old snow is what Polybius describes (3.55.1-3), and the Traversette is the only candidate above the firn line; Part I (p. 167) makes this a criterion ("the only pass that reaches to within the snowline (~2800 m asl)"). But Polybius' snow is a hardship that killed animals, not a criterion the army chose, and it is not clear that the text requires permanent snow rather than the first autumn snowfall. On feasibility for elephants there is no physical evidence either way. Berti and Vollrath 2026 (PNAS, `texts/berti_vollrath2026_pnas_energy_costs.xml`) compute energy costs and find "the Col de la Traversette would have been the shortest and energetically most efficient route, with a total cost for the whole army of 5.42 TJ", the Montgenevre, Clapier and Mont-Cenis routes costing 11 %, 16 % and 19 % more; the model is a route-length-and-gradient calculation that does not evaluate the mire evidence, the descent's technical difficulty, or the rockfall. Lazenby, as reported by Mahaney (2024 p. 65), "considered the Traversette too difficult for Hannibal and his elephants to negotiate". Mahaney (2020 p. 16) argues from elephants working "above 4500 m" on Mount Kenya. The Cordoba elephant bone (Martinez Sanchez et al., JAS Reports; radiocarbon 4th-3rd c. BC; species unconfirmed; found under collapsed adobe walls with catapult shot) is the only dated Punic-period elephant bone reported from Europe and has nothing to do with the Alps. The 2024 claim of AMS-dated Hannibalic elephant remains on the Arno is wrong (section 1, row 11). --- ## 4. Reading notes ### 4.1 Mahaney, W. C., Allen, C. C. R., Pentlavalli, P., Kulakova, A., Young, J. M., Dirszowsky, R. W., West, A., Kelleher, B., Jordan, S., Pulleyblank, C., O'Reilly, S., Murphy, B. T., Lasberg, K., Somelar, P., Garneau, M., Finkelstein, S. A., Sobol, M. K., Kalm, V., Costa, P. J. M., Hancock, R. G. V., Hart, K. M., Tricart, P., Barendregt, R. W., Bunch, T. E. and Milner, M. W. 2017. Biostratigraphic evidence relating to the age-old question of Hannibal's invasion of Italy, I: History and geological reconstruction. Archaeometry 59(1): 164-178. doi:10.1111/arcm.12231. Text: `texts/mahaney2017_archaeometry_partI_arcm12231.txt` (QUB early-view PDF). - p. 164-165 (PDF 2-3): abstract and framing; the mire "potentially identifies the invasion route as the one originally proposed by Sir Gavin de Beer"; the MAD bed "was produced by the passage of thousands of animals and humans". - p. 166-167 (PDF 4-5): the rockfall and frozen ground as "vital clues"; the firing "never happened"; "The Col de la Traversette ... is the only pass that reaches to within the snowline (~2800 m asl)". - p. 169 (PDF 7): site description quoted in 3(a) and 3(d). - p. 170 (PDF 8): methods; "All 14C dates are from peat samples in two cores and one trench". - p. 171 (PDF 9): Table 1 with the "Hannibal: 218 BC 2168 0 2168 0" row; "The two dates highlighted with lighter shading bracket the age of Hannibal's invasion in darker shading." - p. 172 (PDF 10): stratigraphy; the bracketing paragraph quoted in 2.4. - p. 173 (PDF 11): "activity of Hannibal's army that disturbed and churned pre-existing, stratified organic beds into one massive, non-laminated bed"; the older, deeper part "interpreted to result from human/animal traffic, possibly from" earlier movements. - p. 174-175 (PDF 12-13): pollen (Cyperaceae spike in the MAD bed; "Cyperaceae pollen is not reliably identified to genus or species"); macrofossils; transhumance: "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"; "the abundance of bacteria -- endospores for the most part -- sourced specifically from mammals ... could only result from the passage of hundreds, if not thousands, of animals". - p. 175-176 (PDF 13-14): conclusions: "Radiocarbon-dating indicates that this disturbance occurred within the Hannibal window"; "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 hope of "coins, belt buckles, daggers, equestrian fasteners". Note the gap between the paper's own hedged conclusion ("cannot determine conclusively ... consistent with ... if confirmed") and the press framing (QUB release: "finally provided solid evidence"; Allen: "it looks like we may just have cracked it"). ### 4.2 Mahaney, W. C., Allen, C. C. R., Pentlavalli, P., Kulakova, A., Young, J. M., Dirszowsky, R. W., West, A., Kelleher, B., Jordan, S., Pulleyblank, C., O'Reilly, S., Murphy, B. T., Lasberg, K., Somelar, P., Garneau, M., Finkelstein, S. A., Sobol, M. K., Kalm, V., Costa, P. J. M., Hancock, R. G. V., Hart, K. M., Tricart, P., Barendregt, R. W., Bunch, T. E. and Milner, M. W. 2017. Biostratigraphic evidence relating to the age-old question of Hannibal's invasion of Italy, II: Chemical biomarkers and microbial signatures. Archaeometry 59(1): 179-190. doi:10.1111/arcm.12228. Text: `texts/mahaney2017_archaeometry_partII_arcm12228.txt`. - p. 179-180 (PDF 2-3): abstract: "The presence of deoxycholic acid and ethylcoprostanol derived from faecal matter, coupled with high relative numbers of Clostridia" in the MAD bed. - p. 182-183 (PDF 5-6): methods (GC-MS of stanols and bile acids; 454 pyrosequencing; QIIME); nine samples at 5-cm intervals, 15-60 cm; stanol ratio; the ruminant/DCA-LCA sentences quoted in 3(b); Table 1 (values in 3(b)). - p. 184 (PDF 7): profile interpretation; MWP transhumance at 20-30 cm; the "surge" at 35-40 cm; "it would appear that the MAD bed is the layer most significantly contaminated with faecal material in the G5a soil profile" (on the basis of the stanol ratio, not concentration); possibility of "anaerobic microbial hydrogenation of higher plant sterols ... which could lead to the false assignment of a faecal source" noted and set aside. - p. 186 (PDF 9): controls negative; leaching considered unlikely; the ibex/chamois/marmot "ruminants" sentence. - p. 187 (PDF 10): Clostridia > 12 % at 40-45 cm; "surprising"; Jansen 2006 benchmark. - p. 188 (PDF 11): "we cannot presently be certain that the Clostridia detected are associated with endospore-forming gut bacteria"; conclusions: "provides some of the first direct evidence of the route followed by the Punic Army ... the most direct evidence yet uncovered to date for the passage of Hannibal's army". ### 4.3 Mahaney, W. C., Somelar, P., Pulleyblank, C., Tricart, P., West, A., Young, J. M. and Allen, C. C. R. 2017. Notes on magnetic susceptibility in the Guil valley alluvial mire correlated with the Punic invasion of Italia in 218 BC. Mediterranean Archaeology and Archaeometry 17(1): 23-35. doi:10.5281/zenodo.258081. Text: `texts/mahaney2017_MAA17_magnetic_susceptibility_zenodo258081.txt`. - p. 23: abstract: "all dated by AMS 14C to 2168 cal yr BP (i.e., 218 BC ...)". - p. 28-29: section coordinates and elevations; the seven new dates; the "218 BC bed" placed by "Bayesian statistical calculations" between dates centuries apart (quoted in 2.5). - p. 31: "the 14C dates within the churned-up bed are age inconsistent as expected within the bioturbated area"; MS higher and more variable in the churned beds. - p. 32: alternative hypotheses (overbank flooding, frost heave) considered and dismissed in one paragraph. ### 4.4 Mahaney, W. C., Allen, C. C. R., Pentlavalli, P., Dirszowsky, R., Tricart, P., Keiser, L., Somelar, P., Kelleher, B., Murphy, B., Costa, P. J. M., Jordan, S., O'Reilly, S. and Juligj, P. [sic, as printed] 2014. Polybius' 'previous landslide': proof that Hannibal's invasion route crossed the Col de la Traversette. Mediterranean Archaeology and Archaeometry 14(2): 77-94. doi:10.5281/zenodo.14744. Text: `texts/mahaney2014_MAA14_previous_landslide_zenodo14744.txt`. - p. 77-78: "the Traversette debris is nearly as Polybius described it"; response to Kuhle and Kuhle 2012; "The two-tier nature of the rockfall is clearly evident in the field, the older mass of Late Glacial age lying astride the younger mass originating during the Neoglacial." - p. 82: dimensions ("300 yards ... ~240 m"), lobes, lichen cover "similar" on both, rind sites V8/V9. - p. 83-84: rind statistics; "middle Neoglacial age" of the younger soil; particle size "characteristic of heterogeneous material produced by large scale mass wasting events". - The title says "proof"; the body dates the deposit to a millennial bracket and identifies it with the text by size alone. ### 4.5 Mahaney, W. C. 2020. The Hannibal enigma of 218 BC: a forensic exercise of importance to historical archaeology. Scientific Culture 6(3): 7-24. doi:10.5281/zenodo.4007566. Text: `texts/mahaney2020_hannibal_enigma_zenodo4007566.txt`. - p. 10: the route question "led to embedded historical hypotheses lacking any evidence providing an outcome with little chance of finding artifacts". - p. 12: "Up to the time of Mahaney (2004, 2008a) no one attempted to target key sites where historical archaeological exploration might reveal artifacts related to the invasion of Hannibal or the later crossing of his brother, Hasdrubal, in 207 BC." - p. 14: "horse and mule manure with roundworm eggs and bacteria, the latter belonging to microbe groups of Firmacuties [sic] and Clostridia related to gut release from horses and mules". - p. 15: "the Col de la Traversette itself could barely hold 100 soldiers". - p. 16: elephants: Hoyte; Mount Kenya; "The remains of Hannibal's elephants have been found in the Arno terraces near Florence (Romano and Palombo, 2017)" -- wrong, see row 11. - p. 18: two further sites to the south "produced inconclusive results". - p. 19: "It remains for DNA analysis to link firmly recovered microbes and helminth parasite eggs with Spanish and/or North African animals." - p. 21: "The only artifacts obtained are in the form of microbes and faecal matter". - p. 21: acknowledgements: National Geographic Society grant 9988-16; Lion TV (London) and ZDF filming, July 2017; "the 2015 and 2017 field teams". ### 4.6 Mahaney, W. C. 2024. Polybius: decoder of Hannibal's alpine invasion route, author of a stratigraphic assessment of the famous blocking rockfall. Scientific Culture 10(1): 59-67. doi:10.5281/zenodo.8297673. Text: `texts/mahaney2024_scientific_culture_polybius_zenodo8297673.txt`. - p. 61: "all 37 of Hannibal's elephants made it over the Alps"; "His and other remains were unearthed along the Arno River near Florence by Italian paleontologists (Romano and Palombo, 2017) and have been AMS C14 dated to within the Hannibal time window." Contradicted by the cited paper's abstract. - p. 62: Figure 2 caption cites "Fink et al., 2024, in progress" (Dijkstra path-planning paper, "Nature, in progress", ref. p. 66); no such publication found (Crossref, Semantic Scholar, OpenAlex searches; section 6). - p. 65: Lazenby "considered the Traversette too difficult for Hannibal and his elephants to negotiate". - p. 66: permits from the Prefet des Hautes-Alpes for excavation on the French side (Reserve Naturelle Ristolas-Mont Viso). ### 4.7 Mahaney, W. C. 2026. Futile route quest of the Hannibal invasion: merging of classical history and surface geology. London Journal of Research in Humanities and Social Sciences (Great Britain Journals Press), doi:10.34257/ljrhss229004uk (from Unpaywall record `texts/meta/upw_10_34257_ljrhss229004uk.json`). Text: `texts/mahaney2026_futile_route_quest_journalspress.txt`. - Age-control section (quoted in 2.4): "one lone AMSC14 age of 2070 +/-31 cal yr BP marks the first recovered organic material following the invasion"; "The thin laminae below the 2070 +/-31 date carries an AMSC14 age of 2530 +/-90 which is positively reworked organic sediment imported to the section from older beds higher in the valley"; "it still took nearly a century for the degraded Hannibal landscape (at ~40-45 cm depth) to recover"; "AMS-C14 dates older than 2244 yr are reworked from soldiers and animals degrading and disrupting surface sediments dating into the Early Holocene". - Microbes: "the twin bacteria Firmacuties and Clostridia ... spiked within the MAD group and adjacent beds, with common depths of 35-60 cm in all samples running up to 2017 (Young, 2018)" -- data not published. - No new fieldwork after 2017 is mentioned; the 2019 upper-Po sections (V21 group) are cited from Mahaney et al. 2019. - The paper is a venue of unknown standing (Journals Press, London); treat as the author's latest statement, not as peer-reviewed confirmation. ### 4.8 Abstract-only items (full texts not obtained; `texts/abstracts_paywalled_papers_openalex.txt`) - Mahaney, W. C. 2008. Hannibal and the Alps: unravelling the invasion route. Geology Today 24(6). Abstract lists the ancient sources' col order and concludes for the Traversette "either by design or happenstance". - Mahaney, W. C., Milner, M. W., Kalm, V., Dirszowsky, R. W., Hancock, R. G. V. and Beukens, R. P. 2010. The Traversette (Italia) rockfall: geomorphological indicator of the Hannibalic invasion route. Archaeometry 52(1): 156-172. Abstract quoted in 3(e). - Mahaney, W. C. et al. 2010. Hannibal's invasion route: an age-old question revisited within a geoarchaeological and palaeobotanical context. Archaeometry 52(6). Abstract quoted in 3(g). - Kuhle, M. and Kuhle, S. 2012. Hannibal gone astray? A critical comment on W. C. Mahaney et al. ... Archaeometry 54(3). Abstract quoted in 3(e). - Mahaney, W. C. 2013. Comments on M. Kuhle and S. Kuhle (2012) ... Archaeometry 55(6): 1196-1204. Abstract: "appears to be mostly a tirade against the Traversette Route". - Kuhle, M. and Kuhle, S. 2015 (online 2014). Lost in translation, or can we still understand what Polybius says about Hannibal's crossing of the Alps? A reply to Mahaney. Archaeometry 57(4). Abstract: Perrin (1887) and the Col de Clapier "concurs with all the descriptive details as given in Polybius' Histories". - Mahaney, W. C. et al. 2019 (online 2018). Reconnaissance of the Hannibalic route in the upper Po valley, Italy: correlation with biostratigraphic historical archaeological evidence in the upper Guil valley, France. Archaeometry 61(1). Abstract: "compelling stratigraphic, geochemical and microbiological evidence ... potentially identifies the invasion route"; coalescing alluvial fans in the upper Po "calibrated by radiocarbon dating". - Mahaney, W. C. 2023. Polybius: decipherer of Hannibal's alpine route and ancient stratigrapher. International Journal of Earth Sciences. No abstract retrievable. - Romano, M. and Palombo, M. R. 2017. When legend, history and science rhyme: Hannibal's war elephants as an explanation to large vertebrate skeletons found in Italy. Historical Biology. Abstract: a study of why 16th-19th-century authors attributed fossil bones to Hannibal's elephants, "critically examined and confuted" by Targioni Tozzetti. ### 4.9 Press and web items (saved) - Queen's University Belfast press release, April 2016 (`texts/qub_news2016_hannibal_press_release.txt`): "Until now, no solid archaeological evidence has been forthcoming. However ... have finally provided solid evidence"; "directly dated to approximately 2168 cal yr BP, i.e. 218 BC"; Allen: "We found scientifically significant evidence of these same bugs in a genetic microbial signature precisely dating to the time of the Punic invasion." - Allen, C. 2016. How ancient horse-dung bacteria is helping our team locate where Hannibal crossed the Alps. The Conversation, April 2016 (`texts/allen2016_conversation_horse_dung_bacteria.txt`; the Guardian reprint was not obtainable). Key sentences: "a mass animal deposition of faecal materials -- probably from horses"; "directly dated to around 200BC through carbon isotope analysis (very close to the date on historical records - 218BC)"; "we cannot yet be absolutely certain that these bacteria do actually come from horses or humans. The gene analysis needs to be expanded"; "We may also be able to find parasite eggs". - Smithsonian smart-news, April 2016 (`texts/smithsonian2016_horse_poop_hannibal_route.txt`): "what may turn out to be a horse tapeworm egg"; "There is even the possibility of finding an elephant tapeworm egg ... the pot of gold at the end of the rainbow." - EARTH Magazine 2016, On the trail of Hannibal's army - and elephants - in the Alps (`texts/earth2016_trail_hannibals_army_elephants_alps.txt`): Mahaney's "tapeworm egg that's only found in horses"; Hunt's "knife-edge", "more than 300 meters below the summit", "such rockfalls are common", "none of the evidence is identifiable uniquely to Hannibal's army ... age dating has wide error margins", "When they find elephant dung, I'll be more persuaded they have the right route."; Mahaney: "An army of nuns could walk right down off that pass [Clapier] into Italy". - EARTH Magazine, 5 January 2012, Erin Wayman, On Hannibal's trail: the clues are in the geology (`texts/earth_hannibals_trail_clues_geology.txt`): rockfall volume; burned rock at the Clapier and its young age by lichen and weathering; Hunt's Bramans anticline as the white rock; the Grenoble "Gateway to the Alps" ambush site. - Smithsonian smart-news, 2025/26 (date not on the saved page; `texts/smithsonian_2200yr_bone_hannibal_elephant.txt`): the Cordoba elephant bone. - de Galbert, G. 2016. La datation par C14 de crotin de cheval trouve au col de la Traversette ne peut pas prouver qu'Hannibal a traverse les Alpes par ce col. LinkedIn, 2 May 2016 (`texts/degalbert2016_linkedin_C14_critique.txt`). OxCal gives "deux intervalles de temps: l'un de 231 a 97 av. J.-C. avec une probabilite de 66 %; l'autre de 355 a 292 av. J.-C avec une probabilite de 30 %" (his own calibration of a "218 BC" input; consistent with this dossier's 380-197 BC); "au moins 5 armees ont traverse les Alpes" in 231-97 BC; the Buco di Viso; General Guillaume 1967. - Berti, E. and Vollrath, F. 2026. Energy costs of Hannibal's crossing. PNAS (`texts/berti_vollrath2026_pnas_energy_costs.xml`): quoted in 3(i). --- ## 5. Criticism: who has said what - Kuhle, M. and Kuhle, S. (Gottingen; geographer and classicist) 2012, 2015, Archaeometry: the two-tier rockfall is a misreading of Polybius; the Col de Clapier fits all the Polybian details (following Perrin 1887); fieldwork "informed only by free translations of the Greek originals" is "misguided". Full texts not obtained; abstracts only. - Hunt, P. (Stanford; Alpine Archaeology Project, Clapier-Savine Coche): no summit camp at the Traversette; two-tier rockfalls common; nothing in the evidence is unique to Hannibal; dating error margins wide; wants elephant dung (EARTH 2016). His book Hannibal (2017) not obtained. - de Galbert, G. (author of Hannibal et Cesar dans les Alpes; Isere valley route) 2016: the radiocarbon window admits at least five armies; Sir Gavin de Beer's thesis contradicts the sources; the Traversette descent does not fit the 1.5-stade break-away (Guillaume 1967); mules could not cross before the 1480 tunnel. - MacDonald, E. (Cardiff), The Conversation, April 2016, "No, we still don't know where Hannibal crossed the Alps" (unverified: seen only in a Brewminate summary; the original was not fetched). - Lazenby, J. F. (as reported by Mahaney 2024 p. 65): Traversette too difficult for elephants; favoured Mont Cenis or Clapier. - Other web commentary: Tour de Travoy blog (the 2070 BP date is c. 200 years too young for Hannibal; medieval salt traffic) and M. Harrsch's blog (unverified; not saved). - Supportive: Berti and Vollrath 2026 (energetics favour the Traversette; no engagement with the mire evidence); the QUB press office; Allen 2016. - Notably absent: no published re-analysis of the radiocarbon model, no published critique from a radiocarbon or ancient-DNA specialist, and no independent re-sampling of the mire were found. The dating critique in this dossier (section 2) appears, apart from de Galbert's short note, to be new. --- ## 6. "Nothing has happened since 2017" -- checked - 2017: a further field season under National Geographic Society grant 9988-16, filmed by Lion TV and ZDF in July 2017 (Mahaney 2020 p. 21). Excavation permits on the French side (Mahaney 2024 p. 66). - 2018/2019: Mahaney et al., Archaeometry 61(1), upper Po valley sections (V21 group), "churned-up bed" at 40-55 cm, dates "~2200 yr" (abstract and Mahaney 2026 summary; full text not obtained). - 2020, 2024, 2026: review and restatement papers by Mahaney alone; the 2026 paper re-reads the 2070 BP date as post-event recovery; no new dates, sequences, or artefacts are reported in any of them. - Allen's 2016 genome programme: no publication found (Crossref, Semantic Scholar, OpenAlex, Europe PMC searches on Allen + Hannibal/Traversette/Clostridia). - Parasite eggs: never published. - "Fink et al. 2024, Nature, in progress" (Dijkstra optimal-path paper): not found in any index. - Two additional mire sites: "inconclusive" (Mahaney 2020 p. 18). - No independent group has re-sampled or re-dated the mire, as far as could be found. - New but unrelated: the Cordoba elephant bone (Spain, 4th-3rd c. BC); Berti and Vollrath 2026 (modelling only). The claim "nothing has happened since" is therefore nearly true in the sense that matters: the evidence base for the Traversette has not grown since the 2017 papers, and the one substantive change is the authors' own re-interpretation of the key date. --- ## 7. Assessment - Resolution, forward: the radiocarbon evidence has a resolution of about two centuries at this point of the curve; the claim ("218 BC", "precisely dating") has a resolution of one year. The precision is supplied by the prior. Downgrade every "dated to 218 BC" statement to "contains material of the 4th-2nd centuries BC". - Resolution, backward: the same coarseness means the dates cannot refute the Traversette. Nothing here shows Hannibal did not cross there. - Species: the authors' own criteria point to ruminants; "horse" is not demonstrated. Downgrade "horse manure" to "herbivore dung". - Scale: an equal signal at the medieval level shows the magnitude is reachable without an army. Downgrade "mass animal deposition ... thousands" to "faecal input comparable to that of the transhumance layer". - Identity: no artefact; several armies and centuries of herds fit the envelope. "Hannibal's" is an attribution, not a finding. - The rockfall: exists, predates 218 BC, but the Polybian passage may describe a broken path rather than a debris sheet, and the dating methods cannot address "recently". - Credit: the Mahaney team did real fieldwork at 2600 m, published their raw dates, and wrote a hedged conclusion in Part I; the overclaim lies in the age model, the press releases and the later single-author papers. Kuhle and Kuhle, Hunt and de Galbert raised most of the substantive objections in 2012-2016; the radiocarbon arithmetic in section 2 has not, as far as found, been published by anyone. Overall: the Traversette physical case is consistent with the hypothesis and does not test it. Its evidential weight for choosing among the candidate passes is small, and it should not be reported as "solid evidence", "direct evidence" or a date of 218 BC. --- ## 8. Access log Obtained (full text): Archaeometry Part I and Part II (QUB Research Portal PDFs, 200); MAA 2014 and MAA 2017 (Zenodo, 200 on retry after a 504); Scientific Culture 2020 and 2024 (Zenodo, 200); Journals Press 2026 (HTML, 200); Berti and Vollrath 2026 (PNAS JATS XML via Europe PMC/Unpaywall route, 200); Fiedler et al. 2025 (Frontiers PDF, 200); Giguet-Covex et al. 2019 (Sci. Rep. PDF, 200); Livy 27 (thelatinlibrary.com, 200); QUB press release (qub.ac.uk via curl, 200; WebFetch 403); de Galbert LinkedIn article (200); The Conversation (Allen 2016, 200); Smithsonian smart-news 2016 and the elephant-bone piece (200); EARTH 2012 and 2016 (200); IntCal20 curve (intcal.org, 200); Crossref, Semantic Scholar (per-DOI), OpenAlex, Unpaywall, Europe PMC, Zenodo API metadata (200). Failed or refused (URL -> result): - https://onlinelibrary.wiley.com/doi/pdfdirect/10.1111/arcm.12231 and .12228 -> 403 (Wiley); the same for 10.1111/j.1475-4754.2009.00463.x (rockfall 2010), 10.1111/j.1475-4754.2010.00526.x (2010b), 10.1111/j.1475-4754.2011.00633.x (Kuhle and Kuhle 2012), 10.1111/j.1475-4754.2012.00724.x (Mahaney 2013), 10.1111/arcm.12115 (Kuhle and Kuhle 2015), 10.1111/arcm.12405 (Mahaney et al. 2019), 10.1111/j.1365-2451.2008.00695.x (Geology Today 2008) -> 403 / paywall; abstracts taken from OpenAlex. - https://link.springer.com/article/10.1007/s00531-023-02336-z (Mahaney 2023 IJES) -> paywall; no abstract in OpenAlex. - https://web.archive.org/web/2024id_/https://doi.org/10.1111/arcm.12231 -> redirected to an unrelated DOI; Wayback availability API -> 429 (repeated); abandoned. - https://www.researchgate.net/... and https://www.academia.edu/... (Mahaney papers, Hunt) -> 403. - https://www.deepdyve.com/... -> 403. - https://www.maajournal.com/ (MAA 16(2) 2016, pp. 97-105, Mahaney et al. 2016 "Hannibal's ... biostratigraphy" and the 2018 MAA paper) -> 403 for both. - https://pure.qub.ac.uk/en/publications/... (2019 Archaeometry record) -> 200 but no file; QUB Pure search -> 403. - https://api.semanticscholar.org/graph/v1/paper/search -> 429; per-DOI lookups worked. - https://nominatim.openstreetmap.org/ -> empty/blocked responses (col coordinates then taken from the project's `data/places.json` and checked with opentopodata). - https://hal.science/... (Walsh et al. 2023, Alpine landscape) -> HTML landing pages, no PDF. - https://www.googleapis.com/books/v1/volumes?q=... (Hunt 2017, Mahaney 2008) -> no usable result. - https://theconversation.com/hannibals-crossing-of-the-alps-the-new-evidence-that-reveals-the-route-he-took-57130 (guessed slug) -> 200 but an unrelated article; correct URL found by search and fetched. - https://www.smithsonianmag.com/smart-news/microbes-and-ancient-manure-may-mark-hannibals-route-across-alps-180962784/ (guessed) -> unrelated article, discarded; https://www.smithsonianmag.com/history/how-hannibal-crossed-the-alps-180963671/ -> 403 (Cloudflare challenge). - https://www.earthmagazine.org/article/hannibals-elephants-and-alps-geoarchaeology-may-reveal-route/ (guessed) -> 404; correct URLs found and fetched. - https://www.theguardian.com/science/2016/apr/04/... and .../2016/apr/05/hannibal-crossed-alps-horse-dung-col-de-traversette (guessed) -> 404; theguardian.com is blocked to the search tool; the Guardian reprint of Allen's piece was not obtained (the Conversation original was). - Brewminate summary of MacDonald 2016 and the Gallia (Petit-Saint-Bernard) article -> seen only through the search tool's summaries; pages not saved. Not obtained at all: Mahaney 2008, Hannibal's Odyssey (book); Hunt 2017, Hannibal (book); Kuhle and Kuhle 2012 and 2015 (full texts); Mahaney 2010a, 2010b, 2013, 2019, 2023 (full texts); Mahaney et al. 2016 and 2018 (MAA); Mahaney 2008 (MAA 8(2)); Walsh et al. 2023; de Beer 1955/1967/1969; Perrin 1887; Guillaume 1967; Young 2018 (thesis cited by Mahaney 2026); any parasitological or genomic report from the Traversette mire (none appears to exist). --- ## 9. Files Files added for this dossier: - `dossiers/geoarchaeology.md` (this file) - `code/calibrate.py` (IntCal20 calibration, HPD, Ward-Wilson test, SVG plot; run: `python3 code/calibrate.py data/intcal20.14c data`) - `data/intcal20.14c`, `data/radiocarbon_calibration.json`, `data/radiocarbon_calibration.md`, `data/radiocarbon_curve_40cm.svg` - `texts/mahaney2017_archaeometry_partI_arcm12231.txt`, `texts/mahaney2017_archaeometry_partII_arcm12228.txt`, `texts/mahaney2017_MAA17_magnetic_susceptibility_zenodo258081.txt`, `texts/mahaney2014_MAA14_previous_landslide_zenodo14744.txt`, `texts/mahaney2020_hannibal_enigma_zenodo4007566.txt`, `texts/mahaney2024_scientific_culture_polybius_zenodo8297673.txt`, `texts/mahaney2026_futile_route_quest_journalspress.txt`, `texts/pdf/` (the six PDFs) - `texts/abstracts_paywalled_papers_openalex.txt`, `texts/berti_vollrath2026_pnas_energy_costs.xml`, `texts/fiedler2025_frontiers_fecal_lipid_markers.txt`, `texts/giguetcovex2019_srep_sedDNA.txt`, `texts/livy_27_latin_latinlibrary.txt`, `texts/degalbert2016_linkedin_C14_critique.txt` (+ .html), `texts/qub_news2016_hannibal_press_release.txt` (+ .html), `texts/allen2016_conversation_horse_dung_bacteria.txt`, `texts/smithsonian2016_horse_poop_hannibal_route.txt`, `texts/earth2016_trail_hannibals_army_elephants_alps.txt`, `texts/earth_hannibals_trail_clues_geology.txt`, `texts/smithsonian_2200yr_bone_hannibal_elephant.txt` - `texts/meta_crossref_arcm12231.json`, `texts/meta_crossref_arcm12228.json`, `texts/meta_s2_arcm12231.json`, `texts/meta_s2_arcm12228.json`, `texts/meta/` (OpenAlex, Semantic Scholar, Unpaywall, Crossref, Zenodo, Europe PMC records; SRTM profile) Used from the other dossiers: `texts/polybius_3_greek.txt`, `texts/polybius_3_english.txt`, `texts/polybius_3_shuckburgh.txt`, `texts/livy_21_latin.txt`, `data/places.json`.