Obsidian evidence and least-cost path scenarios in the Taş Tepeler region

This map accompanies From distant volcanoes: obsidian and early Neolithic connections on the Şanlıurfa plateau. The register covers 24 selected localities with reported Pre-Pottery Neolithic components. It is not an exhaustive survey. Nine have independently verified locations and are mapped; the other 15 remain in the evidence register and Table S1. A site’s Neolithic occupation does not date every surface obsidian find. Local site markers record reported obsidian presence; their source-attribution status is stated separately.

The model pairs each of three source zones with six local endpoints, giving 18 local paths, and adds four regional comparator paths. Nine local paths are principal and nine supplementary; Taşlı and Hamzan remain illustrative. Figure 2 displays 14 paths: seven principal, three supplementary and four regional comparator paths. The principal Nemrut Dağ–Karahan and Nemrut Dağ–Göbekli paths are not drawn; both are reported in Tables 2 and S2. The comparison covers only the three selected source zones. The lines neither identify sources nor reconstruct journeys.

Evidence and model coverageEvidence selection · 12 September 202624 retained cases · 9 verified occurrence locations9 local markers · 4 comparators · 22 calculated model paths
Archaeological evidence and conditional spatial scenarios
Each line is the deterministic least-cost path returned for a stated spatial test. Line style distinguishes regional comparator scenarios, preliminary local source-zone tests and unsourced source–site tests. A line is not an archaeological route and does not demonstrate direct procurement, exchange, pilgrimage, redistribution or repeated use.

Dataset view

Shows 9 verified local occurrence locations, three representative source endpoints and four regional comparators. All 24 retained local cases, including 15 unmapped cases, are in the evidence table. All 22 calculated scenarios retain the original model endpoints and cost assumptions.
Separates published source assignments at comparator sites, Harbetsuvan's published grouped comparison and reported analyses without auditable datasets at Göbekli Tepe, Gürcütepe and Tell Abu Hureyra.
Shows all 22 scenarios on a shared zoom-8/downsample-4 grid. Dijkstra's algorithm uses a direction-sensitive Tobler-style function, terrain modifiers and a 5 × 5 movement neighbourhood. The output is deterministic and comparative; it does not recover journeys.
Removes modelled paths so that sites and source landscapes can be read against modern waterways and terrain. Figure 2 does not model river travel. Hydrology supplies geographical context only.
Open a site record to read its evidence, chronology, source-attribution status, location checks and limitations. All 24 cases are listed, including the 15 without independently verified occurrence locations. Modelling is recorded separately.

24 retained local cases; 9 mapped. Model paths are conditional scenarios.

Map filters

Filter the archaeological evidence categories and model paths separately. A modelled path does not grade the archaeological evidence.

Four archaeological evidence states
Shows all four archaeological evidence states, representative source endpoints and the separate unsourced source–site test layer.
Published analytical comparisons support source assignments for the sampled assemblage. The completeness of individual measurements, assignments and contexts is assessed separately. A displayed path remains modelled.
Harbetsuvan’s published group-level major-oxide results permit preliminary comparison with broad source zones. Wavelength-dispersive X-ray fluorescence (WDXRF) measures the glass’s main chemical components, reported here as oxide percentages. Results for trenches or sample groups cannot establish individual sources or source diversity.
Conference reports, posters and later published mentions document analyses or source claims at Göbekli Tepe, Gürcütepe and Tell Abu Hureyra. They do not supply the contextual artefact-level results needed for independent checking.
Obsidian is reported, with context and chronological limitations stated for each case. Presence does not establish a specific source, travelled path or transfer mechanism.
Separate spatial model layer
Displays deterministic least-cost paths from representative source endpoints to unsourced local sites. Endpoint choice is illustrative and does not imply source presence or preference.

Showing all four archaeological evidence states, representative source endpoints and the separate model layer.

Model assumptions, distances and data

Evidence selection: 12 September 2026. The source-check records give the dates and scope of subsequent verification. Model: all 22 base paths and 54 principal sensitivity runs use full grid-path lengths plus both endpoint connectors. The supplement records the grids, endpoints, cost assumptions and reconstruction checks.

Modelling compares the geographical implications of candidate source–site relationships. A source zone is an area that may contain several obsidian outcrops or lava flows. Its representative endpoint does not identify a quarry or collection place. Bingöl/Solhan is a geographical label; it does not combine the chemically distinct Bingöl A and B sources.

All 22 base scenarios use Dijkstra search over modern Mapzen terrain data: a shared zoom-8 Web Mercator raster downsampled by 4, a 0.9° search margin and a 5 × 5 movement neighbourhood. All base and sensitivity paths were calculated from source to site; reverse-direction paths were not calculated. Grid-step distances are derived from latitude/longitude cell centres rather than projected map distances. The 24 permitted offsets include longer moves costed through intervening cells. Tobler-style walking time is multiplied by terrain factors for altitude and steepness, ruggedness, topographic position and flow accumulation; adjusted step costs are summed along the path.

River crossings are not modelled explicitly. High-flow cells receive a finite terrain-derived penalty, but mapped river lines do not enter the cost calculation. There is no river or lake mask and no representation of channel width, depth or crossing points. A path crossing a mapped waterway does not establish that the crossing was feasible. The supplementary methods give the exact formulas and implementation settings.

Path length and cost measure different things. Reported lengths sum every elementary grid step used to calculate cost, including intermediate cells within longer moves, plus connectors to the exact original endpoints. Both connectors are included even when very short; they are reported separately and do not contribute to movement cost. Lines are simplified only for display. The composite cost index is a comparative measure, not journey duration. Neither metric reconstructs travel over the ground.

The nine principal source–site paths concern Harbetsuvan, Karahan and Göbekli. They form six contrasts: Bingöl/Solhan and Nemrut Dağ against Göllü Dağ/Kömürcü at each of the three sites. Their paths are 7.0–16.6% longer than straight-line distances between the same endpoints. Bingöl/Solhan is already closer to all three sites in a straight line; these results do not isolate the effect of terrain. Its modelled paths are 40.1%, 42.8% and 37.6% shorter, respectively, than the Göllü Dağ/Kömürcü alternatives. Nemrut paths are 28.0%, 30.8% and 15.8% shorter at Harbetsuvan, Karahan and Göbekli respectively. These percentages use unrounded total analytical lengths, including both endpoint connectors.

View all 22 calculated paths (Table S2)

Components and totals are rounded independently, so displayed components may not sum exactly. Percentages use unrounded totals. Relative cost sets Göllü Dağ/Kömürcü–Harbetsuvan to 100. It is a cost comparison under the same assumptions, not a percentage of journey time.

{{PRINCIPAL_TABLE}}

Download Table S2 (Excel) · Full values and definitions (JSON)

Each of six sensitivity variants was applied to all nine principal paths (54 runs). Across the tested changes in grid resolution, search extent and terrain weights, the nine principal path lengths change by −3.08% to +4.72%; eastern paths remain shorter than the central alternatives. The 3 × 3 and 5 × 5 movement settings yield identical expanded paths and lengths for these nine cases, so they do not provide an independent test of geometrical robustness. This does not demonstrate source use, preference or stable travel corridors.

Location checks: the verified positions of Harbetsuvan, Karahan, Göbekli and Sefer remain in the original 1.93 km grid cells. The calculated cell paths and costs are unchanged. Updating only the short endpoint connectors changes total analytical lengths by less than 0.06% and the six principal source contrasts by less than 0.02 percentage points. The Karahan–Nemrut contrast rounds to 30.7% under that substitution, rather than 30.8%. This map retains the original endpoints and reported values.

Placement within source zones remains untested. Taşlı and Hamzan use illustrative points near reported localities whose exact locations remain unverified; all six scenarios at these two endpoints are excluded from the nine principal source–site paths. The model uses present terrain and drainage. It does not reconstruct Early Holocene vegetation, palaeohydrology, seasonality, snow, soils, social constraints, boat travel or repeated use.

  • Complete matrix and supplementary methods
  • All 22 base and 54 sensitivity results (CSV)
  • All 22 complete paths (JSON)
  • Complete matrix calculation code
  • Site-coordinate check (JSON)
  • Inputs for all 22 paths and six sensitivity variants
  • Evidence selection record
  • Supplementary Table S1
  • Comparator transcription and reporting checks
  • Figure 1 PNG
  • Figure 1 PDF
  • Figure 1 SVG
  • Figure 2 PNG
  • Figure 2 PDF
  • Figure 2 SVG
  • Terrain model used by the complete-matrix calculation
  • Check the supplied results, tables and map
  • Map build source
  • These files accompany the submission. The public interactive map may differ from the supplied copy.
  • Supplementary file guide and package details
  • Reproduction instructions
  • Build manifest
About this map

This map makes the paper’s evidence and modelling assumptions available for inspection. Taş Tepeler is a geographical label for this research; it does not mean that the sites were contemporary or formed a single social unit.

The selected register is not an exhaustive review of every obsidian occurrence. It includes published or conference-reported obsidian evidence at 24 localities with reported Pre-Pottery Neolithic components. Çakmaktepe is excluded because the report reviewed describes its chipped stone as entirely flint (Şahin and Miyake 2025: 9). The linked selection record includes a reconstructed query list, not a complete search log recorded at the time.

  • Evidence state, source-attribution status, spatial treatment and social interpretation are reported separately.
  • Least-cost paths compare stated source–site tests; they are not recovered journeys.
  • Modern waterways provide geographical context; Figure 2 does not model river transport.
Contribute obsidian data

This evidence register is designed to be revised. Readers with relevant obsidian evidence for Taş Tepeler, the Upper Tigris, the Middle Euphrates or related source landscapes are invited to contact the authors with bibliographic references, analytical data, site coordinates where available and excavation or museum context.

Submissions will be evaluated before any future update. Contact: d.schwartzman@sussex.ac.uk.

Selected feature

Map feature

Sources, data and code

Sources for the evidence summary, regional comparison, least-cost path method, base geography and these supplementary materials.

Local evidence register

The full 24-case local register distinguishes reported evidence, site chronology, the date of the obsidian, source attribution and location status. Regional comparators remain a separate map layer. See Supplementary Table S1 for the full evidence register.