The Charun–Buni red belt

The view looks north-north-west across the Mastuj River, from above the ridge south of Charun Oweer. I geocoded all 24 village labels in the screenshot from OpenStreetMap and used them to fit a perspective camera. The fit places the viewpoint over 36.11°N 72.22°E, heading about 349°.

Village labels all sit on the valley floor, so they fix direction well but distance poorly. An independent render from that camera does not reproduce the foreground (see Methods). The location is therefore fixed from the red bands themselves. In map view they are unmistakable (§2). Your view spans the valley from Sholiri Gohkir (72.107°E) to Pachili Awi Lusht (72.300°E). It runs from the Mulkhow villages in the north (36.33°N) to the ridges south of the red belt (≈36.16°N).

Karl Stauffer's 1975 USGS reconnaissance of the central Mastuj valley covered exactly this ground. He mapped the Reshun Formation as "massive, light-brown, ridge-forming conglomerate and beds of red siltstone". He traced it along the left bank from Reshun to Buni and noted that its upper and lower limits are north-west-dipping thrust faults. Terraces fed by these siltstones carry soil that he called always bright red.

A redness index on the Sentinel-2 12-scene composite reproduces that belt in map view. It follows the same bank, the same ENE strike and the same end points.

An unsupervised clustering of seven illumination-robust Sentinel-2 features splits the exposed rock into seven colour units. I then described each unit with two independent sensors: ASTER thermal emissivity (quartz and carbonate) and the minerals EMIT identified on the same ground.

EMIT columns: share of unit pixels where the mineral family was identified on ≥3 of 5 dates. QI = quartz index B11²/(B10·B12), higher = more quartz. CI = carbonate index B13/B14, higher = more carbonate.

The units agree across sensors. The red core is 96% ferric by EMIT. The pale units have the highest carbonate index. The grey unit has the highest quartz index (slates are quartz-rich) and the most chlorite and epidote. This is ordinary low-grade metamorphism, which the KP Directorate's report puts at no higher than greenschist facies for Chitral.

Terrain here is arid and bare. After masking cloud, shadow, snow, vegetation and water, and applying a 0.02 reflectance floor, 68% of the grid is visible rock. At PL-53 in Upper Dir the figure was 11%. A null here therefore carries weight.

The method has range in this scene. The raw ferric index sits at a median robust z of +2.0 inside the red belt and −0.1 outside it. To stop the red beds swamping everything, I searched for anomalies within each colour unit. A pixel counted when it was anomalous in at least two of four indices (ferric B4/B2, gossan B11/B4, Al-OH/carbonate B11/B12, 900 nm Fe³⁺).

That first pass flagged 78 clusters. Almost all sat above 4,500 m on the Buni Zom flank, a median 350 m from snow. I then normalised by elevation band as well as unit, and the high-altitude cluster vanished completely. What remains at z > 3 is 8 specks of 0.3–2 ha, mostly outside your view. None coincides with an EMIT hit. Frost-shattered fresh rock, weathering rinds and mixed snow/debris pixels explain the first pass better than sulphide oxidation does.

Five EMIT L2B mineral-ID dates cover the area (Apr 2023 to Oct 2024), with retrievals on 93–100% of pixels. At PL-53 only 1 of 15 granules retrieved anything. For each mineral family I counted how many dates it was identified in each cell. I compared that with the count expected if identifications were random.

E1, at 36.2090°N 72.2700°E and 3,590–3,835 m, is about 9 cells of jarosite/copiapite, identified on up to 4 of 5 dates. It sits about 1 km from the red-belt margin. Sentinel-2 shows nothing there: the ferric and Al-OH z-scores are both about 0.3–0.5. Jarosite also forms on ordinary pyritic black shale, and there is plenty of that in the Chitral Slate and Sarikol Shale.

E2, near Buni at 36.262°N 72.289°E, is 23 cells matched to azurite, scattered over about 1 km². Azurite's SWIR features resemble carbonate, and the surrounding rock is carbonate-rich. Treat it as unconfirmed.

Five pre-2008 AST_07XT scenes passed QC, with no clipped bands and 10–72% usable after the same masking. They give Al-OH, kaolinite/alunite and Mg-OH/carbonate indices. They match the unit picture above and add no independent alteration zone. There is no phyllic (sericite) or advanced-argillic (kaolinite/alunite) halo of the kind a porphyry would leave.

EnMAP has no archive scene over the core area (DLR STAC, 0 results). EnMAP and PRISMA can both be tasked at 30 m, and a summer EnMAP acquisition is the realistic upgrade over EMIT. Any hyperspectral sensor is still optical and still limited by snow and shadow.

Sentinel-1 radar sees through cloud and at night, but in 4 km of relief the viewing geometry matters. I modelled layover and shadow from the DEM separately for each orbit. Orbit 173 ascending and orbit 5 descending each observe 92% of the grid cleanly, and 97% is clean in at least one of them. Orbit 100 clips the edge (23%) and was set aside. Following earlier Karakoram work, backscatter is compared only within an orbit, never across orbits.

Radar does not detect ore minerals. It measures surface roughness and moisture, and here it confirms the colour units from an independent physical property. The same pattern appears in both geometries, so it is not a look-angle artefact:

The red belt returns 1.1–1.5 dB less than its surroundings: the soft siltstone breaks down to fine debris. Stauffer described exactly this weakness. The siltstones acted as a lubricant along the thrusts. He mapped a landslide about 1 mile wide and 3 miles long near Reri that is largely red Reshun siltstone. He also found the Sarikol Shale between Lon and Kosht moving downslope as a whole. For any future access road or camp, the red belt is the ground to watch. InSAR time series from these same two orbits would measure that creep directly.

The mineral record for this corridor is thin, and the published sources are candid about it. Stauffer found no economic deposit in the central Mastuj valley. He recorded three things of interest:

The KP Minerals Department's investor package places this ground in two proposed exploration blocks for Au–Ag–Sb along the Reshun Fault Zone: Block-05 "Mastuj" (184 km²) and Block-06 "Reshun" (191 km²). The same report is sober about scale. After 50 million years of uplift and erosion, it concludes that the chances of world-class ore bodies are minimal. The realistic target is small, structurally controlled, metamorphogenic vein deposits.

That fits everything the satellites show. A quartz–stibnite vein a metre wide is invisible in a 20 m or 60 m pixel, whatever the number of bands. Satellites can map the structure that hosts such veins, and the belt's thrust margins are that structure. The ground has to be sampled to find the veins themselves.

I computed the drainage network from the DEM (40 m D8, priority-flood filled). Every place where a stream of 1 km² or more crosses the red-belt margin is a candidate sampling site. After merging duplicates, 14 remain. A tributary crossing a thrust carries debris from any vein along it. Sediment at the crossing samples the whole upstream reach, and that is how Chitral's known Sb–Pb districts show up in regional geochemistry.

Recommended programme, cheapest first

At each site, take a −80 mesh stream sediment and a panned heavy-mineral concentrate. Analyse the sediment for Sb, As, Au, Ag, Pb, Zn, Cu, W and Hg by aqua-regia ICP-MS, with fire assay for gold. Count scheelite in the concentrate under UV light. Visit E1 on the same trip and take rock chips; send any sulphide-bearing sample for four-acid or fusion ICP to get total grades. Order archival GSP and Austromineral (1974–78) maps and data before fieldwork; the Austromineral work explored Buni Zom iron. After that, depending on results: ground magnetics or self-potential along the belt margins, then IP, then trenching. That is the sequence that found the shallow sulphides at Drosh.

Grade is a measurement, not an inference. Nothing in this report is an assay. No figure here estimates tonnage or grade for any site in this area.

Location. I fitted a physical camera to 24 OSM-geocoded labels plus GLO-30 elevations, with an RMS error of 19 px. Every fit variant converges on the same position: free or fixed field of view, 1.0–2.0× terrain exaggeration, free principal point. Yet the render below hides the red belt behind a foreground ridge, which your screenshot does not. Labels on a near-planar valley floor cannot separate camera distance from zoom. Location therefore rests on the map-view match of the red belt, not on the camera.

Data.

• Copernicus GLO-30 DEM (AWS open data)
• Sentinel-2 L2A: 12 scenes, Aug–Sep 2023–2026, tile 43SBA (Element84 earth-search). BOA offset already applied, so no −1000 correction.
• ASTER AST_07XT and AST_05: 5 scenes, 2001–2003, before the SWIR detector failure in April 2008
• EMIT L2B MIN: 5 dates (NASA LP DAAC), orthorectified by GLT gather. Masked (−9999) and "no mineral" (0) pixels were kept apart.
• Sentinel-1 RTC: summer 2025, per orbit, never mixed across orbits (Microsoft Planetary Computer)

Known limits.

• At 20–60 m the imagery cannot detect veins.
• Colour units come from clustering and are not a geological map.
• The red-belt outline approximates the Reshun Fm. Where it abuts other red rock (terrace soils, fault slices) it will be wrong locally.
• The trunk river's catchment is truncated at the analysis box. Tributary catchments are complete.
• EMIT library matches have a measurable false-positive rate (see the negative control).