-2.87411, 32.21903 · 2000 m × 2000 m window (1000 m each side) · AOI_E0320_S030 · generated 5 Nov 2026, 03:54 EAT
Grid squares are 200 m. Dashed rings sit at 200 m, 500 m and 1000 m from the point. Traces are clipped to the window; a line may continue well beyond it.
What each bearing meets between the point and the edge of the window.
| Bearing | Reach | First structure met | Distance | What it is |
|---|---|---|---|---|
| N (0°) | 1000 m | nothing crossed on this bearing | ||
| NE (45°) | 1414 m | nothing crossed on this bearing | ||
| E (90°) | 1000 m | nothing crossed on this bearing | ||
| SE (135°) | 1414 m | nothing crossed on this bearing | ||
| S (180°) | 1000 m | nothing crossed on this bearing | ||
| SW (225°) | 1414 m | nothing crossed on this bearing | ||
| W (270°) | 1000 m | nothing crossed on this bearing | ||
| NW (315°) | 1414 m | nothing crossed on this bearing | ||
Traces leaving the window: north edge 2, east 0, south 0, west 1. A structure that exits an edge continues beyond this report — prospect along it, not across it.
| # | Verdict | Distance | Bearing to it | Strike | In window | Full length | Relief | Throw | Agreement |
|---|---|---|---|---|---|---|---|---|---|
| L214 | scarp | 678 m | NNW (334°) | 35° | 479 m | 1.09 km | 181 m | 52.1 m | 100% |
| L241 | scarp | 764 m | WNW (288°) | 62° | 310 m | 1.05 km | 142 m | 22.1 m | 67% |
| L306 | scarp | 826 m | NW (325°) | 54° | 563 m | 995 m | 180 m | 46.2 m | 78% |
Strikes 35° (NE–SW), carrying 479 m of trace through the window out of 1.09 km in total. The cross-section pass called it scarp — a real step in the ground — the strongest kind of evidence this pipeline produces. It leaves the window through the N edge. The point sits 678 m to the SSE of it.
In plain terms. The ground drops about 52 m from one side of this line to the other, and that step survives once the general slope of the land is taken out. Rock moved here. It behaves the same way along most of its length, which is what a single continuous structure looks like. What it does not tell you is whether there is gold in it. Only digging across it does that.
Median elevation across the trace, sampled 400 m either side at 9 places along the line; the shaded band is the 10th–90th percentile of those samples. Noise floor 8 m. A real fault steps, a fracture zone notches, a field edge stays flat.
Strikes 62° (NE–SW), carrying 310 m of trace through the window out of 1.05 km in total. The cross-section pass called it scarp — a real step in the ground — the strongest kind of evidence this pipeline produces. It leaves the window through the W edge. The point sits 764 m to the ESE of it.
In plain terms. The ground drops about 22 m from one side of this line to the other, and that step survives once the general slope of the land is taken out. Rock moved here. It behaves this way along a little over half its length, so it is probably more than one thing joined end to end — worth walking before choosing where to dig. What it does not tell you is whether there is gold in it. Only digging across it does that.
Median elevation across the trace, sampled 400 m either side at 9 places along the line; the shaded band is the 10th–90th percentile of those samples. Noise floor 8 m. A real fault steps, a fracture zone notches, a field edge stays flat.
Strikes 54° (NE–SW), carrying 563 m of trace through the window out of 995 m in total. The cross-section pass called it scarp — a real step in the ground — the strongest kind of evidence this pipeline produces. It leaves the window through the N edge. The point sits 826 m to the SE of it.
In plain terms. The ground drops about 46 m from one side of this line to the other, and that step survives once the general slope of the land is taken out. Rock moved here. It behaves the same way along most of its length, which is what a single continuous structure looks like. What it does not tell you is whether there is gold in it. Only digging across it does that.
Median elevation across the trace, sampled 400 m either side at 9 places along the line; the shaded band is the 10th–90th percentile of those samples. Noise floor 8 m. A real fault steps, a fracture zone notches, a field edge stays flat.
| Verdict | Distance | Bearing | Strike | Full length |
|---|---|---|---|---|
| scarp | 1.20 km | WNW (291°) | 61° | 2.16 km |
| scarp | 1.38 km | NNW (344°) | 53° | 1.08 km |
| mixed | 1.58 km | WNW (292°) | 50° | 2.17 km |
| ridge | 1.63 km | WNW (285°) | 15° | 1.72 km |
| scarp | 1.65 km | W (281°) | 61° | 1.16 km |
| scarp | 1.71 km | NNW (328°) | 22° | 1.10 km |
The orientations here are statistically indistinguishable from random (p = 0.07). With 3 lines in a 2000 m square that is expected — treat the dominant strike as descriptive, not diagnostic.
Sense of throw. The nearest scarp (L214) stands about 23.2 m lower on one flank once the regional slope is fitted out, so the downthrown side faces roughly NW (305°). The stored throw for that line is 52.1 m — if those two numbers disagree badly, the profile is being dragged by terrain the fit could not remove. This reads the profile's offset sign as increasing toward strike + 90°. Check it once against a scarp whose downthrown side you know on the ground before relying on it.
Petals weighted by trace length inside the window, 10° bins, mirrored — a line has no head or tail. The dashed line is the length-weighted mean strike.
No two structures cross inside this window.
| Distance | Bearing | Structures | Angle | Opening potential | Why | |
|---|---|---|---|---|---|---|
| out | 1.01 km | N | 2 | 19° | low · crossing | same-family crossing (19 deg) |
| out | 1.63 km | WNW | 2 | 36° | med · crossing | cross-family crossing (36 deg) |
| out | 1.66 km | W | 2 | 46° | med · crossing | wide cross-family crossing (46 deg), only 2 structures |
| out | 1.78 km | NNW | 2 | 24° | low · crossing | same-family crossing (23 deg) |
Node favourability is an untested hypothesis layer at Geita — the mines calibration found no enrichment for map-view crossings there, where gold sits on sheared BIF contacts instead. Read these as geometry, not as a score you can trust yet.
Every target below is a place you can stand, with the work to do when you get there. The ordering ranks structure — how strongly the break is expressed, how well the cross-section pass verified it, and how two of them meet. Nothing in this pipeline measures metal.
| # | Target | Priority | From the point | Bearing | Coordinates | UTM Arc 1960 |
|---|---|---|---|---|---|---|
| T1 | Trench across L214 (scarp) | first pass | 678 m | NNW (334°) | -2.86861, 32.21639 | 36M 412 828E 9 683 203N |
| T2 | Trench across L306 (scarp) | first pass | 826 m | NW (325°) | -2.86806, 32.21472 | 36M 412 642E 9 683 264N |
| T3 | Trench across L241 (scarp) | first pass | 764 m | WNW (288°) | -2.87194, 32.21250 | 36M 412 396E 9 682 834N |
Cut across the grain, not along it: these run 125°, square to the strongest structure in the window (L214, 35°) — the orientations here are too scattered for a mean strike to mean anything. A sample line that follows a fault instead of crossing it is the one orientation guaranteed to teach you nothing.
| Line | Bearing | Length | Start | End |
|---|---|---|---|---|
| Line 1 | 125° | 1600 m | -2.88120, 32.22291 | -2.87298, 32.21106 |
| Line 2 | 125° | 1600 m | -2.87822, 32.22495 | -2.87000, 32.21311 |
| Line 3 | 125° | 1600 m | -2.87524, 32.22700 | -2.86702, 32.21515 |
Sample every 25 m along each line, at a consistent depth below the organic layer — a shallow sample and a deep one on the same line are two different datasets. Termite mounds are a good alternative sample medium in this terrain where soil cover is thick. Line spacing here is 400 m; tighten it to 50 m once a line gives you something. The lines run past the edge of the window on the plan — that is deliberate, since an anomaly does not stop where this report does.
1 — Walk it first (a day or two, no cost). Go to T1 and T2 and look at the ground. Is there outcrop? Is there quartz? Is there old working? Does the break the satellite drew actually exist where it says? Take rock chips from any outcrop and photograph everything with the GPS on. Half of these targets will fail this step, and that is the cheapest failure available to you.
2 — Sample before you dig. Soil or termite-mound sampling across the lines above costs a fraction of a trench and tells you where along a 479 m structure to put one. Trenching a fault at a random point along its length is the most common way to waste money in early exploration.
3 — Trench where the geochemistry and the structure agree. Not where either one alone looks good. Use the trench lines above: square across strike, long enough to catch both walls of the damage zone, channel-sampled continuously so you get a grade over a width and not a grab sample.
4 — Only then think about drilling. A trench that returns nothing is worth knowing; a hole that returns nothing costs twenty times as much to learn the same thing.
Before any of it: confirm the licence position. These coordinates carry no tenure information — a target inside someone else's PML or PL is not your target.
How the ordering was made. A target's score combines the cross-section verdict (a scarp counts fully, a valley 0.85, a ridge 0.55, a flat line zero), how far the terrain agrees along the line, the measured throw or relief, and how much trace runs through the window. Crossings add a term for the angle the two lines meet at, because a high angle means two real families rather than one broken line, plus up to six points for being close enough to walk to. The tiers are positions in that queue, not thresholds — the score has no calibrated meaning in absolute terms, and the top target in a poor window is still a poor target. The only part of this platform tested against real deposits is lineament proximity — 74% of fault-following Tanzanian occurrences lie within 3 km of a DEM lineament, roughly 12× the rate of random ground at Geita. Node favourability was tested at the same place and showed no signal, because gold there sits on sheared BIF contacts rather than map-view crossings. Read the crossing targets accordingly.
Every field in the structure tables is tagged with where it came from, because the four kinds carry different weight and should not be read as though they were the same.
| Field | Kind | What it is |
|---|---|---|
| Line rank | derived | The identifier for this trace within its cell. It has no geological meaning and does not rank anything. |
| Strike (°) | measured | The compass direction the break runs, quoted 0–180°. This is the best-constrained number in the report; strike is reliable even where everything else is marginal. |
| Total trace length (m) | measured | The full length of the trace end to end, including the part outside this window. |
| Fragments merged | derived | How many separate fragments the pipeline joined into this one trace. More than one means the line is a reconstruction, so treat its ends with care. |
| Cross-section verdict | derived | What the ground actually does across the line: steps down one side (scarp), is eroded out along it (valley), stands up along it (ridge), or does nothing (flat). Read from the measured profile, not from the image. |
| Relief (m) | measured | Total height difference across the sampled ground, lowest to highest. It includes the regional slope, so a large relief on a hillside does not by itself mean a large structure. |
| Throw (m) | measured | Vertical separation between the two flanks once the regional slope has been fitted out. Zero means the ground sits at the same height either side — the feature is a shape in the land, not an offset of it. |
| Expression (m) | derived | How strongly the break shows in the ground: the larger of relief and throw. A convenience number for ranking, not a physical quantity. |
| Agreement along line | derived | The share of cross-sections along the line that agreed on the same verdict. Low values mean the line changes character along its length, which usually means it is more than one thing. |
| Sections measured | measured | How many cross-sections were cut along the line to reach the verdict. |
| Notch depth (m) | measured | Depth of the notch the line sits in, where there is one. |
| Notch width (m) | measured | Width of that notch across strike. A wide, shallow notch is a valley; a narrow, deep one is a fracture. |
| Wall asymmetry | measured | Steepness of one flank divided by the other. 1.0 is symmetric. A strong asymmetry can point to the dip direction, but it can equally be the drainage, so it is a hint and not a measurement of dip. |
| V-shape index | derived | How V-shaped the profile across the line is, 0 to 1. A V is erosion cutting into a fracture. |
| U-shape index | derived | How U-shaped it is, 0 to 1. A U is a broader trough — often a softer unit rather than a single break. |
| Valley score | modelled | The classifier's confidence that the profile is a valley, 0 to 1. |
| Ridge score | modelled | Its confidence that the profile is a ridge, 0 to 1. |
| Structural class | modelled | Where this line sits in a Riedel-style scheme — R, R′, P, Y or oblique — measured by angle against the regional trend fitted upstream in the lineament pass. It is an assignment by geometry alone. It is not a field observation of shear sense, and it should not be quoted as one without structural data from the ground. |
| Affinity | modelled | How well the line fits the class above, 0 to 1. Low values mean the class is close to arbitrary. |
| Angle to regional trend (°) | derived | The angle between this line and the regional trend the classifier uses. |
| Near a class boundary | modelled | The line sits close to the angular boundary between two classes, so the class above could reasonably have been the other one. |
Two things this report does not carry. There is no dip — strike is measured, but the angle a structure leans at is not something a plan-view DEM can see, so a trench length here is a working figure and should be widened if the break turns out to be shallow-dipping. And there is no tenure: nothing in this pipeline knows who holds the ground.
| Cell | Layer | File | Analysed |
|---|---|---|---|
| AOI_E0320_S030 | dem.sections | AOI_E0320_S030_sections.geojson | 8 Sep 2026 |
Lineaments come from NASADEM elevation: multi-azimuth hillshade, Canny edge detection and a
probabilistic Hough transform. Cross-section verdicts come from lineament_sections.py, which samples
the real ground across each line and asks whether it steps, notches, rises or stays flat.
What this report is. Every line here is a break in the shape of the land, read from satellite elevation data. It is structure, never gold. Faults are where gold-bearing fluid can travel and collect, which is why the ground near one is worth more attention than the ground away from one — but a fault with no gold in it is the ordinary case, not the exception.
Accuracy. Distances are straight-line, not walking distance. Trace positions carry the DEM's own positional error — roughly a pixel, about 30 m — so treat anything under 50 m as "at the point". Strike is reliable; the sense of throw is inferred, not surveyed.
GoldRadar · AOI_E0320_S030 · generated 2026-11-05 03:54 EAT · -2.87411, 32.21903