Axine Labs
Free / commercial imagery

District Prospectivity Ranking

A district-scale likelihood surface you can rank targets against, with the cross-validation scheme stated on the layer.

Regional likelihood scoring for hydrothermal systems, backed by leave-one-region-out cross-validation.

District Prospectivity Ranking
In view

Open-pit iron ore operations. District scores rank ground like this before anyone commits to it.

WHAT THE SENSOR SEES

What this layer looks for.

The laboratory reference for each material, and what the sensor's band spacing captures of it. The shaded band marks the range the method reads.

0.602.002 µm2.090 µm2.179 µm2.268 µm2.357 µm2.446 µmWavelength (µm)Reflectance
Laboratory reference
Sensor sampling
Working range
Band depth0.28
Feature width30 nm
Band spacing7.4 nm

WHAT IT FINDS

What it detects.

MaterialTypical settingWhat can confuse it
Hydrothermal Core ComplexesIntrusive centers, volcanic calderas, and porphyry lithocapsExtensive post-mineral volcanic cover
Structural Alteration CorridorsRegional faults, shear zones, and detachment systemsQuaternary colluvium in fault trenches

METHOD

How it works.

Calibrated reflectance goes in, and every pixel is fitted against a public reference library.

01Step 1

District alteration clustering

Pulls spatial clusters of advanced argillic, phyllic and gossanous pixels out of contiguous scenes.

02Step 2

Leave-one-region-out calibration

Whole geological provinces are withheld during fitting, so a score cannot borrow its evidence from a neighbouring pixel in the same district.

03Step 3

Label permutation as a negative control

Discrimination is retested against randomised occurrence coordinates. A model that scores well on shuffled labels is measuring the sampling, not the ground.

04Step 4

Score generation with provenance

The output is a continuous district score carrying its model version, its input set and its interval. Expert-system features are deliberately kept out of it, so the two engines stay independent.

SPECIFICATIONS

What you get, and what it runs on.

Values deliveredDistrict score, floating point, reported to two decimals with its interval
OutputDistrict evidence surface
TurnaroundRuns once the contributing scenes are on disk; area drives the queue, not the model
Formats
Cloud-Optimized GeoTIFF (COG)Vector Likelihood Contours (GeoJSON)Provenance record (model version, input set, interval)

Conditions it needs

Sun angle limit:65°
Cloud cover limit:15%
Vegetation cover limit:40%

Method and checks are documented in Axine Applied Spectroscopy Paper #1 · Section 5.1.

Read the paper

Scope

Ranks where to look first from surface evidence, not what a target holds.