A commercial-grade desktop application for geothermal exploration geoscientists. Ingest heterogeneous evidence — heat-flow points, fault traces, lithology polygons, geochemistry, seismicity — score each into a common [0, 1] favorability scale, propagate with explicit spatial influence, and combine into a composite prospectivity surface with quantified uncertainty.
Unlike petroleum systems — where buoyant hydrocarbons migrate through a charge–trap–seal architecture — geothermal systems are static coincidence problems. The question is not whether a fluid has accumulated; it is whether sufficient heat, open permeability, and circulating fluid occupy the same rock volume. The PFA application treats each element as an independent risk surface, scores it from geoscience evidence, and combines the three into a composite favorability map.
A critical implementation rule: favorability scoring is applied at the data point, then the scored surface is interpolated. Reversing this order — interpolating raw data, then scoring — violates Jensen's inequality for any non-linear scoring function, and it smooths genuine anomalies into the background.
Element confidence is tracked as a separate surface, propagated independently via kernel density estimation with Sheather–Jones bandwidth. Well-drilled regions show high confidence; frontier basins show low confidence. The composite output carries both signals — never one without the other.
Does anomalous heat reach the exploration depth? The PFA stack aggregates direct measurements (heat-flow wells, BHT, equilibrated temperature logs) with proxy evidence: Quaternary volcanism proximity, crustal thinning, radiogenic heat production from the Lithology Library (heat_production_uw_m3), and regional thermal gradient gridding.
Heat without fluid pathways is unreachable. Permeability is dominantly fracture-controlled in crystalline and competent sedimentary settings. The system ingests mapped fault traces, computes density via KDE, reads perm_enhancement_class from the Structure Library, and layers in stress regime favorability and seismicity-derived strain.
Sufficient fluid volume must circulate through the permeable horizon to harvest the thermal energy. Evidence: recharge-zone elevation and catchment area, cation geothermometry (Na–K, Na–K–Ca), spring and seep inventories, and hydrogeological unit context from the hydrogeology_library. Isotopic signatures (δ¹â¸O, δ²H) separate deep-circulated fluids from shallow aquifer water.
Geothermal datasets are heterogeneous: legacy well files with Bottom-Hole Temperature, shapefiles of mapped faults from state geological surveys, GeoTIFF heat-flow grids from academic compilations, LAS logs from gradient wells, CSV tables of spring chemistry. The PFA importer handles each, projects into a user-chosen Coordinate Reference System, and enrolls the layer with provenance in the project database.
Every evidence layer follows the same three-stage transformation: score the data at its native geometry, propagate the scored values into the target grid with an explicit spatial model, and combine across elements. Every parameter — scoring function, decay kernel, background value, combination weight — is recorded to the project database for audit.
Each measurement is transformed into a [0, 1] favorability value by a calibrated scoring function. For continuous
data, the user selects from monotonic, bell, sigmoid, piecewise-linear
and specifies parameters (thresholds, inflection points). For categorical polygon layers — lithology, structure,
hydrogeology, alteration — scores auto-populate from the built-in 5-tier ordinal mapping
(NONE→0.0, POOR→0.15, FAIR→0.35, GOOD→0.70,
EXCELLENT→0.90). Every mapping is per-scenario editable and audit-tracked.
Scored values are propagated into a common grid. Point evidence uses Gaussian decay, IDW, kriging, thin-plate
spline or minimum curvature — anisotropy (azimuth + ratio) is first-class. Polygon evidence is center-point
rasterized with the layer score. Cells outside evidence extent receive an
explicit background: pessimistic (0), agnostic (0.5), optimistic (1), or no-data.
agnostic 0.5; pessimistic must be chosen deliberately.
Element surfaces combine via geometric-mean, multiplication, minimum,
OWA or weights-of-evidence. Weights are normalized to the simplex and are sensitivity-
perturbable. Output is the Composite Composite Risk Surface (CCRS) plus a paired composite confidence surface.
minimum combiner encodes the limiting-factor nature of the
three-element problem. A site can have abundant heat and permeability, but with no fluid it is not a play.
MINIMUM or a geometric-mean with weights biased toward the limiting element (historically heat, but in the Great Basin it is often fluid). Always run a weight-perturbation sensitivity study before publishing a prospect map.
The PFA system treats favorability and confidence as paired surfaces. Confidence is computed independently of the decay kernel via Kernel Density Estimation with Sheather–Jones bandwidth (Silverman fallback), domain-bounded by the project AOI polygon. Users see both layers side by side — and the composite can be confidence-weighted, so poorly sampled cells cannot masquerade as high-confidence anomalies.
The most methodologically mature geothermal play-fairway study is the DOE-funded Snake River Plain PFA (Shervais et al., Geothermics 2024). The PFA application loads the published evidence layers natively and reproduces the composite within the scoring system described above. It also exposes the levers the original study did not formally vary — weights, background values, decay lengths — so explorers can stress-test the conclusions rather than consume them as given.
Geothermal exploration datasets are often proprietary — state-survey compilations, partner well data, unpublished gradient surveys. The application is a native desktop product with an embedded spatial database. No upload, no cloud round-trip, no third-party data sharing. Projects are single files that can be versioned like any engineering artefact.
MathTransform.
.pfa project · full spatial SQL · pure Java, zero install. Metadata, layer registry, well data, vector features, GDE + Lithology + Structure + Alteration + Hydrogeology libraries, and scoring scenarios all persist transactionally. Raster tiles live on filesystem; the database indexes them.
We are actively working with academic partners and commercial explorers. If you are running a favorability study on a basin-scale or regional dataset — conventional hydrothermal, EGS, or sedimentary low-temperature — we would like to talk.