Question Commercial mining software (Leapfrog, Surpac, Datamine, Micromine) dominates 3D geological modelling and resource estimation, but licenses are expensive and closed. Is there a reproducible, fully open-source pipeline that can take public drillhole data and produce (a) a 3D block model and (b) a tonnage/grade (resource) estimate? What are the practical gotchas? Answer Yes. A free stack — Python + GDAL + gstlearn + pyvista — covers the whole main workflow. I verified it end-to-end on real public data (Tom's Gully gold deposit, Pine Creek, Northern Territory, Australia; drillhole + assay data from the NTGS open-data portals). The five-step pipeline * Ingest — parse tabular drillhole exports (NTGS format) and read the regional borehole GPKG (pyogrio/geopandas for geometry, sqlite3 for attributes). * Desurvey — compute true 3D coordinates along inclined holes with the minimum-curvature method. * Variography — fit a 3D variogram with gstlearn. * Kriging — ordinary kriging onto a regular block grid. * Estimate — tonnage = block volume × specific gravity; metal = tonnage × grade; build a tonnage-grade curve across cut-offs. Minimal gstlearn 3D setup (this is where most people trip): import gstlearn as gst gst.setDefaultSpace(gst.SpaceRN.create(3)) # NOT defineDefaultSpaceRnT(3)! db = gst.Db.fromTL(df[['X','Y','Z','AU']]) for name, i in (('X',0), ('Y',1), ('Z',2)): db.setLocator(name, gst.ELoc.X, i) db.setLocator('AU', gst.ELoc.Z, 0) vp = gst.VarioParam.createFromSpaceDimension(nlag=12, dlag=25.0, toldis=0.5) va = gst.Vario(vp); va.compute(db=db) mo = gst.Model.createFromVario(va); mo.fit(va) gst.kriging(db, grid, mo, gst.NeighMoving.create(True, 8, 40.0)) Result on the test deposit: spherical model, sill 7.76, ranges (E,N,vertical) = (72.5, 18.9, 10.2) m; 10×10×5 m blocks; at a 0.5 g/t cut-off the shallow public data yields ≈174 kt @ 2.33 g/t (≈13,000 oz Au). Gotchas worth knowing (all reproducible) * Silent TEXT columns — numeric fields stored as TEXT return 0 rows for a numeric BETWEEN with no error. Always CAST(col AS REAL). * defineDefaultSpaceRnT(n) builds RN(n+1) — use SpaceRN.create(3) for true 3D. * File name ≠ resolution — a glo30 DEM file was actually 8 arcsec (~241 m), not 30 m; read the geotransform, not the name. * Vertical datum mismatch — mine "Reduced Level" (~1002 m) vs sea-level DEM (~16 m), a ~986 m offset. Align vertical datums before modelling. * Same name, different place — "Toms Gully" lithology drilling sits ~1.2 km from the official "Tom's Gully" deposit point; always cross-check coordinates against the authoritative deposit record, not the project name. * Open-data boundary — the public archive only holds shallow legacy assays (0–60 m); deeper resource-definition assays live in company PDFs. Open tools reproduce the method and a magnitude estimate, not the full deposit resource. Full scripts, data provenance and boundary statements are in the project repository https://github.com/lucia-lee-geo/tomgully-3d-geology. Formal JORC classification still requires a Competent Person sign-off — scripts do not replace that.

From noise to art: how to generate high-quality images using diffusion models

Vulnerability-Lookup - Fast vulnerability lookup correlation from different sources.

Overview Similar to #7122. The user_dict is proving to be stubborn and doesn't clear properly when clearing field data. import pyvista as pv mesh = pv.Sphere() mesh.user_dict = dict(a=0) mesh.f...

Overview 🎉 HAPPY NEW YEAR 2025 🎉 relate to #5417 Details Updated the year of LICENSE to 2025.

3D plotting and mesh analysis through a streamlined interface for the Visualization Toolkit (VTK) - pyvista/pyvista

Convert raster to unstructured mesh structure. Contribute to changliao1025/uraster development by creating an account on GitHub.