Overview
Once you have exported your results, the package includes a VPS .csv, VPS raster, and Feature Importance (SHAP) files. This article explains what each output contains and how to use them for field follow-up and interpretation.
For instructions on how to export your results, see [DORA 2.0] Export Prediction Map Results.
Working With the VPS Results CSV
The VPS results can be downloaded in a .csv format, with each row representing one pixel.
Field | Description |
X, Y | Spatial coordinates of the pixel. |
pred_z | Predicted depth of the prospective signal. |
VPS | Prospectivity score for that pixel (0–1). |
targets | Target cluster identifier; filter by this field to isolate individual targets in GIS. |
dist_learn | Distance in metres to the nearest learning point. |
VPS_noDDH | VPS score calculated with a buffer applied around drill hole locations (approximately 100–150 m), reducing the influence of densely drilled areas on the VPS. |
VPS_var | How much the VPS score varied across model runs for that pixel. Lower is more consistent. |
VPS_uncertainty | Per-pixel measure of how strongly the prediction is supported by nearby data. Higher values indicate greater distance from known data points and less prediction confidence |
Z_var | How much the predicted depth varied across model runs for that pixel. Lower is more consistent. |
The following workflow is one approach to using the results for field follow-up:
Import the
.csvinto Leapfrog, QGIS, or equivalent software.Filter or symbolise by target group to isolate individual targets.
Calculate the average VPS score across each target's pixels as a relative ranking between targets.
Cross-reference with the Feature Importance plots (included in the export as images and a CSV) to understand what is driving each target.
Generate polygons around priority targets and extract coordinates for ground truthing.
If you have an existing target ranking spreadsheet (incorporating permitting status, access, data coverage, etc.), add the average VPS score as an additional weighted column.
Working With the VPS Raster
A 2D raster of VPS results is also included in the export and can be visualized in GIS software. To focus only on high-confidence results, apply a VPS threshold filter (e.g., show only values above 0.7).
Adjust layer transparency or symbology to highlight these zones without obscuring underlying geological layers. This helps replicate your DORA view and maintain clarity in your own environment.
Remember to watch for edge effects or anomalies in raster visuals. These may indicate data misalignment or overly influential layers and should be reviewed carefully before decision-making.
A Note on Depth vs. X/Y
DORA is best understood as a 2.5D model. Depth estimates provide directional guidance, but X/Y targeting is the primary output for field planning purposes. Keep in mind that depth projections alone are not sufficient to accept or reject a target.
Working With Feature Importance (SHAP)
The export includes both PNG summary plots and a .csv file:
Field | Description |
Feature | The input raster that contributed to the prediction. |
Contribution % | The relative importance of that feature to the modelled result. |
Target | The target group the contribution applies to. |
You can import the .csv into QGIS, Leapfrog, or a spreadsheet to filter by target and explore which features are driving each modelled result, and whether they align with your geological expectations.
Learn More
Still Have Questions?
Reach out to your dedicated DORA contact or email support@VRIFY.com for more information.