Overview
DORA generates a VRIFY Prospectivity Score (VPS) and an uncertainty value for every grid cell in your Prediction Map. Uncertainty measures how well-supported each VPS prediction is by the model. Where predictions consistently agree, uncertainty is low; where they diverge, uncertainty is high.
Two features make uncertainty visible in your workflow:
Prediction Uncertainty Layer — a spatial raster across your Area of Interest (AOI) that shows where confidence is highest and lowest at a glance
VPS x Uncertainty Filter — a colour-coded grid on the Prediction Map that lets you filter targets by both prospectivity and prediction confidence at once
Note: If you ran your prediction before this feature was released, the VPS x Uncertainty filter and Prediction Uncertainty layer will not be available until you return to Step 4: Select Input Features and rerun your prediction. |
What Is VPS Uncertainty
When you click Run Prediction in Step 4: Select Input Features, DORA generates a Prediction Map showing a VRIFY Prospectivity Score (VPS) and uncertainty value for each grid cell where a prediction was produced. uncertainty measures how well-supported each of those predictions is by the model.
To produce a VPS, DORA runs hundreds of predictions using different combinations of your input data. Where those predictions consistently agree, uncertainty is low. Where they diverge, uncertainty is high. This commonly occurs in areas where the model encountered limited, conflicting, or unfamiliar data during training. The uncertainty shown in DORA is known as “epistemic uncertainty,” or uncertainty due to limited knowledge.
A grid cell in a data-rich area of your Area of Interest (AOI) will typically carry lower uncertainty than one where the model had to extrapolate, even if both cells share the same VPS.
To quickly view the values for a specific point, make sure that the VRIFY Prospectivity Score layer is toggled on in the 3D Layers List, then hover over any VPS point on the map. A tooltip will display both the VPS and the uncertainty value for that location:
The Uncertainty value is also included as a column in the VPS results CSV, so you can filter and analyze targets outside of DORA.
Viewing the Prediction Uncertainty Layer
Additionally, you can toggle the Prediction Uncertainty raster layer on and off using the eye icon in the 3D Layers List on the right:
Click the three dots to the right of the Prediction Uncertainty layer to adjust the transparency:
The layer shows you where uncertainty is distributed spatially across your entire AOI at a glance. Use it as a first pass to identify zones of high and low uncertainty before using the more granular VPS x Uncertainty Filter for target-level filtering.
Using the VPS x Uncertainty Filter
Two grid cells can share the same VPS but carry very different levels of prediction uncertainty. The VPS x Visualize Uncertainty filter lets you see both dimensions at once on the Prediction Map, so you can factor in how well-supported each VPS prediction is before selecting targets.
To enable it, click the checkbox next to Visualize Uncertainty under the VPS slider in the lower left of the Prediction Map:
The VPS x Uncertainty filter will appear, displaying a colour-coded grid where:
The Y-axis represents Uncertainty, from High (top) to Low (bottom)
The X-axis represents the VPS, from Low (left) to High (right)
Each cell corresponds to a combination of VPS and uncertainty range. Hover over any cell to see the full range it represents:
The VPS points on the Prediction Map are re-coloured according to which cell they fall into, so each point reflects both its prospectivity and the confidence behind it.
By default, all cells are selected and all points are visible. To filter the Prediction Map:
Click individual cells to select or deselect them. Only points that fall within your selected cells will remain visible on the map:
Click and drag to select multiple cells at once:
Use the grid size dropdown (4x4, 5x5, 6x6) to adjust the granularity of the filter. A larger grid gives you finer control over VPS and uncertainty ranges:
To isolate your most well-supported targets, select cells in the high VPS, low uncertainty area of the filter.
The panel displays a count of data points currently visible on the Prediction Map:
Surface VPS values are displayed as bars, where multiple points are grouped into a single bar.
Sub-surface VPS values are displayed as individual points.
The count updates as you adjust your filter selection:
VPS, Uncertainty, and Target Review
You can use the VPS and Uncertainty results together when reviewing targets:
VPS | Uncertainty | What it Means |
High | Low | Strong, well-supported target |
High | High | Positive prediction, but treat with caution and seek additional geological evidence |
Low | Low | Confident negative prediction |
Low | High | Insufficient information; prediction may be less reliable |
A high VPS with high uncertainty in DORA may still represent a valid target, but it carries more risk and may warrant further investigation.
High uncertainty in DORA does not necessarily mean a target should be dismissed. It indicates that the model had less data support in that area when generating the VPS prediction.
Uncertainty can potentially be reduced by collecting more representative training data or adding more informative input features.
Learn More
Still Have Questions?
Reach out to your dedicated DORA contact or email support@vrify.com for more information.