Magnetic transformations¶
A magnetic grid is a measured field, and the field is not the geology. The transformations here re-express that field to make source positions and source boundaries easier to read. Each answers a narrow question and introduces its own artifacts.
Read all of them as contrast rather than absolute value. What matters is where the field changes.
Which grid you are transforming decides how much any of this is worth. See Magnetics for line spacing, coverage and the reason cell size is not resolution. Transformations of a coarse regional grid are still a coarse regional grid.
What is computed from what¶
The layer names tell you the lineage, and the lineage matters more than it looks. Layers sharing an input share its artifacts, so two layers computed from the same intermediate are one calculation seen twice.
| Layer | Computed from |
|---|---|
| TMI | the survey, after main-field removal |
| TMI AS | TMI |
| TMI RTP | TMI |
| TMI RTP 1VD, 2VD | RTP |
| TMI RTP HG | RTP |
| TMI RTP Residual | RTP |
| TMI RTP Tilt | RTP |
Everything except Analytic Signal runs on RTP. That is the single most useful fact on this page when you are weighing whether two layers agreeing means anything.
Total Magnetic Intensity (TMI)¶
The measured anomaly field, in nanoteslas, after the core field has been removed. Everything else here derives from it.
TMI anomalies are offset from the bodies that cause them, by an amount that depends on magnetic inclination, on the body's own magnetization, and on how deep it sits. The bigger reading problem is shape rather than position: away from the poles the anomaly is dipolar, a paired high and low straddling the source, so a single body can present as two features and be mapped as two. RTP exists to fix this.
Reduction to the Pole (RTP)¶
RTP recomputes the field as it would have been measured at the magnetic pole, where induced anomalies sit symmetrically over their sources. After RTP a magnetic high is roughly above the body that produced it, which is what makes the layer readable next to a geological map.
Load this first if magnetics is not your background. Two different things break it, with different symptoms.
Low inclination breaks it numerically. The filter divides by a term that goes small for wavenumbers running near-perpendicular to the declination, which is precisely why the resulting artifacts strike parallel to the declination — close to north-south where declination is small. Below roughly 15 to 20 degrees of inclination this noise amplification produces streaks indistinguishable from real lineaments. Degradation is continuous rather than sudden, so treat anything below about 30 degrees with care, and noisy or coarsely gridded data brings it on at higher inclinations too.
Strong remanence breaks it as a model. The arithmetic stays stable, but RTP assumes the body's magnetization follows the present field, and with remanence it applies the wrong direction. Anomalies are not centered on their sources, dipolar asymmetry survives the transform, and apparent edges shift. In the worst case a body carrying reversed remanence reads as a magnetic low over itself, and the rule that a high sits above its source fails outright.
Near the magnetic equator, or anywhere you suspect remanence, work from Analytic Signal and treat RTP lineaments as unproven until something else reproduces them.
RTP also needs the field's inclination and declination at the survey's location and epoch. Secular variation moves both over decades, so a historical grid reduced with present-day values is wrong. Send the survey date with your data.
Analytic Signal (AS)¶
AS combines the field's gradients into a single positive quantity whose peaks sit over magnetic sources: over the margins for a body wide relative to its depth, and over the body itself for a thin dyke or a deep source, where the two peaks merge.
Its value is that it depends very little on magnetization direction, which makes it the choice exactly where RTP fails. Stable against direction is not stable against noise: AS is built from first derivatives and amplifies noise like any gradient product, so on a coarse or noisy grid treat it as skeptically as 2VD.
It is computed from TMI, so it does not inherit RTP's assumption about magnetization. If an edge appears in both AS and an RTP-derived layer, you have eliminated that assumption as the explanation, which is worth doing and is the reason to run AS at low inclination. It is not independent confirmation. Both filter the same grid, so a leveling stripe, a survey seam, a gridding artifact or a pipeline appears in both. Same-grid agreement is one observation seen through two filters. Independent confirmation means a different survey, a different method, or the ground.
Vertical Derivatives (1VD, 2VD)¶
The first vertical derivative sharpens the field, suppressing broad deep-sourced signal in favor of shallow near-surface detail. Its maximum sits over the body, not over its edges — for edges use Horizontal Gradient or Tilt. The second vertical derivative resolves finer structure still, and its zero crossings approximate edges, at the cost of much more noise.
Each derivative amplifies noise. 2VD on a noisy or coarsely gridded dataset can produce a texture that looks like structural fabric and is entirely numerical.
Horizontal Gradient (HG)¶
Peaks over contacts and the edges of magnetic sources, and less noise-prone than the vertical derivatives.
Two limits. Over a dipping contact the peak sits down-dip of the surface trace, by an amount that grows with depth and with shallower dip, so do not read an HG maximum as a surveyed contact position. And because HG is computed from RTP, it is not an independent check on Tilt or on the vertical derivatives — use it to cross-check against Analytic Signal, which has a different input.
Residual¶
Removes the regional trend so local anomalies stand out. Useful when your target sits on a strong regional gradient that dominates the display range.
The regional field it removes is a modeling choice, which means a residual layer shows you anomalies at one particular scale. An anomaly that appears only in the residual is worth confirming in TMI or RTP.
Tilt Derivative¶
The arctangent of the ratio of the vertical to the horizontal gradient, which is why it is bounded between −90° and +90°: that bounding is the normalization. Its zero crossings fall over the edges of magnetic sources, and because the ratio is normalized it brings out weak anomalies alongside strong ones in a single display. The half-distance between the +45° and −45° contours is also a serviceable depth-to-top estimate for a vertical contact.
Tilt is computed from RTP, so Tilt and RTP agreeing is one calculation seen twice, and Tilt inherits every RTP artifact described above.
Displaying it¶
- Open the Browse Data tab, open Geophysics → Magnetometry, and click Add layer.
- Pick the product from the dropdown.
- Switch between Raster and Shaded.
- In Shaded mode, set the Light Controls: Azimuth (0 to 360°), Altitude (1 to 90°), and Z-Factor (0.1 to 100) for vertical exaggeration.
- Choose a Shading model: Standard Lambertian for uniform illumination, or Aspect-Adaptive, which adjusts to surface orientation.
- Use Clip Out-of-Range to hide values outside the selected range, and Show Contours for contour lines.
- Set Opacity to blend against the base map.
Shaded mode is not decoration. A magnetic grid behaves like a surface, and shading brightens features running across the illumination while hiding features running along it. So rotate the azimuth through a full turn before you conclude a fabric is absent: a north-south lineament only shows under roughly east-west light, and its disappearance under north-south light is expected rather than evidence against it.
What makes a lineament worth interpreting is that it appears whenever the illumination is favorable to its strike, and that it appears in transformations with different input grids. A feature visible in RTP, Tilt and HG has been seen once, because all three come from RTP.
Automated lineament extraction is not available yet, so this reading is manual.