Skip to content

Interpretation of Simple Bodies, Forward Modelling and Inversion

Gravity and magnetic anomalies of spheres, cylinders, slabs and dykes, depth rules, Gauss excess mass, Euler deconvolution, Talwani modelling and inversion.

  • 23 min read
  • Updated
On this page · 15 sections
  1. Prerequisites
  2. The Hook: One Anomaly, Infinitely Many Bodies
  3. What Is Unique: Gauss’s Theorem and Excess Mass
  4. Gravity Anomalies of Simple Bodies
  5. Magnetic Anomalies of Simple Bodies
  6. Euler Deconvolution
  7. Forward Modelling
  8. Inversion
  9. Memory Tricks
  10. Don’t Confuse
  11. Common Mistakes
  12. Practice
  13. Quick Revision Table
  14. Key Takeaways
  15. What’s Next

Prerequisites

The Hook: One Anomaly, Infinitely Many Bodies


What Is Unique: Gauss’s Theorem and Excess Mass

Apply Gauss’s theorem to a hemisphere enclosing the source, with its flat face on the observation plane. If the anomaly is known over the whole plane, the total anomalous (excess) mass is

$$\boxed{M_e = \frac{1}{2\pi G}\iint \Delta g\,dx\,dy}$$

This holds whatever the shape and depth of the source. In practice the integral is the sum of grid values times cell area, and the anomaly must be well enclosed by the survey (the tails matter).

The actual mass of an ore body of density \( \rho_1 \) in host rock \( \rho_0 \) follows from the excess mass:

$$M = M_e\,\frac{\rho_1}{\rho_1 - \rho_0}$$

What is not unique: the depth, the shape and the density contrast separately. An infinite slab gives \( 2\pi G\Delta\rho t \) whatever its depth, and only the product \( \Delta\rho\,t \) is constrained.


Gravity Anomalies of Simple Bodies

Profiles are along \( x \), with the body’s centre (or edge) below \( x = 0 \) at depth \( z \); \( \Delta\rho \) is the density contrast.

BodyAnomaly \( \Delta g(x) \)MaximumHalf-width \( x_{1/2} \) at half maximumDepth rule
Sphere (point mass, \( M = \frac{4}{3}\pi R^3\Delta\rho \))\( \dfrac{GMz}{(x^2 + z^2)^{3/2}} \)\( GM/z^2 \)\( 0.766\,z \)\( z = 1.305\,x_{1/2} \)
Horizontal cylinder (2D line mass, \( \lambda = \pi R^2\Delta\rho \))\( \dfrac{2G\lambda z}{x^2 + z^2} \)\( 2\pi G\Delta\rho R^2/z \)\( z \)\( z = x_{1/2} \)
Vertical cylinder (on axis; top \( z_1 \), bottom \( z_2 \), radius \( R \))\( 2\pi G\Delta\rho\left[(z_2 - z_1) + \sqrt{z_1^2 + R^2} - \sqrt{z_2^2 + R^2}\right] \)on axis—as \( R\to\infty \), slab
Infinite horizontal slab (thickness \( t \))\( 2\pi G\Delta\rho t = 0.04193\,\Delta\rho\,t \) mGalconstant—no depth information
Semi-infinite thin slab (faulted bed, edge at \( x = 0 \), extending to \( +x \))\( 2G\Delta\rho t\left[\dfrac{\pi}{2} + \tan^{-1}\dfrac{x}{z}\right] \)\( 2\pi G\Delta\rho t \) far over the slabhalf of maximum directly over the edgequarter- and three-quarter-amplitude points are \( 2z \) apart

Derivations of the half-widths:

  • Sphere: \( (1 + u^2)^{-3/2} = \tfrac12 \) with \( u = x/z \) gives \( 1 + u^2 = 2^{2/3} = 1.587 \), so \( u = 0.766 \).
  • Cylinder: \( (1 + u^2)^{-1} = \tfrac12 \) gives \( u = 1 \).
  • Thin slab edge: \( \tan^{-1}(x/z) = \pm\pi/4 \) at \( x = \pm z \), which are the 3/4 and 1/4 amplitude points.

Limiting-depth (Smith) rules

Without assuming a shape, the ratio of the maximum anomaly \( A_{max} \) to its maximum horizontal gradient \( A'_{max} \) sets an upper limit to the depth of the top of the body:

$$\boxed{z \le 0.86\,\frac{A_{max}}{A'_{max}}\ \ (\text{3D bodies})} \qquad \boxed{z \le 0.65\,\frac{A_{max}}{A'_{max}}\ \ (\text{2D bodies})}$$

The constants are exact for a point mass and a line mass: for the sphere \( A'_{max} = 0.859\,GM/z^3 \) (at \( x = z/2 \)), so \( A_{max}/A'_{max} = 1.164z \); for the cylinder the steepest slope is at \( x = z/\sqrt3 \) and the ratio is \( 1.540z \). Any body of finite size has its top shallower than these values. A rule of this kind gives a maximum depth: Part C likes to ask whether it is a maximum or a minimum.


Magnetic Anomalies of Simple Bodies

Magnetisation

$$\mathbf{M} = \mathbf{M}_i + \mathbf{M}_r = \chi\,\frac{\mathbf{F}}{\mu_0} + \mathbf{M}_r \qquad Q = \frac{\lvert\mathbf{M}_r\rvert}{\lvert\mathbf{M}_i\rvert}$$

\( Q \) is the Koenigsberger ratio. For \( Q \ll 1 \) the magnetisation is induced and parallel to the present field; for \( Q > 1 \) remanence dominates and its direction may be anything (reversed, rotated), so anomaly shapes can be unexpected.

The dipole (sphere) at the pole

A uniformly magnetised sphere acts outside as a dipole of moment \( m = MV \) at its centre. At the magnetic pole, with vertical field and vertical induced magnetisation, the vertical (and total-field) anomaly along a profile is

$$\boxed{\Delta Z = \frac{\mu_0}{4\pi}\,m\,\frac{2z^2 - x^2}{(x^2 + z^2)^{5/2}}}$$
  • Maximum over the centre: \( \frac{\mu_0}{4\pi}\frac{2m}{z^3} \). For an induced sphere of radius \( R \) (ignoring self-demagnetisation) this becomes \( \Delta Z_{max} = \tfrac23\,\chi F\,(R/z)^3 \).
  • Zero crossing at \( x = \sqrt{2}\,z \); a weak negative ring beyond it.
  • Half-width at half maximum about \( 0.50z \), so \( z \approx 2x_{1/2} \).
  • Fall-off \( \propto 1/z^3 \), faster than gravity’s \( 1/z^2 \) for the same body: magnetic anomalies are sharper and more sensitive to shallow sources.
Magnetic source at the poleAnomaly formHalf-widthDepth
Monopole (top of a long vertical pipe)\( \propto z/(x^2 + z^2)^{3/2} \)\( 0.766z \)\( 1.305\,x_{1/2} \)
Dipole (sphere)\( \propto (2z^2 - x^2)/(x^2 + z^2)^{5/2} \)about \( 0.50z \)about \( 2x_{1/2} \)
Line of dipoles (horizontal cylinder, vertical magnetisation)\( \propto (z^2 - x^2)/(x^2 + z^2)^2 \)\( 0.486z \)about \( 2.06\,x_{1/2} \)

Anomaly shape and latitude

For an induced, compact body:

LocationTotal-field anomaly over the body
Magnetic pole (\( I = \pm 90° \))symmetric high directly over the body, weak surrounding low
Mid-latitudes, northern hemispherehigh displaced towards the equator (south), low on the poleward (north) side
Mid-latitudes, southern hemispherehigh displaced towards the equator (north), low on the poleward (south) side
Magnetic equator (\( I = 0 \)), profile north–southlow over the body with highs on its north and south flanks
Magnetic equator, profile east–westlow only, no flanking highs, about half the amplitude of the polar peak

The rule to remember: for induced magnetisation, the high is on the equatorward side and the low on the poleward side. RTP (previous page) moves the high back over the body.

Thin dykes and contacts

The anomaly of a long thin dyke depends on its dip, its strike relative to magnetic north, and the field inclination; it is in general asymmetric. Depth to top is read from the sharpness of the anomaly: classical graphical methods use the horizontal distances between characteristic points (half-maximum points, or points where the slope is half its maximum, as in Peters’ half-slope method), multiplied by constants that depend on the assumed body. Today the same information is extracted automatically with the analytic signal (FWHM \( = 2h \) for a thin dyke, \( 2\sqrt3h \) for a contact; see the previous page) and with Euler or Werner deconvolution.


Euler Deconvolution

A field \( T \) that is homogeneous of degree \( -N \) in the coordinates relative to its source satisfies Euler’s homogeneity equation:

$$\boxed{(x - x_0)\frac{\partial T}{\partial x} + (y - y_0)\frac{\partial T}{\partial y} + (z - z_0)\frac{\partial T}{\partial z} = N\,(B - T)}$$

where \( (x_0, y_0, z_0) \) is the source position, \( B \) a background value and \( N \) the structural index, the rate at which the field falls off with distance. With the measured field and its three gradients at several points in a moving window, the equation is solved by least squares for \( x_0, y_0, z_0, B \). Clusters of solutions mark sources; scattered solutions are discarded.

Source geometryStructural index, magneticsStructural index, gravity
Sphere (point dipole / point mass)32
Horizontal or vertical cylinder (line of dipoles / line mass)21
Thin dyke, sill edge, thin sheet10 (edge of a thin sheet; conventions for gravity dykes vary)
Contact (thick step)0—

For the same geometry the magnetic index is one more than the gravity index, because a magnetic field is a derivative of a gravity-like potential. Choosing the wrong \( N \) biases the depth: too high an index gives too great a depth, too low an index too shallow. Werner deconvolution is a related 2D method for thin-dyke and contact sources along profiles.


Forward Modelling

Forward modelling computes the anomaly of a chosen model and compares it with the data; the model is adjusted until they match.

graph LR
    A[Initial model from geology, seismic, drilling, densities and susceptibilities] --> B[Compute anomaly]
    B --> C{Fits the residual within error?}
    C -- No --> D[Adjust shape, depth or contrast]
    D --> B
    C -- Yes --> E[Accepted model: one of many that fit]

Polygon method (2D)

For a 2D body of arbitrary cross-section, approximate the cross-section by a polygon. The gravity at a station reduces to a line integral around the boundary:

$$\Delta g = 2G\Delta\rho\oint z\,d\theta$$

where \( \theta \) is the angle from the station to points on the boundary. Each polygon side contributes a closed-form term, which makes the method fast and exact for the polygon. This is the Talwani method (Talwani, Worzel and Landisman, 1959), the standard tool for 2D gravity profiles. Equivalent formulas exist for the magnetic case.

Extensions:

  • 2.5D: bodies of finite strike length, for anomalies that are elongated but not infinitely long.
  • 3D: a body sliced into horizontal polygonal laminae (contours), each integrated in depth; or assembled from rectangular prisms, each with a closed-form field.

Wavenumber-domain forward modelling

For an interface of relief \( h(x, y) \) about a mean depth \( z_0 \), with density contrast \( \Delta\rho \) across it, Parker’s method expresses the Fourier transform of the anomaly as a series:

$$\tilde{\Delta g}(k) = 2\pi G\Delta\rho\,e^{-\lvert k\rvert z_0}\sum_{n = 1}^{\infty}\frac{\lvert k\rvert^{\,n-1}}{n!}\,\mathcal{F}\left[h^n(x, y)\right]$$

(with the sign fixed by whether \( h \) is measured upward or downward). The first term is just a Bouguer slab of thickness \( h \), continued up from depth \( z_0 \); the higher terms correct for relief that is not small compared with \( z_0 \). It is fast on large grids and is used for basement, Moho and bathymetry models. Rearranged and iterated, the same series gives \( h \) from \( \Delta g \): the Parker–Oldenburg inversion for interface depth.


Inversion

Inversion asks the computer to find a model that fits the data, subject to constraints.

Linear and nonlinear problems

Divide the subsurface into cells with unknown densities \( m_j \). The anomaly at station \( i \) is

$$d_i = \sum_j G_{ij}\,m_j \qquad \mathbf{d} = \mathbf{G}\mathbf{m}$$

where \( G_{ij} \) is the anomaly at station \( i \) of cell \( j \) with unit density. This is linear in the densities. Unknown geometry (depth to basement, shape of a body with known contrast) makes the problem nonlinear, solved iteratively by linearising about a current model.

SituationSolutionComment
More data than unknowns (overdetermined)least squares \( \mathbf{m} = (\mathbf{G}^T\mathbf{G})^{-1}\mathbf{G}^T\mathbf{d} \)fails if \( \mathbf{G}^T\mathbf{G} \) is near-singular
Fewer data than unknowns (underdetermined, the usual 3D case)minimum norm \( \mathbf{m} = \mathbf{G}^T(\mathbf{G}\mathbf{G}^T)^{-1}\mathbf{d} \)concentrates mass near the surface
Either, with noiseregularised (damped, Tikhonov): minimise \( \lVert\mathbf{d} - \mathbf{G}\mathbf{m}\rVert^2 + \lambda\lVert\mathbf{W}\mathbf{m}\rVert^2 \); \( \mathbf{m} = (\mathbf{G}^T\mathbf{G} + \lambda\mathbf{W}^T\mathbf{W})^{-1}\mathbf{G}^T\mathbf{d} \)\( \lambda \) trades data fit against model simplicity; \( \mathbf{W} \) can impose smoothness and depth weighting

Bott’s iterative method for a sedimentary basin

For a 2D basin of known density contrast \( \Delta\rho \) (negative), model the basin as vertical prisms, one below each station:

  1. First estimate from the slab formula: \( t_i^{(1)} = \Delta g_i/(2\pi G\Delta\rho) \).
  2. Compute the anomaly of the prism model, \( \Delta g_i^{calc} \).
  3. Update each thickness: \( t_i^{(n+1)} = t_i^{(n)} + (\Delta g_i^{obs} - \Delta g_i^{calc})/(2\pi G\Delta\rho) \).
  4. Repeat until the misfit is within the data error.

Constraints and honest answers

  • Non-uniqueness is fundamental: it follows from potential theory (equivalent layers can reproduce any external field), not from data quality.
  • Useful inversions use prior information: densities from logs and samples, depths from seismic and drilling, positivity or bounds on contrasts, smoothness or compactness.
  • Ideal-body analysis finds bounds that hold for every model fitting the data (for example the least density contrast able to explain an anomaly); such bounds are unique even though models are not.
  • Joint inversion of gravity and magnetics (and seismic) with shared structure reduces ambiguity.
  • Report the misfit, the regularisation, the constraints, and the range of acceptable models.

Memory Tricks

Don’t Confuse

ThisvsThat
Excess mass from Gauss’s theorem: uniquedepth and shape: non-unique
Sphere half-width \( 0.766z \)cylinder half-width \( z \)
Gravity of a sphere falls as \( 1/z^2 \)magnetic dipole field falls as \( 1/z^3 \)
Smith rule: upper bound on depth to tophalf-width rule: depth to centre for an assumed shape
Induced magnetisation: parallel to present fieldremanent: any direction; \( Q > 1 \) means remanence dominates
Forward modelling: model to data, trial and errorinversion: data to model, by optimisation with constraints
Linear inversion: unknown densities of fixed cellsnonlinear inversion: unknown geometry
Minimum-norm solution: shallow massdepth-weighted solution: spreads mass with depth
Talwani: 2D polygons in the space domainParker: interface relief in the wavenumber domain

Common Mistakes


Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. What is the half-width of the gravity anomaly of a buried sphere, in terms of the depth to its centre?
  2. What does Gauss’s theorem give uniquely from a complete gravity anomaly map?
  3. What is the structural index of a thin dyke in magnetic Euler deconvolution?
Answers
  1. \( x_{1/2} = 0.766z \) (so \( z = 1.305x_{1/2} \)).
  2. The total anomalous (excess) mass, \( \iint\Delta g\,dA/2\pi G \), independent of shape and depth.
  3. 1.

Level 2: Part B Style

Practice Q1. An elongated gravity anomaly over a buried channel fill has a half-width of 350 m. Treating the source as a horizontal cylinder, the depth to its axis is:

(a) 268 m (b) 350 m (c) 457 m (d) 700 m

Answer: (b) 350 m

For a horizontal cylinder \( x_{1/2} = z \). Option (c) applies the sphere rule (\( 1.305\times 350 = 457 \)); (a) multiplies by 0.766; (d) confuses half-width with full width.

Practice Q2. A gravity anomaly has a maximum of 3.0 mGal and a maximum horizontal gradient of 2.0 mGal km\(^{-1}\). If the source is three-dimensional, the depth to its top is:

(a) at least 1.29 km (b) at most 1.29 km (c) at most 0.98 km (d) exactly 1.5 km

Answer: (b) at most 1.29 km

\( z \le 0.86\times 3.0/2.0 = 1.29 \) km. It is an upper bound. Option (c) uses the 2D constant (\( 0.65\times 1.5 = 0.975 \)); (d) simply takes the ratio.

Practice Q3. In the northern hemisphere at mid-latitudes, the total-field anomaly over a compact body with induced magnetisation typically shows:

(a) a symmetric high centred over the body (b) a high on the south side and a low on the north side (c) a high on the north side and a low on the south side (d) a low centred over the body with equal highs east and west

Answer: (b)

For induced magnetisation the high lies on the equatorward side and the low on the poleward side. (a) is the pole case; (d) resembles the equatorial case but on the wrong (east–west) axis.

Practice Q4. A 2D basin with density contrast −0.30 g cm\(^{-3}\) gives a residual of −6.0 mGal at its centre. Bott’s first estimate of the basin thickness there is about:

(a) 54 m (b) 477 m (c) 143 m (d) 954 m

Answer: (b) 477 m

\( t = \Delta g/(0.04193\,\Delta\rho) = -6.0/(0.04193\times(-0.30)) = 6.0/0.012579 = 477 \) m.

The distractors: (a) uses the rock density 2.67 instead of the contrast (\( 6.0/0.1120 = 54 \)); (c) leaves out the density altogether (\( 6.0/0.04193 = 143 \)); (d) uses \( \pi G \) in place of \( 2\pi G \) (\( 2\times 477 = 954 \)).

Level 3: Part C Style

Practice Q5. A residual gravity grid over a massive sulphide prospect integrates to \( 6.3\times10^{6} \) mGal m\(^2\). The anomaly peak is 2.3 mGal and its half-width 500 m. The ore density is 4.5 g cm\(^{-3}\) and the host 2.7 g cm\(^{-3}\). Which set is correct?

(a) Excess mass \( 1.5\times10^{11} \) kg; ore mass \( 3.76\times10^{11} \) kg; sphere centre at 653 m (b) Excess mass \( 1.5\times10^{11} \) kg; ore mass \( 2.5\times10^{11} \) kg; sphere centre at 383 m (c) Excess mass \( 7.5\times10^{10} \) kg; ore mass \( 1.9\times10^{11} \) kg; sphere centre at 653 m (d) Excess mass \( 9.4\times10^{11} \) kg; ore mass \( 2.4\times10^{12} \) kg; sphere centre at 500 m

Answer: (a)

Integral \( = 6.3\times10^6\times10^{-5} = 63 \) m\(^3\) s\(^{-2}\). \( M_e = 63/4.193\times10^{-10} = 1.50\times10^{11} \) kg.

Ore mass \( = 1.50\times10^{11}\times 4.5/(4.5 - 2.7) = 1.50\times10^{11}\times 2.5 = 3.76\times10^{11} \) kg.

Sphere depth \( = 1.305\times 500 = 653 \) m. Consistency check from the peak: \( GM_e/z^2 = 6.674\times10^{-11}\times 1.50\times10^{11}/653^2 = 2.35\times10^{-5} \) m s\(^{-2}\) \( \approx 2.3 \) mGal, matching the observed peak, so a compact source is plausible. Gauss’s theorem gives the excess mass whatever the shape; the sphere model adds only the depth.

The distractors: (b) scales by \( \rho_1/\rho_0 = 4.5/2.7 \) instead of \( \rho_1/(\rho_1 - \rho_0) \) and uses 0.766 instead of 1.305; (c) divides by \( 4\pi G \) instead of \( 2\pi G \); (d) divides by \( G \) alone and takes the depth equal to the half-width.

Practice Q6. An induced sphere (susceptibility 0.02 SI, radius 40 m) lies at the magnetic pole in a field of 60,000 nT, with its centre 120 m deep. Consider:

  1. The peak vertical anomaly is about 29.6 nT.
  2. The anomaly changes sign about 170 m from the point above the centre.
  3. The half-width at half maximum is about 60 m.
  4. Doubling the depth halves the peak anomaly.

Which are correct?

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2 and 4 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 3 only
  1. \( \tfrac23\times 0.02\times 60{,}000\times(40/120)^3 = 800\times(1/27) = 29.6 \) nT. Correct.
  2. Zero at \( x = \sqrt2z = 1.414\times120 = 170 \) m. Correct.
  3. Half-width \( \approx 0.50z = 60 \) m. Correct.
  4. Wrong. A dipole field falls as \( z^{-3} \), so doubling the depth reduces the peak by a factor of 8.

Practice Q7. Consider these statements about inversion of gravity data for density:

  1. With more cells than data, the minimum-norm solution tends to place anomalous density close to the surface.
  2. Increasing the regularisation parameter \( \lambda \) gives a smoother model and a larger data misfit.
  3. Depth weighting makes the solution unique.
  4. Knowing the density contrast and inverting for the depth to an interface is a nonlinear problem.

Which are correct?

(a) 1 and 3 only (b) 1, 2 and 4 only (c) 2, 3 and 4 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 4 only
  1. Correct: kernels decay with depth, so the smallest-norm model uses shallow cells.
  2. Correct: \( \lambda \) trades misfit against model norm or roughness.
  3. Wrong: depth weighting selects one solution by a chosen criterion; non-uniqueness remains.
  4. Correct: the anomaly depends nonlinearly on the geometry (as in Parker’s series).

Quick Revision Table

IdeaFormula or fact
Excess mass\( M_e = \iint\Delta g\,dA/2\pi G \); actual \( M = M_e\rho_1/(\rho_1 - \rho_0) \)
Sphere\( GMz/(x^2+z^2)^{3/2} \); max \( GM/z^2 \); \( z = 1.305x_{1/2} \)
Horizontal cylinder\( 2G\lambda z/(x^2+z^2) \); max \( 2\pi G\Delta\rho R^2/z \); \( z = x_{1/2} \)
Slab\( 2\pi G\Delta\rho t = 0.04193\Delta\rho t \) mGal; no depth
Thin slab edge\( 2G\Delta\rho t[\pi/2 + \tan^{-1}(x/z)] \); 1/4–3/4 points \( 2z \) apart
Smith\( z \le 0.86A/A' \) (3D), \( 0.65A/A' \) (2D)
Magnetic dipole at pole\( \frac{\mu_0}{4\pi}m(2z^2 - x^2)/(x^2+z^2)^{5/2} \); zero at \( \sqrt2z \); \( z\approx 2x_{1/2} \)
Induced sphere at pole\( \Delta Z_{max} = \tfrac23\chi F(R/z)^3 \)
Latitudehigh equatorward, low poleward
Koenigsberger\( Q = M_r/M_i \)
Euler\( (\mathbf{r} - \mathbf{r}_0)\cdot\nabla T = N(B - T) \); SI magnetics 3, 2, 1, 0
Talwani\( \Delta g = 2G\Delta\rho\oint z\,d\theta \), polygons
Parker\( 2\pi G\Delta\rho e^{-kz_0}\sum k^{n-1}\mathcal{F}[h^n]/n! \)
Inversion\( (\mathbf{G}^T\mathbf{G} + \lambda\mathbf{W}^T\mathbf{W})^{-1}\mathbf{G}^T\mathbf{d} \); depth weighting
Bott\( t \mathrel{+}= (\Delta g_{obs} - \Delta g_{calc})/2\pi G\Delta\rho \)

Key Takeaways

What’s Next

Advertisement
Advertisement