Corrections and Regional–Residual Separation
Gravity drift, tide, latitude, free-air, Bouguer and terrain corrections, density by Nettleton and Parasnis, magnetic reductions and regional-residual separation.
- 20 min
On this page · 13 sections
Prerequisites
- Gravimeters, Magnetometers and Data Acquisition: loops, drift, base stations and diurnal records.
- Gravity, Geomagnetism, Heat Flow and Isostasy: the normal gravity formula, free-air and Bouguer anomalies at Part B level, and isostasy.
- Geoid; Free-Air, Bouguer and Isostatic Anomalies: the Part C treatment of the anomalies whose field reduction this page carries out.
- Least squares fitting, revised in Numerical Ability for the arithmetic and in Filters, Deconvolution and Fractal Analysis for the filtering view.
The Hook: The Anomaly That Followed the Hills
Gravity Reductions in Order
graph LR
A[Dial reading] --> B[Calibration to mGal]
B --> C[Drift and tide]
C --> D[Tie to base: observed g]
D --> E[Subtract normal gravity: latitude]
E --> F[Add free-air: free-air anomaly]
F --> G[Subtract Bouguer slab]
G --> H[Add terrain: complete Bouguer anomaly]
H --> I[Isostatic correction: isostatic anomaly]$$\boxed{\Delta g_B = g_{obs} - g_n(\phi) + 0.3086\,h - 0.04193\,\rho\,h + TC} \qquad (h\ \text{in m},\ \rho\ \text{in g cm}^{-3},\ \text{all in mGal})$$1. Drift and tide
Drift. From repeated base readings, a drift rate is found and each station corrected by linear interpolation in time. Worked numbers are in Practice Q5 of the instruments page.
Earth tide. The Sun and Moon change gravity by up to about 0.3 mGal peak to peak over a day, with a period near 12 h. The effect is computed from astronomical tables for each station time and subtracted. If the base is reoccupied every hour or so, much of the tide is absorbed into the drift correction, but in precise work the two are removed separately.
2. Latitude (normal gravity)
Subtract the normal gravity \( g_n(\phi) \) from the reference-ellipsoid formula (see the Part B page). For a small survey it is enough to use the north–south gradient
$$\frac{\partial g_n}{\partial x} \approx 0.81\sin 2\phi\ \text{mGal km}^{-1}$$subtracted for stations poleward of the base (gravity increases towards the poles).
3. Free-air correction
Gravity falls off with height above the datum. For a sphere \( g = GM/R^2 \), so \( \partial g/\partial R = -2g/R \approx -0.308 \) mGal m\(^{-1}\); the value adopted for the ellipsoid is
$$\boxed{FAC = 0.3086\,h\ \text{mGal}}$$added for a station above the datum (it reads too little because it is farther from the centre). It treats the space between station and datum as empty, hence “free air”.
$$\Delta g_{FA} = g_{obs} - g_n + 0.3086\,h$$4. Bouguer correction
The rock between the station and the datum attracts the gravimeter. Approximating it by an infinite horizontal slab of thickness \( h \) and density \( \rho \):
$$\boxed{BC = 2\pi G\rho h = 0.04193\,\rho\,h\ \text{mGal}}$$subtracted for a station above the datum. For \( \rho = 2.67 \) g cm\(^{-3}\) it is 0.1119 mGal m\(^{-1}\); combined with the free-air correction the net elevation factor is about \( 0.3086 - 0.1119 = 0.1967 \) mGal m\(^{-1}\).
At sea the reduction is reversed in spirit: the water layer of depth \( d \) is replaced by rock, adding \( 0.04193(\rho_r - \rho_w)d \). That is why marine Bouguer anomalies are large and positive over the deep ocean.
5. Terrain correction
The slab is flat; real topography is not. A hill above the station pulls upward, reducing measured gravity; a valley below the station is a hole in the slab that the Bouguer correction wrongly subtracted. Both effects are corrected by adding: the terrain correction is always positive (on land, with the station as reference).
It is computed by dividing the surroundings into rings and sectors (Hammer charts, or a digital elevation model). For a sector of angle \( \theta \) between radii \( r_1 \) and \( r_2 \), whose mean height differs from the station by \( z \):
$$\boxed{\Delta g_T = G\rho\theta\left[(r_2 - r_1) + \sqrt{r_1^2 + z^2} - \sqrt{r_2^2 + z^2}\right]}$$Terrain corrections matter most in rugged ground and close to the station; inner zones need accurate heights.
6. Eötvös (moving platforms) and isostatic correction
- Eötvös for ship and aircraft data: see the instruments page.
- Isostatic correction removes the predicted effect of the compensating root (Airy) or density variation (Pratt) beneath topography. The result, the isostatic anomaly, is near zero over compensated regions. For exploration it removes a long-wavelength regional that has nothing to do with the target.
The anomalies and what they show
| Anomaly | Correction applied | Use |
|---|---|---|
| Free-air | latitude + free-air | marine and satellite work, isostatic state (near zero over compensated broad features) |
| Simple Bouguer | + slab | land exploration, flat areas |
| Complete Bouguer | + terrain | land exploration, rugged areas; mapping density variations below the datum |
| Isostatic residual | + isostatic root | removing the compensation regional |
Choosing the Bouguer Density
The slab and terrain corrections need the density of the rock between the station and the datum, which is rarely measured. Two field methods estimate it from the gravity data themselves.
Nettleton’s method
Run a gravity profile across a topographic feature (a hill or valley) that is not expected to have a geological structure beneath it. Reduce the profile with a series of trial densities (say 1.8 to 2.8 g cm\(^{-3}\)). The correct density gives the Bouguer anomaly least correlated with topography: too low a density leaves a positive anomaly over hills; too high a density gives a negative one.
Assumption: no real density structure mirrors the topography. If the hill is a resistant dense dyke, the method fails.
Parasnis’ method
Rearrange the Bouguer anomaly:
$$g_{obs} - g_n + 0.3086\,h = \rho\,(0.04193\,h - t) + \Delta g_B$$where \( t \) is the terrain correction computed for unit density (so \( TC = \rho t \)). Ignoring terrain (or keeping \( t \) in the \( x \) term), plot \( y = \Delta g_{FA} \) against \( x = 0.04193\,h \). If \( \Delta g_B \) is random with respect to height, the points fall on a straight line of slope \( \rho \); the intercept is the mean Bouguer anomaly.
Other routes: density logs and core measurements (direct but local), and seismic velocity converted with an empirical velocity–density relation.
Magnetic Data Reduction
$$\boxed{\Delta T = T_{obs} - \delta T_{diurnal}(t) - T_{IGRF}(x, y, z, t) - C}$$| Step | What it removes | Notes |
|---|---|---|
| Diurnal | time variation of the external field | base-station record; subtract the change since a reference time |
| Main field (IGRF) | the core field, a long-wavelength regional | International Geomagnetic Reference Field: a spherical harmonic model revised every five years under IAGA; evaluated at each station’s position, elevation and date |
| Heading and lag (airborne) | sensor and aircraft effects | test lines in opposite directions; position shift |
| Levelling (airborne) | residual errors between lines | tie-line crossover differences minimised |
| Constant \( C \) | arbitrary base level | anomalies are relative |
Points to note:
- No elevation correction in the gravity sense. The main field changes with height by roughly \( 3F/R \approx 3\times 50{,}000/6.371\times10^6 \approx 0.024 \) nT m\(^{-1}\) (from the \( r^{-3} \) dipole law), negligible for ground surveys and absorbed by the IGRF in airborne work. Topography matters for magnetics only through the magnetisation of the terrain itself.
- No latitude correction separate from the IGRF, which already contains the latitude dependence.
- Magnetic storms cannot be corrected reliably; data are reflown or remeasured.
- After reduction the anomaly is a total-field anomaly \( \Delta T \): its shape still depends on the inclination and declination of the main field and of the body’s magnetisation. Reduction to the pole (next page) is a transformation, not a correction.
Regional–Residual Separation
The Bouguer or magnetic anomaly is the sum of everything below the station. Deep, broad sources produce a smooth, long-wavelength regional; shallow, local sources produce short-wavelength residuals. The target is usually in the residual:
$$\boxed{\text{Residual} = \text{Observed anomaly} - \text{Regional}}$$There is no unique way to do this: the regional is a choice, and each method makes a different one.
graph TD
A[Regional-residual methods] --> B[Graphical: hand-smoothed contours or profiles]
A --> C[Averaging: moving average, Griffin ring method]
A --> D[Trend surface: least-squares polynomial]
A --> E[Wavelength filtering: low-pass, band-pass]
A --> F[Upward continuation as regional]
A --> G[Derivatives as residual enhancers]
A --> H[Modelling: compute and remove a known regional source]Graphical smoothing
Draw smooth regional contours (or a smooth line on a profile) by eye, guided by geology. Subjective, but it lets an experienced interpreter use outside knowledge, such as a known basin margin. Still used to check automated results.
Averaging methods
- Moving average. The regional at a point is the mean of the anomaly over a window of length \( L \) centred on it. It is a low-pass filter; \( L \) must be wider than the target anomaly, or part of the target ends up in the regional.
- Griffin’s ring method. The regional at a grid point is the average of values on a circle of radius \( r \) around it; the residual is the centre value minus that average:
With centre value 12.0 mGal and four ring values 10.0, 9.0, 9.5 and 9.5 mGal, the regional is 9.5 mGal and the residual 2.5 mGal. The choice of \( r \) controls which sizes of anomaly survive.
Trend-surface (polynomial) fitting
Fit a low-order polynomial in \( x \) and \( y \) to the data by least squares and call it the regional:
$$g_{reg}(x, y) = a_0 + a_1x + a_2y + a_3x^2 + a_4xy + a_5y^2 + \cdots$$A first-order surface is a plane (a uniform regional gradient). Higher orders follow the data more closely and absorb more of the target. A common refinement is to fit the polynomial only to stations outside the target area.
Wavelength filtering and upward continuation
In the wavenumber domain the separation is a filter: the regional is the low-wavenumber part. Options:
- Low-pass / high-pass / band-pass filters with a cut-off wavelength chosen from the power spectrum, where the deep and shallow source ensembles show different slopes (see Continuation, Derivatives and Frequency-Domain Methods).
- Upward continuation: continue the field upward by a height \( h \); short wavelengths are attenuated by \( e^{-kh} \), and the result is a physically meaningful regional (it is the field that would be measured at that height). Residual = observed − upward-continued.
These are objective and repeatable, but sources at different depths have overlapping spectra, so no filter separates them cleanly.
Derivatives as residual maps
The second vertical derivative (and other derivatives) suppresses broad anomalies and sharpens local ones. It is not a residual in the strict sense (its units are mGal per km\(^{2}\)) but is used the same way: to locate shallow sources and their edges. Details on the next page.
Comparing methods
| Method | Objective? | Strength | Weakness |
|---|---|---|---|
| Graphical | no | uses geology | not reproducible |
| Moving average / Griffin | yes (given window) | simple | window choice controls result; distorts near edges |
| Polynomial trend | yes (given order) | smooth, analytic regional | absorbs target if fitted over it; edge effects with high orders |
| Wavelength filter | yes (given cut-off) | fast on grids | ringing with sharp cut-offs; overlapping spectra |
| Upward continuation | yes (given height) | physically meaningful | still a choice of height |
| Derivatives | yes | sharp edges | amplify noise; not true residual units |
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Free-air correction: removes the effect of height (distance from centre) | Bouguer correction: removes the attraction of the rock slab below the station | |
| Free-air anomaly: small over isostatically compensated broad features | Bouguer anomaly: strongly negative over compensated mountains (the root) | |
| Terrain correction: always added (hills and valleys) on land | Bouguer correction: sign depends on station above or below datum | |
| Simple Bouguer anomaly: flat slab | complete Bouguer anomaly: slab + terrain | |
| Drift (instrument) | tide (Sun and Moon, computed) | |
| Diurnal correction (external field, time) | IGRF removal (core field, position and date) | |
| Regional: deep, broad, low wavenumber | residual: shallow, local, high wavenumber | |
| Nettleton: minimise correlation with topography (profile) | Parasnis: slope of free-air anomaly vs \( 0.04193h \) |
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
- Is the terrain correction on land added or subtracted?
- Write the Bouguer correction for a slab of density \( \rho \) (g cm\(^{-3}\)) and thickness \( h \) (m) in mGal.
- Name the model used to remove the core field from magnetic survey data.
Answers
- Added, for both hills and valleys around the station.
- \( 0.04193\,\rho\,h \) mGal (\( 2\pi G\rho h \)).
- The International Geomagnetic Reference Field (IGRF).
Level 2: Part B Style
Practice Q1. A station is 120 m above the datum. With a reduction density of 2.40 g cm\(^{-3}\), the combined free-air and Bouguer elevation correction is:
(a) +25.0 mGal (b) +37.0 mGal (c) +49.1 mGal (d) −12.1 mGal
Answer: (a) +25.0 mGal
Free-air: \( 0.3086\times 120 = 37.03 \) mGal (added). Bouguer: \( 0.04193\times 2.40\times 120 = 12.08 \) mGal (subtracted). Net \( 37.03 - 12.08 = +24.96 \approx +25.0 \) mGal.
Option (b) is the free-air term alone; (c) adds the Bouguer term instead of subtracting it; (d) is the Bouguer term with the wrong sign and no free-air.
Practice Q2. In Nettleton’s method the correct reduction density is the one for which the Bouguer anomaly along a profile across a hill:
(a) has the largest amplitude (b) is least correlated with the topography (c) equals the free-air anomaly (d) is zero at the summit
Answer: (b)
The method assumes no density structure follows the topography, so the right density leaves no topographic signature in the Bouguer anomaly.
Practice Q3. A ship measures gravity over water 4,000 m deep. To compute a Bouguer anomaly with rock density 2.67 and water density 1.03 g cm\(^{-3}\), the correction added is about:
(a) 112 mGal (b) 172 mGal (c) 275 mGal (d) 448 mGal
Answer: (c) 275 mGal
\( 0.04193\times(2.67 - 1.03)\times 4000 = 0.04193\times 1.64\times 4000 = 275.1 \) mGal.
Option (d) uses the full rock density (\( 0.04193\times 2.67\times 4000 = 447.8 \)); option (b) uses only the water density (\( 172.7 \)).
Practice Q4. Which of the following is NOT part of the routine reduction of a ground total-field magnetic survey?
(a) Diurnal correction from a base-station record (b) Removal of the IGRF (c) A free-air correction for station elevation (d) Removal of a constant base level
Answer: (c)
There is no free-air step in magnetics; the main-field vertical gradient (about 0.02 nT m\(^{-1}\)) is negligible on the ground and handled by the IGRF in airborne work.
Level 3: Part C Style
Practice Q5. Free-air anomalies \( y \) (mGal) at four stations of height \( h \) are: \( h = 50 \) m, \( y = 3.1 \); \( h = 150 \) m, \( y = 12.3 \); \( h = 250 \) m, \( y = 21.5 \); \( h = 350 \) m, \( y = 30.7 \). Using the Parasnis method (ignore terrain), the reduction density and the mean Bouguer anomaly are:
(a) 2.19 g cm\(^{-3}\), −1.5 mGal (b) 2.67 g cm\(^{-3}\), −2.5 mGal (c) 2.19 g cm\(^{-3}\), +1.5 mGal (d) 0.092 g cm\(^{-3}\), −1.5 mGal
Answer: (a) 2.19 g cm−3, −1.5 mGal
\( x = 0.04193h \): 2.0965, 6.2895, 10.4825, 14.6755.
The \( y \) values rise by 9.2 mGal for each 100 m, that is for each 4.193 in \( x \). Slope \( = 9.2/4.193 = 2.194 \) g cm\(^{-3}\).
Intercept: \( 3.1 - 2.194\times 2.0965 = 3.1 - 4.600 = -1.50 \) mGal (same from any point, since they are collinear).
Option (d) is the slope against \( h \) itself (9.2 mGal per 100 m = 0.092), forgetting to divide by \( 2\pi G \).
Practice Q6. Consider these statements about regional–residual separation:
- A higher-order trend surface fitted over the whole area tends to absorb part of a local anomaly into the regional.
- Upward continuation gives a regional that is the field actually expected at a greater height.
- The second vertical derivative map has the same units as the Bouguer anomaly.
- Different separation methods applied to the same data generally give different residuals.
Which are correct?
(a) 1 and 2 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
- Correct: more terms, more flexibility, more of the target in the regional.
- Correct: upward continuation is a solution of Laplace’s equation, so it is a physical field.
- Wrong: a second derivative has units of mGal per length squared (mGal km\(^{-2}\)).
- Correct: the regional is a choice; residuals depend on it.
Practice Q7. A gravity station lies in a valley. A full ring of terrain between 0.5 km and 1.0 km from the station has a uniform mean height 100 m above the station level. With \( \rho = 2.67 \) g cm\(^{-3}\), its terrain correction is closest to:
(a) 0.10 mGal (b) 0.55 mGal (c) 1.10 mGal (d) 2.20 mGal
Answer: (b) 0.55 mGal
Full ring, \( \theta = 2\pi \): \( G\rho\theta = 6.674\times10^{-11}\times 2670\times 2\pi = 1.1196\times 10^{-6} \) s\(^{-2}\).
Bracket: \( (1000 - 500) + \sqrt{500^2 + 100^2} - \sqrt{1000^2 + 100^2} = 500 + 509.90 - 1004.99 = 4.91 \) m.
\( \Delta g_T = 1.1196\times 10^{-6}\times 4.91 = 5.50\times 10^{-6} \) m s\(^{-2}\) \( \approx 0.55 \) mGal, added. Options (c) and (d) double and quadruple it (for example by using \( 4\pi \) for a full ring, or by doubling the bracket).
Quick Revision Table
| Idea | Formula or fact |
|---|---|
| Bouguer anomaly | \( g_{obs} - g_n + 0.3086h - 0.04193\rho h + TC \) |
| Free-air | \( 0.3086 \) mGal/m, added above datum |
| Bouguer slab | \( 2\pi G\rho h = 0.04193\rho h \) mGal; 0.1119 mGal/m at 2.67 |
| Net elevation factor | 0.1967 mGal/m at 2.67 |
| Latitude (local) | \( 0.81\sin 2\phi \) mGal/km, subtract poleward |
| Tide | up to about 0.3 mGal, computed |
| Terrain sector | \( G\rho\theta[(r_2 - r_1) + \sqrt{r_1^2+z^2} - \sqrt{r_2^2+z^2}] \), always added on land |
| Marine Bouguer | add \( 0.04193(\rho_r - \rho_w)d \) |
| Nettleton | density giving least correlation with topography |
| Parasnis | slope of \( \Delta g_{FA} \) vs \( 0.04193h \) = \( \rho \) |
| Magnetic reduction | \( T_{obs} \) − diurnal − IGRF − constant |
| Residual | observed − regional; method-dependent |
| Griffin | centre value − mean on ring |
Key Takeaways
What’s Next
- Continuation, Derivatives and Frequency-Domain Methods: upward and downward continuation, derivative maps, reduction to the pole and spectral depth estimates.
- Interpretation of Simple Bodies, Forward Modelling and Inversion.
- Gravity and Magnetic Methods: Quick Revision.