GM-3D Gravity Modeling Software

Welcome to the home of the GM3D program! GM3D stands for Gravity Modelling in 3 Dimensions, a powerful computer program that extracts geologically meaningful density anomalies from gravity data.


GM3D was originally written to help model the gravity effects of shallow features like the ones in this diagram:




But over the last year or so it has been modified so that it can be applied to crustal scale models like this 600km-wide model of the South Island plate boundary, New Zealand:



In both cases, GM3d helps the user include everything thing they might already know about the density structure (topography, buildings, geological layers etc), and then it uses the residuals to estimate any remaining density anomaly. Because there are generally any number of models that would agree with the data GM3D employs some clever constraints to ensure that it produces geologically meaningful ones; i.e. bodies with minimum volume, minimum mass, maximum depth etc

In the first example, the geometry of the topography and buildings are always measurable. Sometimes the subsurface is at least partially known as well, e.g. buried valleys can be profiled using seismics or resistivity, caves and mine workings may be at least partly mapped, and geological contacts may be known both from surface outcrops and drillholes. The problem is how to include this information, and how best to use the gravity anomalies to increase knowledge of the the parts of the model where the structure was unknown.

In the second example, in the New Zealnad model, the topography and bathymetry are known and the geometry of the sediments and crustal root have been determined by the SIGHT experiment. The quetsion is whether or not sensible values for density can reconcile this model with the observed gravity field across the region and if not; could some mantle anomaly help fit the data?

How does it work?

In brief, GM3D uses simulated annealing to solve for anomalous density structure, and this allows more flexibility for building in constraints than other programs allow. E.g. you can solve for the best-fitting continuous body that passes through some set of coordinates, or the best-fitting surface layer etc. It also lets you forward-model anything you might know already, including bodies that extend above the topgraphy, (effectively a terrain correction for the walls of buildings). The model is built from a hybrid of rectangulare blocks and smoothly varying surface-layers (blocks are inefficient at modelling shallow topography and shallow structure and in general these is known-apriori anyway). The effect of each model parameter is calculated at each station and held in RAM as, pointers to pointers to pointers to pointers!! This means that the results of any change to the model can be looked up very quickly so the new residuals can be displayed nearly instantaneously. In this way users quickly forward-model structures, built up from recatngular elements, that fit the observations.

The problem with rectangular elements is that it's impossible to deal with smoothly varying structures without making the block sizes so small that building a model becomes cumbersome.

GM3D estimates the gravity effect of bicubic splines to get around this problem. In this way it allows precise modelling of density right up to the topographic surface if a DEM is avalaiable. It also means that smoothly varying surface layers can be included in the model with ease.








email: Robert Davies for more information