• Direct and Radial gridding methods combined to form new Global method that performs many localized direct solutions followed by secondary gridding where needed. This provides a solution that has the smoothness and appeal of Direct with the ability to handle 100,000s of points and faults.
• Added new optional initial step that assigns z values to the endpoints of faults this improves connectivity and smoothness with sparse data.
• Secondary estimates improved to make Radial results look more like Direct.
• If a boundary is specified, an extra step is performed to ensure all nodes interior to the boundary are defined.  Another optional step is also now available that will fill even the interior of empty fault blocks with a constant based on neighboring blocks. This is useful for doing conformal mapping.
• Estimates of surface/fault intersections improved to improve smoothness of surface near faults. Also does better job of dealing with intersecting faults when interpolating along the edges
• Improved resolution of fault-fault intersections, resulting in more cases being resolved correctly.
• Also have added several new features to ensure faults intersect each other properly.  Faults that terminate internal to another are trimmed, those that terminate close are extended and then trimmed.  Similar changes are applied to faults near polygon boundary.
• All interpolation methods can now be assigned boundary control based on one of four approaches: unassigned, concave shell, user-supplied polygon, or entire grid space.  Use of shell includes options to control degree of concavity and the inclusion of fault endpoints.  Both shells and user polygons can be extended outward and then smoothed.
• Display of lighted 3D images has been facilitated by providing convenient access to gradient values at corners of triangles provided for drawing that are consistent with surface.
• New functions are available that allow surface to surface operations, both arithmetic and logical, logical being things like intersect, keeping selected portions and generating intersection curves at boundaries.
• Several single surface operations are also available. Smoothing is one of these, the best method uses a 9 node filter applied iteratively that minimizes the local curvature. This produces very natural surfaces.

This image shows how well the new improvements work on a very complicated fault.

The following image demonstrates boundary improvements in GTGRID. This allows Z values to be attached to boundary polygons so it provides a 3-dimensional boundary. If you look close at the image below, it is clear that the surface shape is linked to the boundary as well as to the internal points. This same capability is also available along fault boundaries.

Additional features added include the ability to automatically clip intersecting fault polygons and to fill them with uncountoured colored surfaces. These are formed as a set of triangles appropriate for 3-D display as well.

This example shows intersection of a surface and a trend fitted to it. White curves are the intersections; this display shows portions of both that are the greater.

Another useful  improvement to the product is a new option that produces a boundary polygon based on the point distribution that is used to produce a surface, as shown in this example.

This is formed by first forming a convex hull from the pointset, eliminating all edges that exceed a specified distance to form a concave hull, then expanding the resulting polygon outward by a specified distance to obtain the result shown.

And the latest update is the support of creases. These are faults with z values attached and can be used to describe terrain features like road edges and ridge lines. Both are shown in this example.

I think this makes GTGRID very attractive to a new class of users that wish to generate digital terrain models from relatively sparse data like contour maps.

The display of faulted surfaces requires special treatment of the edges. The grid cells that are divided by a fault or crease must be triangulated, as shown here, in order to be properly displayed. In addition, GTGRID offers subdividing of normal cells into smaller sub-cells in order to implement non-linear interpolation across the cells. A bi-cubic interpolation is performed within each normal cell so that contours are smooth.