@thomas.engel’s XYZ data is in nice orderly rows (rasters) and columns like a 2D image. Extending the image analogy by giving each pixel a gray-scale “depth”, height maps essentially provide the same information.
You may, indeed
I think that the accuracy is dependent on the original input datum points. Whether it is rasterized or vectorized, the result is still just an approximation of the real world using best-fit algorithms through the dataset.
The Grand Canyon model I indicated in a previous post was made from 1/3 arc-second (~10 meter) data from the nationalmap data source. There is also some 1/9-arc-second data available for certain areas of the US (Boulder, CO, for example). This is about 3.4 meters of resolution. There is also 1-meter data available in Erdas IMG file format. The most accurate data seems to be obtained by low-flying planes using LIDAR or other sensing schemes. Personally, I find it frustrating that I can’t simply download an area at 1mm resolution to get a really high-res definition
As you know, this gets more complicated when translating between the various planar, cylindrical, and spherical coordinate systems that have been used in one context or another. The OP’s XYZ data uses a perfectly rectangular grid to map a portion of a spherical surface. It also specifies the elevation in meters to four decimal places. This implies a precision of 0.1mm in the vertical axis(?). For more precise work, I think that site scanners (such as Trimble makes), are the most accurate solution by their very nature. However, at the end of the day, it all comes back to some sort of interpolative manipulation of the point cloud to accurately depict the real world.