This question comes from a discussion in Where to find author/creator of tag?.
I think there is a need of making more explicit the kind of transformation we are referring to in questions tagged as transformation. I created the tag affine-transformation because the transformation tag does not express the same meaning.
'Transformation' is a broad term. affine-transformation helps narrowing down topics and, thus, searches. If tags are to provide users (and search engines) with semantics, I do not see why not expressing different meanings with different tags. I could think of datum transformations, geometry transformations (e.g., scaling, shifting, rotating, and skewing geometries), and PolyGeo mentioned also vertical transformations.
In concrete, to enrich questions about transformations, we could start using (at least) these tags:
- datum-transformation (which could be a synonym of transformation, because it's arguably the most common meaning intended by the tag transformation). According to (Iliffe and Lott, 2008), datum transformation is a misnomer and one should use coordinate transformation instead. (Knippers, 2009) uses the term datum transformation to refer to transformations between 3D CRS. In summary, datum-transformation could be a synonym of coordinate-transformation, both defined as a "Change of coordinates from one CRS to another in which the CRSs are based on different datums." (Iliffe and Lott, 2008, p.177)
- vertical-transformation Change of coordinates from one vertical CRS to another. Vertical CRSs have one coordinate (height) based on a vertical datum, which is usually related to mean sea level. (Iliffe and Lott, 2008, p.180)
- affine-transformation A transformation between 2-dimensional plane (cartesian) coordinates that can rotate, shift, scale (even applying different factors on each axis) and skew geometries.
- polynomial-transformation Is a non-linear transformation that translates, rotates and changes scale in a variable way. At least 6 control points are required to calculate the 12 parameters of a (second-order) polynomial transformation. Could be synonym of rubbersheeting.
Perhaps when one of the aforementioned tags appears on a question, most probably the tag transformation shouldn't be there.
How are datum and affine/polynomial transformations different?
Unlike coordinate (datum) transformations, affine/polynomial transformations are local in nature. For coordinate transformations, national mapping agencies promote parameters to the public, whereas for affine/polynomial transformations, the users of the data calculate their own parameters, and those are only valid for that particular case. Moreover, in affine/polynomial transformations, one or two of the underlying CRS may be unknown. It is often the case that a CRS is missing or, due to some errors in the acquisition or processing of the data, a CRS is actually wrong, which makes it impossible to apply a coordinate transformation.
Transformations and conversions
Transformation and conversion are different things in the context of coordinate reference systems. Transformations take place between different datums, conversions take place between the same datum.
Regarding geometries, it seems that geometry conversion is preferred over geometry transformation (which I proposed in an earlier version of this post) in GIS terminology. Therefore, I propose to use geometry-conversion like this:
- geometry-conversion A change from a geometry type to another, e.g., points to lines, lines to polygons, polygons to points, and the like.
What do you community think?
References
Iliffe, J. and Lott, R. Datums and map projections: For remote sensing, GIS and surveying. Chapter 4, 2008.
Knippers, R. Geometric aspects of mapping. International Institute for Geo-Information Science and Earth Observation (ITC), Enschede. Section 5.4. 2009. http://kartoweb.itc.nl/geometrics/Coordinate%20transformations/coordtrans.html
projection-conversionswould fit within this proposal. Tks.projection-conversionsanddatum-transformationis that conversions are reversible with no accuracy loss. Datum transforms will always lose accuracy; how much depends on the method.