Google Sky paper in arXiv.org

The Daily ACK‘s Alasdair Allan has a great find on arXiv.org, the online repository for physics papers:

Sky in Google Earth: The Next Frontier in Astronomical Data Discovery and Visualization

It’s by the people who built the Sky functionality in Google Earth, and the whole thing is a worthy read. Some of the highlights that caught my eye:

The underlying imagery used in Sky resides in a lat/long projection (Snyder 1926). This results in substantial distortion at the poles even when re-projecting onto the sphere. Thus, for regions within five degrees of the pole we replace the original images with a lower resolution view of the sky derived from the Tycho II catalog (Høg et al. 2000). [...]

The underlying projection and registration of images in Sky is based on the technology used in Google Earth. This provides a mature visualization platform on which to develop Sky, a very large user base, a simple but extensible interface and a well-defined and on-going support and development mechanism. It does, however, lead to a small number of trade-offs — related to the way geospatial data is served — that were made in the course of adapting the system to serve astronomical data. [...]

As with Google Earth, the basemap is not static and we anticipate continued improvements as we refine our treatment of the current imagery and new large-scale datasets become available. [...]

It is worth noting that planets’ and the Moon’s icons are not scaled according to their actual appearance on the sky so that they can be more easily distinguished from the background stars. [...]

(If that was the reasoning behind not showing real-size icons then I think this was a mistake. In the Google Earth environment, you can always zoom in more to see detail if you wish. More likely, the limitations on controlling the size of an icon via KML as rendered by Google Earth is to blame. Another problem that will need to be solved: Which planetary icon comes before which, and when? Layers in Google Sky behave like Ptolemaic spheres inside spheres, but planetary orbits do not. Sometimes Venus occults Mercury, sometimes Mercury occults Venus…)

The KML tag name GroundOverlay is inherited from Earth, but these are simply images projected against the basemap in Sky. [...]

(Pace Vint Cerf:-)

As mentioned previously, since Sky shares a rendering engine with Earth, the geometry of the sky is, in fact, a slightly oblate spheroid (technically, the WGS84 projection). The GIS community has developed a number of tools for handling re-projections of images using this geometry, most notably the Geospatial Data Abstraction Library (GDAL). This software can warp images from one projection (including a tangent plane) to another and encodes the geometric information necessary for registering the image on the sphere in either an image header (analogous to a FITS header) or an associated “world file”, depending on the image format. While users may want to become more familiar with the GDAL software themselves, we provide a simple, open source tool wcs2kml which will read in an image in a variety of formats and WCS information from a FITS header and generate a properly warped image and overlay KML for you. This tool is available in both Python and C++ versions.

[What! The sky is not a perfect spere? :-)]