Database Reference
In-Depth Information
12.5.1 Astronomy
Astronomy has used, for over 25 years, the Flexible Image Transport Sys-
tem (FITS) standard
65
for platform-independent data interchange, archival
storage of images and spectra, and all associated metadata. It is endorsed
by the U.S. National Aeronautics and Space Administration (NASA) and the
International Astronomical Union. By design, it is flexible and extensible and
accommodates observations made from telescopes on the ground and from
sensors aboard spacecrafts. Briefly, a FITS data file is composed of a fixed
logical record length of 2,880 bytes. The file can contain an unlimited number
of header records, 80 bytes long, having a “keyword=value” format and writ-
ten as ASCII strings. These headers describe the organization of the binary
data and the format of the contents. The headers are followed by the data
themselves, which are represented as binary records. The headers record all
the metadata describing the science data. Figure 12.2 depicts an instance of
this for the metadata of an image of the galaxy NGC 5584 measured by the
Two-Micron All Sky Survey (2MASS).
66
The complete metadata specification of an astronomical observation not
only includes obvious quantities such as the time of the observation, its posi-
tion and footprint on the sky, and the instrument used to make the observa-
tion, but also includes much information custom to particular datasets. FITS
therefore was designed to allow astronomers to define keywords as needed.
Nevertheless, FITS has predefined reserved keywords to describe metadata
common to many observations. For example, the relationship between the
pixel coordinates in an image and physical units is defined by the World Co-
ordinate System (WCS),
65
which defines how celestial coordinates and pro-
jections are represented in the FITS format as keyword=value pairs. These
keywords are listed in sequence in Figure 12.2: They start at CTYPE and end
at CDELT.
Tabular data and associated metadata are often represented in FITS format,
but the FITS standard is poorly specified for tabular data. Instead, tabular
material, whether they are catalogs of sources or catalogs of metadata, are
generally stored in relational databases to support ecient searches. Transfer
of these data is generally in the form of ASCII files, but XML formats such
as VOTable
67
are growing in use. Metadata describing catalog data can be
grouped into three types: semantics, which describe science content (units,
standards, etc.); logistical, which describe the structure of the table (data
types, representation of null values, etc.) and statistical, which summarize
the contents (number of sources, ranges of data in each column, etc.). The
absence of standard specifications for columns has complicated and confused
data discovery and access to tabular data. Work originating at the Centre
de Donnees Astronomiques de Strasbourg (CDS)
68
has been embraced by the
International Virtual Observatory Alliance (IVOA)
69
as part of an interna-
tional effort to rectify this problem. When this work is complete, all column
descriptors will have a well-defined meaning connected to a hierarchical data
model.
