2002 Annual

Wednesday, 16 January 2002
Development of Gtooln4 netCDF Conventions and Processor—Self Descriptive File Format for Multi-dimensional Data
Eizi Toyoda, Japan Meteorological Agency, Tokyo, Japan; and M. Ishiwatari, S. I. Takehiro, K. Nakajima, and Y. Y. Hayashi
Poster PDF (73.5 kB)
Scientific numerical data is growing larger and more complex to behave like storm in which scientists are easily got a lost. Data engineering technology is becoming more and more important to deal with those data storms in the fields of science.

It has been common that data are not self-descriptive; Scientists are supposed to remember how the data are obtained, how the data are processed, and how the figures are plotted from the data. When the amount of data and/or number of procedures are large, appropriate management of those information associated with data heavily bothers scientists. Here, metadata is referred to as associated data which descirbe those information associated with the core data. The question is `how metadata should be given?'

The authors have been trying to figure out a needed and sufficient list of metadata through the implementation experiment of "gtool4". Gtool4 is a general name of netCDF conventions and the toolkit for them. Gtool4 toolkit is a toolkit for visualization and simple analyses of multidimensional numerical data, The catchword of gtool4 is `figure from data with one-clicking'.

Gtool4 is named after GTOOL3. GTOOL3 is a toolkit written in FORTRAN 77, and reads/writes the sequential unformatted file. GTOOL3 has been used for data manipulation in CCSR/NIES GCM. By the use of netCDF, gtool4 is free from the problems of GTOOL3, such as binary format unportability, restriction into 3-dimensional space. Using Object-Oriented implementation with Fortran 90, the coding of gtool4 becomes much flexible and compact.

Design of gtool4 netCDF conventions consists of two parts: data variable and structured variable. Data variable is normal netCDF variable. Its attributes are almost compatible with existing netCDF conventions. Structured variable is used for storing graphic objects. The graphic object classes are designed for DCL (GFD-Dennou Club library). For compatibility, all attributes that are not used commonly among major meteorlogical netCDF conventions have their name with a common prefix 'gt_', so that they can be safely ignored with existing software and conventions. Both data and structured variables can be contained in one netCDF file, so that it will be `all-in-one' self-descriptive figure that can be reusable as numerical data.

Though all properties of graphic objects can be stored into structured variables, they does not simplify the initial creation of objects from data variable. The data variable should `know' how it will be rendered. So, the attibutes of graphic structured variable can attached also to the data variables, and they describes `the default graphic behaviour' of the data.

The software and document is available at http://www.gfd-dennou.org/arch/gtool4/.

Supplementary URL: http://www.gfd-dennou.org/arch/gtool4/