Key words: Open Source Software, GRASS GIS, Modern Statistics, Spatial Data Analysis, Database Integration
The practice of geocomputation is evolving to take advantage of the peer-review qualities brought to software development by access to source code. The progress reported in this paper covers recent advances in the GRASS geographical information system, in particular the introduction of floating point raster cell values and an enhanced sites data format in the R data analysis language environment, and in their interfacing both directly and through the PostgreSQL data base system.
The R language is eminently extensible, providing for functions written by users, the dynamic loading of compiled code into user functions, and the packaging of libraries of such functions for release to others. R, unlike S-PLUS, its commercial equivalent, conducts all computation in memory, presenting a potential difficulty for analysis of very large data sets. One way to overcome this is to utilize a proxy mechanism through an external data base system, such as PostgreSQL; others will be forthcoming as the Omegahat project matures.
While R is a general data analysis environment, it has been extensively used for modelling and simulation, and for comparison of modern statistical classification methods with, for example, neural nets. It is well suited to assist in unlocking the potential of data held in GIS. Over and above classical, graphical, and modern statistical techniques in the base R library and supplementary packages, packages for point pattern analysis, geostatistics, exploratory spatial data analysis, and spatial econometrics are available. The paper is illustrated with worked examples showing how open source software development can benefit geocomputation.