necessity for physical presence at a center. Implementing such a system could help researchers in the geographical sciences gain access to, and use, the social data they need to address key questions about the changing human geography of the planet.
An NSF report (2003) argued that in the future, science will require a new kind of infrastructure—a cyberinfrastructure—to respond effectively to emerging challenges. Cyberinfrastructure encompasses the computers, networks, and storage devices of scientific computing; the sensors, software, tools, and communications of a networked scientific world; and the virtual communities that have to collaborate to make substantial progress. All of the foregoing require massive investment. Substantial investments have already been made by the NSF in building cyberinfrastructure through awards for the acquisition of high-performance computing systems and for the building of virtual communities of networked scholars. The 2003 NSF report has also been followed by others offering more specialized perspectives on the role of cyberinfrastructure in the social sciences (Berman and Brady, 2005), the humanities (American Council of Learned Societies, 2006), and several individual disciplines.1 Yet to date little of the cyberinfrastructure discussion or investment has focused on the geographical sciences.
Because the geographical world is infinitely complex, any attempt to represent it, whether in the form of a paper map or a digital database, requires making difficult decisions about what to include and what to leave out. One strategy is to ignore spatial detail, rejecting information about variation that occurs over distances smaller than some declared spatial resolution. A related approach is to lump together approximately homogeneous areas or regions and ignore all variation within them. Other types of variation can be adequately captured by taking samples at appropriately spaced measurement points. Numerous commercial firms have adopted their own proprietary formats in the past (e.g., ESRI), and several national and international organizations have promulgated standards (e.g., Open Geospatial Consortium and the Federal Geographic Data Committee). As a result, many hundreds of geographical data formats now exist, creating headaches for anyone wanting to share data or integrate data from multiple sources (Goodchild et al., 1999).
Nonetheless much progress has been made in achieving a greater degree of interoperability. The Open Geospatial Consortium was formed in the 1990s to address interoperability issues, and most high-income countries are now actively engaged in developing their spatial data infrastructures, following the guidance provided by a report under the aegis of the NRC’s Mapping Science Committee (NRC, 1993). In this regard, the geographical sciences are at a distinct advantage relative to many sciences, because not only researchers but also government agencies, corporations, and nongovernmental organizations are willing to support and invest in steps to improve the sharing of geographical data. The creation of mashups by combining geographically referenced information from different Web sites is one demonstration of the power of this new level of interoperability (Chapter 10).
In other respects, however, the research community is unable to benefit by hitching its wagon to broader efforts. Although great progress has been made in the sharing of data, there has been little comparable progress in the infrastructure needed to share the tools of analysis or the software of simulation modeling. The computer codes being created to model complex geographical systems are largely developed in low-level languages and are unlikely to be reusable by others. There are no digital libraries of tools and software, and no standards for their documentation and description. There are no GIS software environments designed specifically for the needs of K-12 education (NRC, 2006), and the products on which most of our students are trained were too often developed to support inventory and management rather than scientific research. By and large, these tools have not received the kinds of funding needed to ensure that they are rigorously engineered, robust, well documented, and widely disseminated.
By its very nature, geographical information is distinct from the kinds of information acquired and analyzed in those disciplines that proceed by controlled experiment. The foundations of statistical analysis were developed in disciplines such as psychology, where it was reasonable to believe that the members of a sample of subjects had been randomly and independently
For a complete listing, see www.nsf.gov/crssprgm/ci-team/ (accessed December 15, 2009).