| Modern telecommunications infrastructure, such as digital switches and fiber optic cables, are the major determinants of the spatial structure of the information society, and are the focus of many policies and regional planning activities. The need for assessing the effectiveness of telecommunications infrastructure requires spatially integrating infrastructure data with demand data. These data refer to incompatible sets of spatial entities and incompatible concepts, which imposes a challenging data integration problem in GIS. There has been a lack of effective methodologies and empirical studies that particularly deal with this type of problems in a GIS environment. The present study identifies the telecommunications infrastructure assessment as an example of a class of data integration problems which can be solved by spatial data modeling approach proposed in this paper. The feasibility of the approach is demonstrated by a full specification of a data integration model in terms of a set of assumptions, constraints, representation, and transformations that compatibility can be achieved between telecommunications infrastructure data and demand data. The linking of infrastructure and demand data is made possible by computing a surface representation of access bandwidth using a spatial transformation modeled by a combination of nearest neighbor, distance decay and buffer zone operations. Based on the data integration model, a set of spatial analysis methods is developed for the discovery of infrastructure gaps and biases. A case study is presented on the assessment of Pennsylvania's telecommunications infrastructure. The results show that Pennsylvania's digital, fiber infrastructure is able to provide T1 (1.544 Mbps) to over 62% of geographical areas, 96% of schools, 94% of residential populations, and 99% of businesses. However, the infrastructure shows significant gaps when a minimum access bandwidth requirement of DS2 (6.312 Mbps) is imposed. The degree of infrastructure biases was compared across geographical regions as well as customer groups, using an access curve as an efficient pattern descriptor. |
