|
|
|
DATA MODELS For computer mapping to be effective all of the information from the original map needs to be in the database. This information must be contained in a central spatial database. The database includes coordinate data, which has to be drawn and feature information for the map such as roads, rivers, etc. Spatial datasets include information for locational and non-locational features and attributes describing these features, and they explain the relationships between them. The world represented by spatial databases must be abstracted, generalized or approximated to make a database because initially it is not in logical units. It is necessary to show an accurate view of the world by using databases and data models. A GIS can only be as efficient as the data provided to construct it. A digital cartographic database is full of information provided by the spatial datasets. Digital cartographic data is constructed when geographic data is first used to make a map and then digitized. On the other hand, digital geographic data is data gathered directly from the environment in digital form. Geographical reality, meaning the empirically verifiable facts about the world, often is not certain. A data model is a limited way to represent geographical reality. Spatial data also includes remotely sensed imagery and census tract descriptions, which are standard products. These datasets have specialized information such as seismic profiles, distribution of relics in an archaeological site or migration statistics. Geographical data can be stored as the tuple T=<x, y, z(1), z(2)…z(n)>. There is an infinite number of tuples available since x,y is continuous, defined as a field. It is difficult to represent the data z(1) to z(n) effectively in a data model. The goal is to have as many real variables represented by a single mapped variable. Many data models are based on discrete objects located in the plane, allowing spatial variation to be represented by a set of tuples <i, a(1), a(2)…a(m)> where i represents the object and a(1) to a(m) are the attributes. Many times objects are not clearly defined and are only generalizations or approximations of variation. Spatial autocorrelation plays a key role in the task of discretizing (similar meaning as data models) spatial variation. The similarities between the tuples T(1) and T(2) increase as they converge. There are two different ways to approach discretization: sampling exploits spatial autocorrelation by assuming that (x1,y1) and (x2,y2) must be more than a certain minimal distance apart before the associated tuples are substantially different; piecewise assumes that the plane can be partitioned into homogeneous, connected regions with variation within each region described by some simple function. (Goodchild, 1992) |