PROGRESS IN METRICS DEVELOPMENT TO MEASURE POSITIONAL ACCURACY OF SPATIAL DATA

The issue of quality in spatial data is a complex one. The ICA Commission in Spatial Data Quality (1995) recognizes three parts of the specification and use of spatial data quality information: (1) defining the elements of spatial data quality; (2) derivation of an easily understood index of spatial quality, composed of a series of metrics that measure the elements of spatial data quality; and, (3) presenting or rendering the known quality for visualization. Efforts in the past twenty years have resulted in the definition of the following elements of spatial data quality: lineage, positional accuracy, attribute accuracy, completeness, logical consistency, semantic accuracy, and temporal information. In the same time period, the second part of spatial data quality information has experienced little progress from the viewpoint of metrics development for the different elements of spatial data quality. This paper describes our efforts in developing metrics for positional accuracy. Current USA spatial positional accuracy standards deal only with the evaluation of points. Examples of this are the National Map Accuracy Standard (1947), and the Geo-spatial Positional Accuracy Standards Part 3: National Standard for Spatial Data Accuracy (FGDC-STD-007.3-1998). Our research focus at this point is the development of metrics to measure the positional accuracy of linear features. Linear features are more complex than cartographic points and are major components of spatial databases. Generally, there are more linear features than cartographic points in most general-purpose topographic maps. Four quality measures have been developed in our research: Bias Factor, Distortion Factor, Fuzziness Factor, and Generalization Factor. The Bias Factor compares the relative location of a linear feature that is less precise to another similar feature that is more precise by super-imposing them and comparing the number and lengths of sub-segments of the less precise feature that fall to the right of the more precise feature with those that fall to the left of the more precise feature. The distortion factor compares the standardized parameterization of two linear features of the same region of the Earth using the equivalent standardized locations. Fuzziness is the factor related to the definition and identification of the end points of two linear features to be compared. The generalization factor compares the lengths of two equivalent linear features. This paper describes in detail these quality measures, their testing, and use to evaluate the positional accuracy of linear features.