GIS I. Final Exam Questions

Answer three out of four questions. Make sure you clearly mark which questions you answered. Your answers should be typed and double spaced and any relevant diagram neatly drawn. Please make an effort to be thorough yet concise (no more than two pages per question, not including graphics). Use all resources available to you including texts, manuals, lab exercises, and help files. The exam is due on December 8 by 11:59 PM.

1. The text (p232, 233) listed nine methods for testing spatial relations between geometric objects such as points, curves (lines), and surfaces (polygons). Some of these methods have recently been implemented in ArcGIS Geodatabase as topology rules, which define the relationships between collections of one or more geometric objects. Figure 11.4 (p233) shows the possible relations of two geographic database operators: contains and touches. An example relation is points contain points, which can be used to validate that the point features in two data layers coincide. For example, points representing the location of hospitals “contain” points showing the location of “Neonatal Intensive Care Facilities (Units)” or street intersections (points) “contain” traffic lights (points). Select and describe 3 other relations in Figure 11.4 and give examples of how they can be used to relate GIS data layers.

2. Spatial analysis is the crux of GIS. What is spatial analysis? Describe the following spatial analytical procedures and give examples of how they can be used: a) measuring compactness of polygons, b) buffering, and c) point in polygon analysis.

3. What is a semivariogram? What information does it tell us? How is such information used in Kriging? Include discussions on the basic components of a semivariogram model (i.e., range, sill, and nugget) and the definitions of isotropy/anisotropy.

4. Statistical models can be used to test hypotheses and draw inferences. The validity of statistical models depends critically upon the accuracy of assumptions. For example, assumptions for regression analysis include no multicollinearity (p117, 118) and no autocorrelation in the observed data. Explain why spatial dependence and spatial heterogeneity often make spatial information less suitable for inferential testing. What are the ways to deal with this problem?