EHASL — Environmental Health Advanced Systems Laboratory

Scale and Representation in GIS and Remote Sensing: Applications of Geographical Principles to Environmental Health Research

Representation

Spatial Data Models — The term "data model" is often used in the GIS context, but should probably more explicitly be termed "spatial data model" since it is space, not data, that is being modeled.

GIS data models — the most commonly used representations of space in a GIS are the Raster and Vector data models.

Raster — grid tessellation of space, grid cells serve as the basic units of analysis. Example: raster pixels of remotely sensed imagery, digital elevation model coverage.

Vector — point, line and polygons based on continuous geometry of space. Analogous to traditional cartographic model.

Other models sometimes incorporated include quadtree, a raster data structure incorporating compression to eliminate redundancy, and TIN -- Triangulated Irregular Network -- an efficient way of representing elevation data often used for terrain analysis.

Some GIS packages contain algorithms for translating between formats, e.g. raster --> vector, vector --> raster, point --> TIN, etc. Some error may be introduced by these data transformation processes.

The data model representation used in a GIS can affect analysis results. Example of determining area using raster and vector models. Obviously scale plays a role here too -- different raster cell size resolutions may allow more precise representations resulting in increased accuracy in determining area.

Why is representation in GIS crucial? Why does the choice of spatial data model matter? When using GIS results for management, decision makers need to fully understand the ways in which representation can affect results. Are data and analyses appropriate and accurate? One important appropriateness issue is that of scale and resolution.

Scale/Resolution

Meanings of scale: cartographic, geographic extent, spatial resolution, operational scale

Cartographic scale: traditional map scale ratio relates the size of a feature on the ground to the size of a feature on the map. USGS topographic maps are commonly used at a 1:24,000 scale, which show a fair amount of detail including roads, water bodies, and in some areas actual structures such as houses or churches.

Geographic Extent: Refers to the size of the study area. For example, a study can be regional scale or global scale. The extent of your study area and/or its subsets can affect the analysis results. Different results might be obtained when looking at cancer incidence in one county versus statewide.

Spatial resolution: Refers to the grain, or smallest, unit that is distinguishable. Map data at different scales will allow for resolution of different objects. For example, a house site represented on a 1:24,000 scale map would not appear on a 1:100,000 scale map. In remotely sensed imagery, resolution is directly related to the pixel size, the area on the ground from which the radiances are integrated. AVHRR data at a 1 km x 1 km pixel resolution may be less useful for some environmental health studies than Landsat TM imagery at a spatial resolution of 30 meters by 30 meters.

Operational scale: Refers to the scale at which the process of interest occurs. For example, erosion may occur at a small or large scale. In addition, processes may be scale dependent, that is, they can be detected at one scale but not another.

Homogeneity and heterogeneity are affected by scale, and may affect the detectability of a process. An example is the infestation of trees by the pine bark beetle, which destroys individual trees and may affect a small area of trees in a stand. However, at the scale of the overall forest the loss is less detectable and the forest may appear homogenous.

Process scale and observational scale -- processes generate forms and patterns on the geographical landscape. When processes are involved the question of scale invariance should be raised, i.e. is the scale of observation relevant? The scale dependency of processes needs to be recognized so that both the appropriate observational scale can be chosen, or a multiscale approach integrated where appropriate.

When considering data for GIS, both the extent and the resolution will have a determination in terms of data volumes. Disk storage space continues to get cheaper, however the volume of data incorporated into a project is still a consideration in terms of both storage and analysis.

Generalization of information -- generalization may introduce uncertainty into a representation, however some phenomena may be more apparent when generalized, i.e. viewing the world from a distance may make the phenomena more apparent (Goodchild and Quattrochi 1997).

Scale/resolution/representation

Even at 1:1, the representation would still only approximate reality because of measurement errors and generalizations introduced by the representation of reality (Goodchild and Quattrochi 1997, p.4)

Error and Uncertainty

Issues having to do with uncertainty in geographic information include those having to do with both accuracy and error.

Accuracy — how well do the GIS data represent reality, in terms of positional accuracy, attribute accuracy, or temporal accuracy.

Positional accuracy may be related to the precision with which measurements were obtained, and the accuracy standards for a certain scale of data.

Attribute accuracy relates to the accuracy of the information linked to the spatial data units, whether they be points, lines, polygons or pixels. Are the polygons or pixels assigned to the proper class? Is the line segment tagged with the correct street information?

Temporal accuracy concerns the appropriateness of using a particular snapshot or snapshots of time for a particular modeling effort. Have data from different time frames been combined into one data set?

Errors in GIS may be categorized as source errors or processing errors. Source errors have to do with the accuracy of the data, i.e. the differences between the GIS data and the "ground truth" reality. Processing errors are those which are introduced into the database due to GIS processing, analysis and modeling. For example, when overlaying 2 different GIS coverages the errors in each will be compounded via error propagation to create additional uncertainty.

Some GIS softwares are including techniques for management of error and uncertainty in spatial data, for example IDRISI now includes some Fuzzy Set Analysis. Also being explored are the generation of fuzzy boundaries and spatially defined zones of uncertainty around point or line data.

Boundaries in GIS are one of the areas that can be used to illustrate issues of uncertainty and error. For some data types, a boundary is a well-defined line that can be delineated using precise surveying methods (e.g. a cadastral boundary between land holdings). Other types of boundaries are fuzzy in nature, such as the gradient between 2 soil types, and may be better visualized using a zone of transition rather than a sharp line. However this latter type of representation raises issues having to do with structuring of data models as well as scale -- a line may be an adequate representation of a boundary at one resolution, say a regional level, but not appropriate at a larger, site-level mapping.