An Essay on Scale

Scale can be described as the level or extent at which observations or phenomena are represented geographically or temporally.  Scale can be described in aggregate or specific terms.  Generally, one can describe varying levels using such terms as “individual, household, neighborhood, city-wide, nation-wide, daily, annually” etc., or conversely by using more discrete definitions to quantify the physical dimensions or duration of represented events.  In regards to cartography, scale refers more specifically to the former; or the ratio and relationship between corresponding elements in reality to those represented on a map.  This ratio can often be expressed as a fraction – for instance 1/100,000 or 1:100,000, where one unit of measure on a map is equal to 100,000 of the same units in reality.

When obtaining data, it is important to consider scale to determine the appropriateness of a GIS coverage for a given project.  For example, analysis conducted on a large scale – such as a parcel of forest land – requires data obtained at a high resolution.  Whether these data are vector rivers or roadways, digitized at a high resolution or captured using many GPS observations, or the data have been obtained by using high resolution raster imagery or interpolated surfaces, it is important for the data to provide enough information about a geographic extent to give a suitable representation of reality for a desired level of study.  Conversely, the same high resolution data may not be necessary for analysis of an area at a smaller scale – such as a state-wide or regional analysis of forested lands.  In this smaller-scale example, data obtained at a scale of 1:100,000 would be more practical to obtain and utilize for analysis, whereas data for parcel-level analysis may need to be obtained at a scale greater than 1:10,000.

These examples briefly illustrate how the accuracy of data is defined by scale.  If features are digitized at a scale greater than 1:10,000, the layer becomes a more accurate depiction of ground conditions as the project’s scale decreases – the 1:10,000-scale stream features are much more accurate at 1:100,000-scale.  In contrast, features digitized at 1:100,000 quickly loose accuracy as the project’s scale increases to parcel-level.  Further, a high resolution raster dataset may affect the efficiency of running analysis over a wide extent, while a low resolution surface may not appropriately represent the events occurring across a given area.  For these reasons, it is apparent that scale characterizes the degrees of suitability for which data is valid and useful.

Scale is an important issue to consider in the world of environmental and ecological studies.  One frequently occurring matter is a mismatch between the scale at which ecological processes occur and the scale at which decisions on them are made regarding these processes.  Many parameters influence the product of any project, and these parameters time and again lay at varying levels of scale.  Reading materials from class explain how focusing upon a single scale may neglect important interactions that are vital to the complex issue at hand.  Further, examination of an ecosystem may be performed at too large or too small of a scale to fully observe an occurring phenomenon.  Finally, the modifiable aerial unit problem (MAUP) is another example of scale-based error that emphasizes the previous challenge; however may be more of a matter of preference to meet a desired result.  The problem stems from the imposition of artificial aggregation of an unbounded area.  When a continuous area is artificially bounded, the results can be unavoidably skewed.  Altering scale, in this matter, allows control of the resulting spatial patterns.  Basing policy on assumptions containing overlooked conditions can easily impose unpredictable and degrading consequences on the environment and society; thus, scale is an important feature to observe and regard when examining spatial information.

A few years ago I contributed to a pilot project at the Florida Department of Environmental Protection to update Florida’s version of the 1:24,000-scale National Hydrography Dataset (24k NHD).  During this project I encountered a fundamental issue previously discussed.  At what scale should features be digitized to provide a valid 1:24,000-scale coverage?  It was determined that digitizing state-wide hydrologic features be captured most efficiently and validly by digitizing at a larger scale: between 1:15,000 to 1:17,000.  By increasing the scale, a slightly more accurate depiction of ground conditions are observed and obtained for the 24k dataset, while providing enough generalization to be able to progress across vast landscapes of the state with some amount of speed. Additionally, several issues of scale were discussed in the article, “Seagrass as pasture for seacows: landscape-level dugong habitat evaluation.” Raster imagery of seagrass meadows were obtained at 200 meter resolution.  At this resolution the continuous nature of seagrasses were suitably sampled, and analysis was performed appropriately.  It was not necessary to acquire imagery at a larger scale, as patches of seagrass species were easily captured at this resolution, while a larger resolution would begin to overlook patches of assorted species.