| Home | Portfolio | Resume | Photos |
Jim Kompanek
The primary role of a cartographer is the portrayal of real-world data with map-based representations (Fairbairn et al. 2001). With modern Geographic Information Systems (GIS), this often takes the role of the multivariate representations of data, as well as the representation of data uncertainty.
Data uncertainty refers to the portrayal of the unknown, as well as the level of confidence a user may have in it. This unknown may often be the result of interpolated data, based on kriging or other methods, which determine the spatial distribution of a phenomenon based on observations made at control points (Wilson and Burrough 1999). According to Pang (2001), the portrayal of uncertainty in geographic data is abundant and can be found in numerous fields, including weather forecasting, Earth Observation Systems (EOS) data, and ocean modeling. In the case of ocean modeling, this uncertainty may take the form of “sparsity of data, noise in measurements, [or] uncertainty in the model” (Pang 2001).
Uncertainty characterizes the unknown and is based on the measurements and observations of phenomenon and predictions made from them. Among many factors, this includes error, [in]accuracy, validity, quality, and noise (Pang 2001). As it may encompass all or some of the previous concepts, data uncertainty is a general term, with no universally agreed upon definition. As previously mentioned, some of the factors of uncertainty may include:
1. Error: The difference between an assigned value versus its given value;
2. [In]accuracy: Accuracy or inaccuracy of data is defined by the consistency (or inconsistency) between the actual value versus its interpolated or predicted value;
3. Validity: A combination of both accuracy of the data and the subsequent procedures applied to the data and is measured by “deductive estimates, inferential evidences, data consistency, and comparison with independent sources and is ratified by testing” (Pang 2001);
4. Quality: Data quality refers to a combination of data validity and lineage, with lineage defined as the characteristics “of data that are monitored and tracked in database operations” (Pang 2001);
5. Noise: Noise is the random interference which is both uncorrelated and independent of the observed phenomenon.
As uncertainty can come in many forms (as well as those in addition to the ones listed above), its representation can pose a great challenge to cartographers.
There are numerous ways data uncertainty can be portrayed but they generally fall into two categories (Pang 2001). This includes “how uncertainty itself is represented, another is by how uncertainty is encoded into visualization.” Some examples of different methods of displaying this uncertainty may include the use of color or transparency, line width, or sharpness or focus. For example, the uncertainty of data may be portrayed with darker hues, higher levels of transparency, thinner contour lines, or fuzzy or out of focus data. By far, the most common portrayal of data uncertainty encountered during this literature review was verbal statements in the map legend and text. For the purpose of my Capstone project, the uncertainty was depicted by the use of color and transparency, as well as verbal statements in the map legend.
More recent advances in GIS systems also allow for use of real time animation (such as moving dots) to portray uncertainty. According to Pang (2001), holistic approaches of displaying uncertainty are becoming more common with improved technology. Instead of portraying uncertainty as another layer of data, this approach attempts to integrate the uncertainty into the data itself.
For the purpose of this paper, the use of multivariate representations were examined within the context of environmental modeling as well as historic changes in boundaries, as well as land use.
Kickapoo Valley, Wisconsin
Heasley (2003) explored the environmental history of the Kickapoo Valley through the context of changing landscape. The author examined many changes in the valley since the 1930’s and used the historical context of the rural setting for explanation. The primary trend of the article was the changing ownership patterns in various rural communities over the twentieth century. Specifically, landownership changes and rates of absentee ownership were the focus of the paper.
These changes were portrayed using nominal visual variables including arrangement and orientation to display the various changes in ownership information, as well “small multiples” (Gruver 2007) to portray changes over time. All of the data used in the paper was constructed from archival and historical research, therefore the level of uncertainty was generally low but dependent on the error and validity of the historical research. This uncertainty was briefly discussed in the text but was not distinctly portrayed in the maps.
Northern England (10,000 – 5,000 BP)
As opposed to the previous article, where historical trends are reflected by the changes in boundaries, Spikens (1999) discussed how ecological changes are reflected in observed cultural constructs. Primarily, Spikens investigated the changes in the distribution of woodland types in England in the Early Holocene period.
Spikens created a series of maps based data derived from soil information, pollen cores, and a mix of environmental and geological datasets. With this data, a series of “small multiples” were created which used hue and arrangement to display “Probable Dominate Woodland Types” over the span of 5,000 years. All of the maps provided were based on interpolated data in the form of a predictive model; as such, all of the potential types of data uncertainty (see Data Uncertainty section of this report) may be encountered. Some of the uncertainty was dealt with by using dashed lines (in the case of probable shore line extents), though much of the uncertainty was simply implied and not portrayed on the maps. The role of uncertainty is discussed thoroughly in the text (primarily the number of assumptions that were made for the purpose of the vegetation model), especially in regards to concern over the use of abrupt lines between environmental zones.
The ability to map the unknown with the help of interpolation and multivariate representation is powerful tool in GIS. After this literary review, it is clear that the ability (or at least the desire) to map the uncertainty is severely lagging. At this juncture, it is almost easier to map the unknown versus mapping the uncertainty regarding these predictions of the unknown. For its simplicity, the use of text based “disclaimers,” perhaps the least holistic approach to modeling uncertainty, appear to be the most commonly utilized type.
Fairbairn, David; Gennady Andrienko; Natalia Andrienko; Gerd Buzik; and Jason Dykes
2001 Representation and its Relationship with Cartographic Visualization: A Research Agenda. Cartography and Geographic Information Science 28(1):13-28.
Gruver, Adrienne
2007 Lesson 07 - Multiple Classifications and Multiple Representations. The Pennsylvania State University Certificate in GIS Program. Accessed 18 June 2007.
Heasley, Lynne
2003 Shifting Boundaries on a Wisconsin Landscape: Can GIS Help Historians Tell a Complicated Story. Human Ecology 31(2):183-213.
Pang, Alex
2001 Visualizing Uncertainty in Geo-spatial Data. The National Academices: Advisers to the Nation on Science, Engineering, and Medicine. Available online: http://www7.nationalacademies.org/CSTB/wp_geo_pang.pdf Accessed 16 June 2007.
Spikins, Penny
1999 GIS Models of Past Vegetation: An Example from Northern England, 10,000-5,000 BP. Journal of Archaeological Science 27:219-234.
Wilson, John P. and Peter A. Burrough
1999 Dynamic Modeling, Geostatistics, and Fuzzy Classification: New Sneakers for a New Geography? Annals of the Association of American Geographers 89(4):736-746.
This document is published in fulfillment of an assignment by a student enrolled in an educational offering of The Pennsylvania State University. The student, named above, retains all rights to the document and responsibility for its accuracy and originality.