Alan MacEachren, David Howard, David Askov, Tim Taormino & Martin von Wyss
RVIS started with a contest sponsored by the NCGIA in the Spring of 1993. The challenge was to find a way or ways to depict the uncertainty in a data set. The data set that was used concerned dissolved inorganic nitrogen in the Chesapeake Bay. Following this introduction are several pieces of the poster that was presented at a special poster session of the 1993 GIS/LIS Conference. RVIS was judged the winner of the NCGIA challenge.
Pressing any of the buttons below will take you to the short segment of this page that is about that function of RVIS. You can also choose to scroll straight through this page in which case you will see the abstract and then a 'sample session' with RVIS.
Following is the abstract that accompanied the poster and then about 25 annotated pictures from a typical RVIS session:
Two important questions related to uncertainty representation have been identified in previous research:
(a) which graphic variables are appropriate for showing different kinds of uncertainty
(b) what kind of user interface is most effective.
The development platform for the ESDA project is IMSL/IDL, running under UNIX on Sun Sparcstations. IDL provides a dynamic environment for addressing the above questions and for extending the first to dynamic as well as static variables.
This is what a user would initially see in RVIS. The red map shows amount of dissolved nitrogen and the blue map shows the size of the 95% confidence interval about the values on the red map. These confidence estimates were obtained through an interpolation process called 'kriging' which returns the 95% confidence interval along with an interpolated grid.
This succession of images involves the exploratory data analysis concept known as focus. The idea is to focus in a certain set of values. This particular implementation greys out successively more of the data values as the slider is moved to higher and higher values.
These four images portray the point icons of uncertainty. The triangles grow larger both in area and in height with a constant base width. The error estimates for these triangles were derived by the jackknifing method. This allows a researcher to check the values at the sampling points.
The maps above are bivariate maps. They show both the value of the data and the value of the 95% confidence interval. The first three maps are 'conventional' color choropleth maps with three different coloring schemes. The bottom full-panel maps show both the original data and another sort of bivariate map attempting to use the graphic variable focus to portray uncertainty. As the data grows more uncertain, the dots get smaller and harder to see.
This is a contour map on top of the data map. The contours show uncertainty and grow thicker as the uncertainty rises.
For more information about RVIS, contact me by e-mail
