This page reports on the creation of an interface to explore spatial data uncertainty. This interface originally ran on Arc/Info 7.1x and was comprised of a modular set of Arc Macro Language (AML) scripts and menus. It has been modified to be displayed through a Web browser. The first section includes the both the HTML version of the interface as well as a link to download the AML and data. The second section describes the development of the interface and assesses the capability of AML for GIScientific research.

Image-Mapped Demo
The following image is the main menu for the aml-based interface developed and described here. It has been image-mapped so that clicking on most buttons will bring up maps and text in separate windows. You may also download the aml scripts and menu files and the grids and coverages used in it.

Questions for those examining the demo:
Does this environment describe the simulation/propagation paradigm adequately?
How do the animations compare with the probability maps for the communication of information about uncertainty?
Does the added information about application uncertainty justify the "overhead" cost?
What conclusions do you draw about the inclusion of uncertainty for the specific GIS applications shown in the demo?

I'd appreciate your answers; please write me at (ashton@geog.ucsb.edu).

Arc/Info is a powerful multi-function geographic information system. It has been used for every GIS application imagineable. Arc Macro Language extends the basic functionality of the GIS by enabling users to write scripts to automate lengthy processing, develop new commands, and construct user interfaces. How useful is AML as a tool for the GIScience research community? To help answer this question, this project's elements are broken down below with the proportion of each stage.

Data Processing (No AML)
Data processing involves obtaining the appropriate data files, clipping them to the area of interest, projecting them to a common format, resampling them to a uniform cell size, and performing reclassification on the land cover set. Some of this was done outside of Arc/Info, and the rest was done at the command line. For this work, aml was not inportant, though Arc/Info certainly was.

Uncertainty Model Development and Simulation (25% AML)
A geostatistical model was used to characterize and simulate uncertainty for the DEM. Such functionality is beyond the current scope of Arc/Info. The GSTAT package was used for this purpose. GSTAT accepts and writes the grid-ascii format for input and output, and AML was used to export and import the DEM realizations. A loosely coupled model could be developed using AML's interface to external processes, though this was not attempted.

The simpler confusion matrix-based landcover model was done in C and AML. A C program took a grid-ascii format landcover file and a text confusion matrix performed the simulation, and wrote the results as grid-ascii files. The resulting "noisy" realizations were imported to Arc using AML and filtered using Grid commands through an AML.

GIS Spatial Model Implementation (100% AML)
Because the models had to be run 100 times each, automation was critical. As written the AML is capable of handling . It can be extended to handle new spatial models easily. The construction of the probability maps would not have been feasible without AML.

Data Visualization (100% AML)
Working with many realizations meant that new display techniques like the "cinema verite" animation were necessary. These techniques were not supported directly by Arc but could be -- and were -- developed using AML. Arc's display environment is not particularly sophisticated by modern graphics standards (multiple, independent plotting windows are apparently not possible in one arc session; display redraw is done onscreen rather than in the buffer, resulting in poor-appearing redraws; legend and title placement and appearance is outdated with limited editing options). This may reflect an old, underlying commitment to the paper paradigm, one unfit for current research needs.

Interface Construction (70% AML)
While not a thing of beauty, the interface apparently does a serviceable job of presenting information to the user.
A more challenging task was reproducing this interface for the Web. The Web results, particularly the animation sequences, are particularly successful and are an improvement over Arc/Info's display. Some of the Web conversion was accomplished with the aid of AML.
 

From this list it is apparent that AML was useful for large portions of most stages of this work. The implementation of the various spatial analyses on a large set of realizations was successful. This indicates that GIS-based spatial analysis can be extended to encompass the propagation/simulation paradigm. Arc/Info's display capabilities are not satisfactory for the communication of spatial information through map graphics. This is a serious limitation, since so much current scientific communication is digital. Finally, AML's capability for developing graphic user interfaces is not flexible and the resulting products (like mine) can be "kludgey". Since interfaces may be very useful in.the communication of ideas, it would be desirable to see greater functionality with this aspect of any future AML version.