1. Results and Discussion
[Xiao Liu:  Should there be a paragraph title here?]

Using correct data is the most important step in the implementation of any spatial decision support system.  It is impossible to provide good decision support using a spatial decision support system without decent data, even if the system is perfect.  Therefore, the data need to be as error free as possible to assure reasonable results.  The quality of data relies on two main factors: accuracy and currency.  Data accuracy indicates how well the data reflect the real world.  It involves data acquisition methods, data resolution, data format, and data units.  Data currency determines how timely the data are for a particular research project.  It is not always true that the more recent XXX data are XXX better.  For example, if the data are used in a research project that is focused on the past, they must reflect the XXX truth at that time.

Data preparation and description

For this research, data were acquired from various sources and were of different formats and coordinate systems. In order to analyze all the data in a single system, they had to be properly digitized into the system. XXX ArcView 3.1 was adopted for digitization and feature editing, and Arcinfo 7.2 was used for projection transformation. The raw data collected for this research included the followingX:

    1. General boundary
    2. To get the data of all the layers within the study area, a rectangular boundary polygon was created.  This rectangle served as a general outline for the study area and was used to clip all  layers that extended beyond its edges.  The specific area for analysis was determined by the distribution of the mine sites within the general geographic vicinity, which was of irregular shape.

    3. Soils
    4. The study area lies on the border of  XXX  Pike County and Warrick County.  The revision dates of soil survey maps of these two counties are 10 years apart.  The Pike County soil map survey was completed in 1989, and the Warrick County soil survey was completed in 1979.  There were some on going mining activities in the study area during  the time gap between 1979 and 1989.   Also,  although the geographic locations of the soil map boundaries of two counties  XXX match perfectly, the attributes of the soils were discontinuous along their borders .
      [Add a paragraph.]
      For those mining areas in Warrick County, which were explored between 1979 and 1989, the soil information recorded on the survey map was merely the soil series and properties that existed before mining activities began. To extract useful soil information, then, the data have to reflect either the current conditions or at least the conditions after mining activities began.  To compensate for this situation in the study area, and to get one continuous soil map, an assumption was made that the mining areas during that particular mining period XXX were covered with a unique soil series.

      First of all, soil map sheets from the two counties were placed side by side, and the study area boundary was drawn on them. With the map registered in a UTM projection, the soil polygons within the boundary were digitized and edited.  [Xiao Liu:  I talked to Steve Connelly a little about your thesis, and mentioned the prolific use of acronyms.  It was his view that they should be defined very clearly, and often, in order that even a lay person can understand what he's reading.  I suggest that you may wish to spell out UTM here.]  During the editing phase, the map of mine sites was overlaid on XXX the digital soil map.  The polygons in the areas that were affected after 1979 in Warrick County were then deleted from the map and the mine site boundary polygon was added to the soil map. [Xiao Liu: in the preceding sentence, does the word map and the words soil map refer to the same map?  I would edit it differently depending on the answer.] After all the polygons were edited, soil attributes were added to each of them.  The information in the general soil map includes the soil series name, the acidity (pH value), and the K value.  In the soil survey maps, the soil reaction (pH value) was recorded as a range of values.  However, to use the pH value in calculations, the values must be saved as a single value for each member of the soil series. The average value of the value range for a soil series was taken as the value for that soil series.  The attributes for those modified polygons in Warrick County were then imported into the data of similar areas during the same mining period in Pike County. [Xiao Liu: Does this last sentence still mean what you want it to mean?]

      The soil layer covers all the soil properties needed in this research. Acidity data were XXX primary criteria in the multiple criteria analysis, and the K values were the input for the RUSLE model. [RSULE ?] It would be quite time-consuming and most resource demanding if a particular property was extracted each time it was accessed in a program.  It is much better to have every particular data set ready as a layer in the database. For this purpose both an acidity map and a K value map were generated from the soil map by using region functions provided by ESRI.[ I know this one, but I didn't till you told me.  Should you spell it out?]

    5. Digital Elevation Model (DEM)
    6. The digital elevation model (DEM) records XXX information regarding the terrain. Originally the values in a DEM were elevations above XXX sea level.  The USGS 7.5 minute DEM, which has 30 meter by 30 meter ground resolution, was used for this research. [Xiao Liu: Do you in fact mean meters here, or do you mean minutes?] The file was stored in USGS DEM format, which could not be processed directly in ArcInfo or ArcView.  However, there was an import module in ArcView that allowed the USGS DEM file to be converted to a grid file. The DEM was mainly used to calculate an LS [LS ?] value for the calculation of erosion rate using the RUSLE model.

    7. Streams
    8. The streams and lakes layer was acquired from the Indiana Geological Survey (IGS).  Since XXX pollutants are carried by water and relocated along the waterways, the stream layer was extremely critical in evaluating the potential pollution caused by abandoned mine sites.  This layer, originally stored in shape file format, was then XXX clipped by the boundary shape file. Using the ArcInfo file format conversion function, the Shapefile was eventually converted to ArcInfo coverage allowing the program to access it directly.

    9. Mine sites
    10. XXX Mine site layer data, collected and prepared by the IGS,  recorded not only the mining areas but also XXX mining periods. This was the main source for delineating the analysis boundary. The intersection of this layer and the general boundary was outlined as the primary analysis area.

    11. Topographic maps
    12. Topographic maps were scanned from the U.S. 7.5 minute topographic maps as digital images. However, the geographic information on the map was not kept with the image file when the maps were scanned; XXX a geo-reference file had to be created for this purpose.  The world file is one that can record XXX geographic locations for any image files.  Primarily, the world file keeps the up-left corner coordinates, horizontal and vertical factor for converting screen units to map units, and rotation factor. [Xiao Liu:  I just don't know what you mean in this sentence, and am therefore reluctant to try to edit it without your input.]  Topographic maps provided general information about the area and served as back draft. [I don't know what back draft is.]  They also were a source of verification for land use and land cover classification from the satellite images.

    13. Aerial photograph
The aerial photograph, which was downloaded from the Microsoft TerraServer at its website http://microsoft.terraserver.com, XXX was in the JPEG image format. Since XXX projection and coordinate system information is lost when saving an image in JPEG format, a world file was created to recover the coordinate system information.  This was accomplished by using control points on the image and on the topographic map, which was already geo-referenced. The images were provided by the U.S. Geological Survey (USGS) in the format of Digital Orthophoto Quadrangles (DOQs), taken over the United States. The photos were originally obtained from the National Aerial Photography Program (NAPP). Each photo covers about one-fourth of a standard, 7.5-minute USGS topographic quadrangle map. The photos then were scanned into digital format and put through some enhancement processes, such as orthorectification, to remove distortions caused by camera angles. This layer could be used as a detailed XXX data field both for classification of the satellite imagery and for the comparison of other layers. The images for the study area in this research were collected on Mar 24, 1992. File format conversion

The Shapefile is a default file format for ArcView, and its coverage is a default format for ArcInfo. To reduce the amount of calculation in the program, all the layers in Shapefile format generated from ArcView were converted to ArcInfo coverage so that the ArcInfo functions could call up the data directly.  This conversion was conducted in ArcInfo using the ‘ShapeArc’ command.

Land use/ land cover classification

  1. Unsupervised classification
  2. Accuracy assessment
  3. Cover management factor generation

System description

System building is another important part of this study, in addition to data collection.  To build an appropriate system, the spatial decision support system must be user friendly and functionally rich. XXX

The SDSS application was built in a multiple document interface (MDI) environment (Figure 4.1). The MDI application comprises a main window and XXX many child, or subservient, windows. It allows users to open child windows within the main application window. More than one document or child window can be opened within a single parent window. Another advantage of using the MDI application is that the child windows can share the toolbars or menus of their parents.

The main menu XXX provides entrance to some typical Windows functions. The "File" menu item contains the functions to operate files, such as opening a child window and printing a map. The user can open a new child window and import any layer for display, such as the ShapefileXXX coverage and image file. XXX   From the "View" menu, the user can navigate the map and load the area of interest by zooming in, by zooming out, and by panning.  The "Modeling" menu item is the main function of this system.  It covers all the models used in the system, which include erosion rate information, stream buffer information, normalization information, and weight matrix information. Each model is represented by a tab form [icon ?] in the modeling window, which provides blanks for the user to operate. By selecting this item, a user is able to interact with the system by typing
XXX preferences in a user-friendly environment. The "Windows" menu item controls the layout of child windows within the frame of their parent window. The windows can be arranged as tile, cascade and auto arrange.

FIGURE 4.1 Main form of SDSS application

Buffering stream proximity

The tab window for the stream proximity calculation XXX comprises four parts (Figure 4.2): a scale bar for values, a scale bar for distance to the stream, a text box, and a panel for the display.  A user can assign values to the ranges of the distance to the streams simply by sliding the scale bar. With a  click of the "Set" button, the user’s selection can be shown in the small text box below.  The "Preview" button sends the user’s assignment to a temporary file and calls ArcInfo functions to calculate the buffering zones.  It eventually displays the result in the display panel.

The buffer function of ArcInfo was integrated into the program to calculate buffering zones. Figure 4.3 illustrates how the function works. The buffer function creates buffer polygons around specified input coverage features. The output is a polygon XXX with a new attribute called "inside" attached to it.  After every buffer function has been called, the output coverage is converted to a grid for further calculation. If only one buffer zone is set, it is quite straightforward to use the buffer function to generate the buffer zones. However, most of the time multiple buffering zones are requested, and the buffer function alone cannot handle it. In this case, other techniques need to be incorporated.
[Xiao Liu:  I just thought a paragraph break here would ease the reader's task.]

The operation of the buffer function, with XXX input of the buffer distance for a stream layer, generates a coverage with three values for the attribute "inside."  The values are: 100 for those inside the buffer zone, 0 for those outside  the buffer zone, and 1 for those in the "donuts" enclosed by pixels inside the buffer zone.  In the case of  more than one buffer distanceXX being defined, the buffer function can calculate only the last defined buffer zone XXX correctly. The result is more likely that a value of 100 will be assigned to  those pixels in the last defined buffer zone, a value of 0 to those pixels outside XXX the buffer zone, and a value of 1 to  those in the donuts (outside the polygons). To redeem the shortcoming of the buffer function, the following scheme was introduced.

For example, the user’s selection is set as:

Distance to streams
0 – D1
D1 –D2
First of all, D1 [Xiao Liu:  This pargraph is quite confusing to me.  What are the units of D1, V1, G1, etc.?]was used as the parameter for buffering. A grid file G1 was generated from the buffer function with values as 0, 1, and 100 according to the distance of the pixel from the streams. Then another buffer function was used on the original stream file using D2 as its input, and a grid file G2 was generated with values as 0, 1, and 100. The next step was to operate on G1 and G2 to produce the final grid file with the value V1, V2 and 0 depending on whether the pixel was within D1 to the streams, whether it lay between D1 and D2 to the streams, or if it lay beyond D2 to the streams. All the possible cases are listed in the table below.  The condition command was used to generate the output grid, as outlined in the table below
G1 pixel value
G2 pixel value
Output value
Any [All ?]
0 or 1
0 or 1
0 or 1
FIGURE 4.2 Tab form for stream proximity
FIGURE 4.3 Buffer function

Erosion rate calculation

The erosion rate window (Figure 4.4) consists of four buttons for conducting the functions. The "Calculate LS" button calls the function that runs the ArcInfo ArcInfo Macro Language (AML) (Appendix A).  It generates the grid file for LS values. The "Statistics" button shows the statistical values for the erosion rate, XXX  such as maximum, minimum and standard deviation. These values can be used when normalization is needed. The "Calculate A" button operates on all the input files for RUSLE [RUSLE ?]calculations.  It also exports of all the files XXX to an output file of erosion rate values. The "Preview" button displays the output erosion rate file.

The RUSLE model expects inputs of five factors, a rainfall factor (R), a soil erosivity factor (K), a slope factor (LS), a cover and management factor (C), and a conservation factor (P).

The R value was acquired from the CITY database which came along with the RUSLE program. This R value was regionally distributed, so it was a constant for the whole study area. The value for the study area is 200 hundred XXX   (ft-tonf-in)/(acre-hr-yr)-1. [Xiao Liu: Don't you think this might be clearer?]

The K value was extracted from the general soil maps that had been digitized from the county soil survey maps.  Using a regional function in ArcView, the soil map was categorized by this K value. The polygons with the same K value were merged into one region. This file was saved as an ArcInfo coverage and then converted to raster format for calculation.  The units for the K value are (ton-acre-hr)/(100-ft-tonf-in).

The P value was not taken into account in this study, since most XXX of the study area has not been restored back to crop fields and therefore no conservation methods were applied.

The C value primarily reflects the coverage conditions of the soil. This layer was generated from the land use/land cover classification extracted from remotely sensed data.

Figure 4.4 Tab form for erosion rate


Like the tab form for proximity to streams, the window for acidity (Figure 4.5) contains the same structure. The two scalebars allow the user to assign normalized values to a pH value and the "Set" button saves the assignment into a text file for final calculation. The "Preview" button reads from the text file that stores the settings, and then reclassifies the acidity file to generate the normalized pH value grid file.

Figure 4.5 Tab form for acidity

Weight matrix

On the weight matrix form (Figure 4.6), three scalebars were designed to allow users to enter their preferences, or weights, for each of XXX three variables: erosion rate, proximity to streams, and acidity. After the weights are set, the current weight matrix shows up in the text box. At that time, the " Preview" XXX  program calculates the final result according to the weight matrix and reclassifies the result so that the output values are normalized on a scale of 1 to 10 for reclamation priority.

Figure 4.6 Tab form for weight matrix
Interface (Delphi, ODE, MO) re-entry.

Normalization (Reclassification)

Normalization is a very important step in multiple criteria analysis. It reassigns new values to original values according to the requirements of the problem. Usually XXX normalization values fall within a range of numeric values. This step assures that every criterion is treated equally before any weight or preference is assigned to it.

The highest priority value for reclamation in this research was set to be 10. Therefore, all the criteria values have to be reclassified into XXX values between 1 and 10 before the final calculation can be made. The higher value reflects the higher priority for reclamation. For the instance of proximity to streams and acidity, as the system was designed, the user assigns a normalized value to each range of the original values at the time user interacts with the system. Higher values should be assigned to those areas that are close to streams and have low pH values. All the settings then are saved in a text file. The ArcInfo reclassify function can read from this  text file and operate on the grid file to generate a normalized output grid.

In the case of erosion rate, the value calculated from the RUSLE model is not easy to predict and not normalized.  Therefore, there needs to be a way to conduct the normalization. At the erosion rate tab form, the user has access to the statistics of the erosion rate result. Based on the statistics, the user is able to get a general idea on how the values are distributed. At the time of normalization, a linear algorithm is used. There are two ways to set the maximum and minimum values for this stretch of XXX data. The default maximum and minimum values from the statistics can be used, or the user can detemine them by analyzing the statistics.


All the operations are conducted behind the screen windows.  Therefore, the visualization of and the presentation of XXX results is another important component of the system.  Displaying the data allows the user to view the geographical distribution of the attributes of a particular layer. Overlaying XXX two or more layers in a single display panel or window provides more detailed information for assessing the results. Multiple display windows provide the capability for the user to make comparisons. If the user has two or more priority maps resulting from different settings of the weights for some of the criteria, he can compare the results on the screen.  A great advantage of multiple display windows XXX is that of allowing the user the option of calling up other data to assist in the comparison.

System requirements

To run the spatial decision support system properly, the following requirements have to be met.

    Hardware  : PC compatible, 32MB RAM
    Software : Windows 95 or NT platform, 32 bit operating system , and ArcInfo library and license

MapObjects DLLs [DLLs ?]

Regarding the physical storage of data and applications, two considerations need to be taken into account : 1.the directory structure, and 2. the naming convention. To simplify the system, a directory structure was created to avoid forcing the user to find the data path XXX when running the system. It is only necessary that the users XXX copy the whole structure. Since the system accesses a fixed directory structure for all the data files and program files, XXX the structure must be intact to ensure that the system gets correct data and functions (AMLs).  Naming convention for this system is critical for XXX users who want to replace the existing data files with their own custom data files.  If any of the data file names are changed, the system will not function normally.

All the data, AMLs, and application files must be saved in the same directory as "working directory."  In the system, the users have the option of changing the system path to the working directory. The files and directories under the working directory should include:

1. DEM directory for digital elevation model data used to generate LS values for erosion rate model,

2. K_Value directory for K value used in erosion rate model,

3. C_Value directory for C value used in erosion rate model,

4. Stream directory for streams and lakes layer,

5. PhGrid directory for pH values,

6. sl.aml, fil.aml, dn_slope.aml, high_pts.aml, s_length.aml, and ls.aml for LS value calculation algorithm ,and [Whew!  What's all this?]

7. SDSS.exe for the executable file of the application.

More coming