Introduction
In the last field activity, group 1 set out to create a digital elevation model (DEM) by collecting data points within a 1 m x 1 m sandbox. The group collected elevation points using a stratified sampling system that also had systematic sampling techniques applied as well. Once the survey was completed, the data was entered into a spreadsheet to be imported into ArcGIS to be mainpulated into the DEM. An important aspect to complete field activity 5 was data normalization. Data normalization was important for group 1 because in order to create a DEM, the data has to be organized properly. This organization allows groups to reduce data redundancy and helps to improve the integrity of the data. For group 1, data normalization was accomplished by creating three columns in a spreadsheet with the field names: "X_Cell", "Y_Cell", and "Z_Value". The X cell corresponds to the x value of the sandbox grid. The Y Cell corresponds to the y value of the sandbox, and the Z Value corresponds to the elevation points collected at each point. This organization allows the data to be imported fairly easy, with the names already correlating to geoprocessing techniques within ArcGIS. The data points group 1 collects are used to facilitate interpolated processes that allows ArcGIS to create a DEM very efficiently. The interpolation procedure takes in the collected data points and creates new points around the known points through a variety of different methods.
Methods
The first step group 1 took to complete the activity was tom import the Excel file with all of the data points into a Geodatabase for the sandbox. The fields were double checked to be numeric, as this allows the points to be converted into data usable for map creation. The data is then added to a blank map in ArcMap as "XY data'. This allows one to convert the points brought in to a feature class that is usable for interpolation later. The next step the group took was to utilize the 'Raster Interpolation' within the 3D Analyst Geoprocessing tools to create maps that replicate the intended terrain.
|
Figure 1: Actual Sandbox Terrain with Amanda adding detail to sand. |
The first interpolation utilized by group 1 is Inverse Distance Weighted (IDW). This method averages points in each neighborhood of a cell, and then weights out the priority of veraging to determine the value each point is given (Figure 2). This method was very successful for the group, as many of the data points were collected in the center of cells, so the output image brought a very realistic image of the terrain to a DEM. The
|
Figure 2: IDW Interpolation Result |
The next method employed by the group was the Natural Neighbor interpolation procedure. It is also known as the "area-stealing" interpolation, as it applies only local values, and establishes a weight that creates a value for each point around the surface. The natural neighbor interpolation provided perhaps the most realistic model of the terrain in the sandbox (Figure 3).
|
Figure 3: Natural Neighbor Interpolation Result |
The third method employed by group 1 in interpolation was the Kriging interpolation process. It uses regionality to establish Z-value patterns throughout the surface, without employing any other method. It assumes the surface is homogeneous, and the same pattern of values are observable at all locations on the terrain. It did not due to well to describe the data collected by group 1 (Figure 4). It centered too much weight around the plateau in the lower right corner of the terrain, and did not do a good job at describing the mountain ridges.
|
Figure 4: Kriging Interpolation Result |
The fourth method employed by group 1 was the Spline interpolation (Figure 5). It utilizes a mathematical function to minimize surface curvature, which in turn maximizes the smooth surface of a surface. It assesses all sample points, and fins the curvature between them. The spline interpolation method did not work well for the data, as it focused very heavily on the data points that were clustered, and formed terrains that were over-exaggerated.
|
Figure 5: Spline Interpolation Result |
The final interpolation method that group 1 employed was the TIN surface. This method employs triangular networks set up between the data points collected. It connects all of the lines to create triangles, which create a network of shapes that resemble the surface. The TIN method worked very well for the data group 1 collected (Figure 6). It portrays all of the various terrains very effectively.
|
Figure 6: TIN Interpolation Result |
The data was manipulated to bring into ArcMap, as the 3D visuals are not apparent in anything but ArcScene and ArcGlobe. To create the scene in ArcScene, the data was interpolated using the various methods. The properties of each individual file is then changed to set the elevation from surfaces to "Floating on a custom surface" to create the 3D effect. Once this is done, one can change the data frame properties of each map to "Calculate From Extent" selection of the vertical exaggeration. This creates the proper scene for analyzing the data. Once the scene is set, the shapefile can be saved as a layer file, and captured as an image. This allows one to open up the two in conjunction with each other in ArcMap to overlay a legend and map properties upon the image. The image was chosen to be landscape orientation, as it displays images such as these the best way.
Results/ Discussion
The best interpolation method for group 1 was the natural neighbor system (Figure 3). It created very accurate local points that portrayed an image that resembled the original sandbox terrain quite accurately. There were very few unaccurate points of the survey. They will be discussed in this section.
Overall, the IDW interpolation provided a very good image of what the sandbox studied looked like. I captured every feature of the box quite well, however the spacing between points portrayed a skewed data set. It did not display the sandbox as a continuous terrain. This could probably be solved by collecting more data points to create a more fluid picture of the terrain. The natural neighbor provided an excellent DEM. It was able to portray the terrains of the different regions with clarity and quality. The only way to make the interpolation more effective would have been to collect more points in different areas. The Kriging interpolation was very poor to represent the groups data. It did not account for the mountain ridge in the sandbox, and created the plateau as a very high hill. This is most likely because the plateau had a key chain size llama on it that the group also collected (which appeared in different interpolations). The spline interpolation was also poor, most likely due to the high concentration of points near the llama shrine again. It provided a good image of the regional if one were to "read between the lines" and see the spots that the reduced curvature removed. The TIN provided an excellent interpolation for the data. The method allowed the data to create well-done triangles that portrayed the terrain as it is. Like the Natural Neighbor interpolation, the TIN scene could have been much better if many more data points were collected to create a continuous data set. Overall, each of the interpolations could have been successful to view the data as a DEM if necessary. The features alone create a very interesting method to collect data. This field activity allowed group 1 to create an effective DEM off of the last field activity in which they collected the data.
Conclusion
This survey was a very effective stratified sampling system. This survey is like other stratified sampling systems in that it collected the elevations of locations, be it for engineering or otherwise. It is different because it was collected using a meter stick rather than a survey GPS. It is not always realistic to perform a grid based system to collect data, as it could intrude upon private land, or buildings one can not enter. A stratified system with a grid influence is probably the best bet for most collection, as it can account for other influences that don't always come up in planning.