Lab 5: Visualizing the Terrain Survey

Introduction

In the previous lab (#4), a four square foot sandbox was used to create a strategic terrain from which elevation data could be collected to create digital models.

Figure 1. Normalized data table.

The purpose of this lab was to create the digital models using elevation data gathered from the previous lab. It is important to consider data normalization when collecting samples in geospatial surveys. Data normalization is the "cleaning up" and organizing of data into columns or tables to improve data integrity. In this case, the data collected in Lab 4 was put into an Excel spreadsheet (Figure 1), because it was compatible with the GIS programs that were used later, ArcMap and ArcScene.

The samples in lab 4 were collected using a grid with 6 cm intervals and a sample was collected at every grid intersect (Figure 2). These samples points were plotted in ArcMap and interpolated to create a 3D model in ArcScene. This lab will examine the effectiveness of the data collection method in its ability to create an accurate digital terrain model using 5 interpolation methods.

Figure 2. Sandbox with terrain and grid.

Methods

Once the data points were imported into ArcMap, interpolation was used to visualize the terrain model. According to ArcHelp, interpolation is used to create a continuous surface from the provided sample point values. Here are the 5 interpolation techniques used in this lab as described by ArcHelp:

1. Inverse Distance Weighted (IDW)-Determines cell values using a linearly weighted combination of a set of sample points.

Advantages: Can set parameters to how heavily the control point is weighted.

Disadvantages: Unevenly spaced data points will result in an unrealistic model.

2. Natural Neighbor-Uses an algorithm that finds the closest subset of input samples to a query point and applies weights to them based on proportionate areas to interpolate a value.

Advantages: It creates a smooth surface, even if data is missing.

Disadvantages: Has a limit of about 15 million input points.

3. Kriging-An advanced geostatistical procedure hat generates an estimated surface from a scattered set of points with z-values.

Advantages: Predicts empty spaces by analyzing spatial behavior derived from z-points, which can create extremely accurate 3D models.

Disadvantages: Can create a highly inaccurate 3D model when not enough data points are present.

4. Spline-Uses a mathematical function to estimate values and minimize overall surface curvature.

Advantages: Results in the smoothest surface of all the other interpolation techniques.

Disadvantages: Can be produce an inaccurate 3D model if the original terrain is not smooth.

5. Triangular Irregular Network (TIN)-Constructs a vector surface made up of triangles by triangulating a set of vertices.

Advantages: Can have a higher resolution in areas where a surface is highly variable.

Disadvantages: The hard-line edges of each triangle are highly unrealistic to any sort of terrain in nature.

In ArcScene the 3D models were exported as JPEGs. 

Results/Discussion

Below are the images of all the 3D models created using the 5 interpolation methods. Since letters were incorporated in the terrain a head-on orientation was used to read the "hidden message". The models were also tilted slightly to show how dynamic the terrain features are.

IDW:

This interpolation method did well in pronouncing the sharp edges of our desired letters. However, the many peaks that result from this method are unrealistic and unattractive.

Figure 3. IDW model

Natural Neighbor:

This method proved to be very accurate, however some random points in the plains areas are too pronounced. This model looks more realistic than the IDW method.

Figure 4. Natural Neighbor model

Kriging:

This method represented the sandbox terrain the worst. It showed hardly any change in elevation and lacked any sharp edges that were present in the sandbox. This may have occurred due to the sharp edged of the terrain and inefficient amount of data points.

Figure 5. Kriging model

Spline:

This method captured the terrain of the sandbox the best. It captures the dynamic change in elevations of the letters while still keep a smooth, realistic surface.

Figure 6. Spline model

TIN:

This model does a great job of capturing elevation. However, the smooth plain looks rougher than it should and the blocky nature of the TIN model creates an unrealistic texture.

Figure 7. TIN model

The 6 cm interval grid proved effective in the number of data points collected in most interpolation models. The only model that had absolutely no resemblance to the sandbox terrain was the kriging model. However, the terrain that was meant to represent a "heart" did not show up in any model as effectively as the other letters. This may be due to the combination of curve and change in elevation occurring at the two upper "humps" of the heart. Such a drastic curve is tough to capture through a grid set-up.

Conclusion

Using a 4 by 4 foot sandbox, a dynamic terrain was created and a systematic grid survey technique was used to capture elevation. Once the elevation data was normalized, it was imported into ArcMap and ArcScene to use various interpolation methods to observe which interpolation method created the most realistic 3D model. This survey technique is comparable to any other survey technique that collects elevation data, either large-scale or small-scale. One of the most important of those survey techniques that is similar to this lab is LiDAR, which uses millions of data points to create terrain models. This difference with this lab is that very basic equipment and techniques were used to collect the data and this survey was at an extremely small scale. The grid based survey is not always necessary, especially when the study area is large. It would be quite difficult to set up a grid when there are objects in the way of intersections and grid lines.