Geo-spatial Field Methods
Interpolation
Introduction:
The previous lab involved a survey of various terrain features generated in a sandbox. In the sandbox we created multiple terrain features which included a ridge, hill, depression, valley, and plain. Survey points were sampled by the use of the systematic surveying technique. A total of 400 survey points were collected using this technique. Data was then recorded and normalized. The normalization of the data requires turning the data into a standard measure. This was done by transforming the surveyed data points into x,y, and z coordinates in excel. These coordinates were then used to map out the surface of our surveyed area (the sandbox 114 by 114 cm). We used multiple interpolation methods to generate estimated values at points not sampled during the survey based off of the known values of the surrounding points. The use of various interpolation techniques allowed us to generate accurate visualizations of our terrain features. However, certain interpolation techniques were more accurate than others.
Methods:
First, survey data had to be normalized into a format that could be used in ArcMap. This involved transforming the data into x, y, and z coordinates in excel.
Figure 1. Above is the format of our normalized survey data points.
Data was imported into ArcMap as points feature class. Ultimately, a layer file was generated which was to be put into our geodatabase. ArcScene could now be used to view the data in a 3-dimensional format.
Figure 2. Above is our data points shown in ArcMap.
After the data was imported into ArcMap, interpolation techniques were used to visually display our data as an accurate representation of the terrain features that we created. The two interpolation methods that most accurately displayed our terrain features were selected. Data interpolation is designed to estimate the values of cells in a raster that could potentially have missing data points/values. Below is our data points in ArcScene before any interpolation techniques were applied.
Figure 3. Above is our data points in ArcScene before any interpolation methods were used.
We applied a total of five different interpolation techniques to our data to determine which technique most accurately portrayed our terrain features. The descriptions of the five interpolation techniques that we used are listed below.
IDW Interpolation: IDW (inverse distance weighted) interpolation presumes that points/cells close together are more similar than points that are far apart from one another. Therefore, IDW predicts cell values based on the averaging of data values in the neighboring cells. Points nearest to the center of the cells being estimated have the most leverage. IDW does have its draw backs. Surfaces cannot be accurately generated if there are points that possess smaller values than the lowest value in the data set. The same holds true with data points that possess larger values than the highest value in the data set. This means that if a very high or low point is not surveyed then IDW will inaccurately display the data. Dense sampling techniques must be used in order for IDW to accurately display what you are surveying. We used a systematic sampling technique that resulted in the collection of over 400 data points in a sandbox. IDW is a good interpolation technique to use in our situation. (ArcGIS Help)
Natural Neighbors Interpolation: The natural neighbors technique of interpolation simply takes the closest subset of samples for a specific point, and then assigns weights to them. The weights are assigned off of proportionate areas so that unrepresented points can be given values. This method has the same draw back of IDW. If there are peaks or extremely low points that are not survey the information will not be accurately represented. Natural neighbors is a solid interpolation technique for our situation because of the large volume of sample points that we took. (ArcGIS Help)
Kriging Interpolation: Kriging interpolation uses statistical models that possess auto-correct functions. This means that kriging interpolation is able to predict a surface, and inform the operator about the accuracy of the predicted surface. (ArcGIS Help)
Spline Interpolation: Spline incorporates a function that reduces surface curvature, which results in the production of a smooth surface. However, the surface must pass through the data points. The greater the number of points the more smooth the surface will be. Since we collected so many data points the spline method proved to be a great method to display our terrain features. (ArcGIS Help)
TIN Interpolation: The raster to TIN tool generates a triangulated irregular network where the surface is not allowed to deviate from the input raster by more than a specified Z tolerance. One downfall of TIN is that the generated image will not have any smooth features. TIN only shows the surveyed points which means there will be a lot of peaks and valleys displayed in the image. This will not accurately show the landscape. (ArcGIS)
ArcScene was then used to generate a 3-D image of our terrain features. Corrections for drastic peaks and valleys had to be used on all the interpolation methods.
Results/Discussion:
Shown below are all of the different interpolation methods results. The first image shown below are the two best interpolation methods that most accurately display our terrain features.
Figure 4. Above is the spline and natural neighbors interpolation methods. These methods produced smooth terrain features that do a great job in displaying our terrain features.
Spline and natural neighbors interpolation best displayed our data (Figure 4). The reason spline and natural neighbors were the best methods was because we sampled a large amount of data points. Since we used a dense sampling technique these methods were able to accurately display our data.
IDW Interpolation
Figure 5. Above is the IDW method of interpolation.
IDW does a great job in showing our hill and ridge feature (Figure 5). However, IDW produced an image that is very bumpy which isn't the most pleasing. IDW does not show the smooth changes in elevation like the natural neighbors method does.
Natural Neighbors Interpolation
Figure 6. Above is the natural neighbors method.
Spline Interpolation
Figure 7. The Spline Interpolation method.
The spline along with the natural neighbors method did the best job in displaying our data (Figure 6) (Figure 7). However, the hill has rounded off slightly. This is one of the down sides to these methods is that when a smooth image is produced drastic changes are not always shown. In our case the hill is less defined than it otherwise would be in a TIN (Figure 8).
TIN Interpolation
Figure 8. The TIN interpolation method.
The TIN image shows a very jagged landscape (Figure 8). This is because the TIN interpolation method only takes into account points that were surveyed which leads to a jagged landscape. However, the TIN most accurately shows our hill while the natural neighbors method shows a more rounded hill (Figure 6).
Kriging Interpolation
Figure 9. Kriging Interpolation.
Kriging interpolation does not to a horrible job of displaying our terrain features (Figure 9). It is certainly a better method than TIN (Figure 8). Kriging does a good job showing features in the 2-D model, but the 3-D model just isn't as nice as the spline and natural neighbors models (Figure 4).
Conclusion:
Overall, it is not always practical to perform an extremely detailed grid survey. However, it is important to concentrate survey points in locations that have drastic elevation changes. This activity greatly improved my knowledge of various interpolation techniques as well as surveying techniques. Understanding the advantages and disadvantages of the different interpolation techniques will help in future geospatial projects. It should also be noted that interpolation can be used many different types of data such as temperature and rainfall data. Although we simply surveyed a sandbox the concepts we learned can be applied to real wold situations. The only difference is the sandbox is much smaller.
No comments:
Post a Comment