Visualizing and Refining Terrain Surface

 

 

Introduction

      The goal of this exercise is to create 3D terrain models, of the landscape created in exercise 1, using various interpolation tools found in ArcMap. The interpolation methods used in this exercise include IDW, Kriging, Natural Neighbor, and Spline. We also use the elevation values of the sampled points to create a Triangulated Irregular Network (TIN). Based on the 3D terrain models we created we are able to determine whether or not our sampling technique was sufficient to produce accurate terrains.

 

     After our original 3D terrains are created we are tasked with determining which areas, if any, need to be resampled to increase the accuracy of the terrain model. In many cases, areas with great variability in elevation will need more sampling points than areas that are relatively flat.

Methods

      We being the process of creating 3D terrains of our landscape by adding the tabular data, from exercise 1, into ArcMap using the "Creat Feature Class from XY Table" feature. This tool creates a point feature class where each point is given its elevation value based on its xy locatioin. Figure 1 below shows the xy coordinate grid that will be used in the interpolation process to great the 3D terrains.

 

Figure 1.  XY grid based on the 10 cm by 10 cm sampling technique used to aquire elevation values.

 

 



     After the point grid was created we used different tools in the Raster Conversion toolset to create 3D terrains of the surface. In this lab we used IDW, Kriging, Natural Neighbor, Spline, and TIN interpolation methods to create 3D terrain surfaces.   

 

 

     The first interpolation method used to create a 3D surface was the IDW, or inverse distance weighted method (Figure 2). IDW calculates elevation values of cells based on the elevation of surrounding elevation values. IDW uses the assumption that points closer will have more of an influence in determining interpolated elevation values than cells farther away. Overall, the IDW model created a terrain with a lot of dimples and does not look like our original landscape.

 

Figure 2.  3D terrain model created using the IDW method.

 

 

 

 

 

 

     The second interpolation method used was the Kriging Method (Figure 3). The Kriging method is a predictive model that uses distance or direction between sample points to create a spatial correlation, which will then be used to estimate elevation.

Figure 3.  3D terrain model created using the Kriging method of raster interpolation.

 

 

 

 

     Natural Neighbor (Figure 4) uses a weighted average of proportionate elevation values of the nearest known points to create a 3D terrain. The interpolated elevation values have a range found between the sampled points, and therefore cannot replicate peaks or valleys that are not already represented by sample inputs.

Figure 4.  3D terrain model created using the Natural Neighbor method of raster interpolation.

 

     Spline uses a mathematical equation to create a surface that passes directly through the exact value of each known point. This equation is aimed at minimizing the curvature of the surface which leads to the smoothest surface of all the interpolation methods. Since our landscape has very smooth features the spline model seems to best replicate the real life landscape.

 

Figure 5.  3D model created using the Spline method of raster interpolation.

 

     Triangulated Irregular Networks are a vector based surface model. TINs take elevation values of known data points and create a series of triangles to create a 3D surface (Figure 6). 

 

 

 

Figure 6.  Triangulated Irregular Network created using the elevation of each point in the sampled grid.

 

     After each of the 3D terrains were created we noticed certain areas that were not modelled accurately. Therefore, we resampled the areas that were not modelled correctly (Figure 7). Instead of the original 10 cm by 10 cm grid we resampled the area with a 5 cm by 5 cm grid to add elevation points to create a more accurate terrain model (Figure 8).

Figure 7.  Resampling the mouth of the valley. This area was resampled using 5 cm by 5 cm grid.

 

  

Figure 8.  The feature class created showing the area that was resampled using the 5 cm by 5 cm grid.

 

 

 

     After the area was resampled we used the same technique described above to import the xy table into ArcMap. We then ran the Spline interpolation tool to create a new, more accurate 3D terrain model of the landscape (Figure 9). Although this model is not exactly identical to the original landscape it is the most accurate.

 

Figure 9.  3D terrain model created using the Spline method after the landscape was resampled.

 

Discussion

     After the original sampling technique was turned into 3D models it was clear that some features were not interpolated correctly due to a lack of data points. Therefore, it was necessary to add more sampling points to those specific areas to ensure a more accurate 3D model. The areas that needed the most sampling points were areas that had large variability in elevation while more gradually changing areas did not require additional sampling points to be modelled accurately.

     In addition to proper sampling of an area, several different user-defined options are available for each interpolation technique. For instance, in each of the interpolation methods we are able to change the output cell size. I believe changing the output cell size would remove the dimples found throughout the IDW model and in the valley of Spline model. For IDW and Kriging we are able to set how each elevation value will be weighted, which will also affect the output terrain.

     After the area was resampled we chose to remodel the landscape using the Spline interpolation. This method was chosen because it creates the smoothest surface, which represents the smoothness of the actual landscape the best. If our landscape had more angular featues it may have been necesary to choose a different interpolation method.

 

 

Conclusion



     In conclusion, the most important factor in determining the accuracy of the terrain model was choosing an appropriate grid to sample from. If our landscape had more variability in the features we would have had to use a smaller grid to capture all of the features. However, The 10 cm by 10 cm sampling grid did a nice job capturing all of the features in the landscape, except a small area at the mouth of the valley. Since this area had the largest variation in elevation we had to resample using a small 5 cm by 5cm grid. This change allowed the valley to be modelled accurately enough for this project.