Creating a Digital Elevation Model from a Geostatistical Layer
In this step, you will choose one of the surface models with the best RMSE value. Then you will export the temporary geostatistical layer and create a permanent raster file. Before exporting, there are a few steps you will need to take to prepare. The first step is to determine which of the surface models to chose. Start by right clicking on one of your geostatistical layers. From the menu select, Compare.
This step opens the Cross Validation Comparison window. Here you can compare the results of any geostatistical layers currently loaded in the Table of Contents by selecting them from the drop-down menu on the upper right. Take a moment to compare the RMSE values. Find the surface model with the best RMSE. This surface model will be the one on which will you base your digital elevation model. The next step is to determine the optimal resolution for the digital elevation model using the average distance between the measured elevation values.
From the Analysing Patterns toolbox under Spatial Statistics tools, launch the Average Nearest Neighbor tool. The Average Nearest Neighbor tool determines the average distance from each feature to its nearest neighboring feature. In this instance, it will calculate the average distance between waypoints.
For the Input Feature Class, choose your elevation data. Leave all other settings as default and click OK.
If you remembered to disable background geoprocessing, the results appear in the geoprocessing window. If not, you may have to search for your geoprocessing results under the Geoprocessing menu from the main menu in ArcMap. Make a note of the observed mean distance. You may need to consider the average distance between waypoints to determine the cell size of your digital elevation model. In the next step, you will export the geostatistical layer with the best RMSE value.
Right click on your chosen geostatistical layer in the Table of Contents. Select Data, then Export to Raster. The GA Layer to Grid tool opens up.
Save your output surface raster to your working folder. The tool will automatically try to determine the cell size based on the dimensions. The tool will automatically try to determine the cell size based on the dimensions, height, and width, of the dataset. This cell size may not always be appropriate. Determining the proper cell size is up to you and mainly a judgment call. The closer the cell size is to the average distance between measured values, the more accurate it might be. However, the purpose of interpolation is to estimate values in between measured points. In most cases, it is ok to have a cell size smaller than the average distance between points. The size of the map and the desired smoothness of the surface is another factor to consider. Larger cell sizes will result in a coarse or pixelated hillshade. I recommend starting with a small cell size. In this example, I chose a cell size of 1, which is about one-quarter the average distance between measured values for my dataset. Leave all other settings as default and click OK.
Feel free to export to raster multiple times, experimenting with different cell sizes.
A digital elevation model will now appear in your Table of Contents. The result may appear very similar to your original geostatistical layer. This visualization is misleading since the values across the surface have a continuous range rather than discrete categories.
To make the digital elevation model appear similar to what you are used to, you may have to change the symbology. Open the properties for the DEM and select the Symbology Tab. Choose Stretched from the options on the left. Choose the black and white color ramp or another color ramp of your choosing.
It should now look more like what you are used to when compared to digital elevation models you have seen in the past.
To test your digital elevation model, create a hillshade. Under the spatial analyst tools and within the Surface toolbox, open the Hillshade tool. Use the default settings, but save the results to your working folder.
Take a moment to look at the results. Does the terrain appear to be similar to what you expected based on your personal experience while collecting the elevation data?