PrintIn this case, we have a 500 meter resolution DEM, so interpolation would not normally be necessary, assuming that this resolution was adequate for our purposes. For the purposes of this project, we will create a random set of spot heights derived from the DEM, so that we can work with those spot heights to evaluate how well different interpolation methods reconstruct the original data.
The easiest way to do this is to use the Geostatistical Analyst - Utilities - Subset features tool. This tool is self-explanatory - you will want to run it on the allSpotHeights layer, and make a sample of 5% of the control points, or, alternatively, you can specify how many points you want (about 1500). You do not need to specify both a training feature class and a test feature class (that is helpful if you want to split a dataset for Machine Learning classifier). Here we are just interested in a 5% sample and only need one feature class. Make sure you turn on the Geostatistical Analyst extension to enable the tool.
An alternative approach might be to use the Data Management - Create Random Points and the Spatial Analyst - Extraction - Extract By Points tools. If you have the appropriate licenses (Advanced), then you can experiment with using these instead to generate around 1500 spot heights.
A typical random selection of spot heights is shown in Figure 6.5 (I've labeled the points with their values).
