Some statistical tests in the study of terrain modelling

Abstract

This paper describes some formative steps taken to create statistical models of the course elevation profiles of the US Army's vehicle test courses at Aberdeen, Maryland and at Yuma, Arizona. It was decided first to test the data, the output from various profilometer runs, for linearity, stationarity and Gaussianity. This paper attempts to explain the statistical terminology and how the statistics are used. The appendices are reserved for the technicalities of the tests. To test for linearity, a test developed by Keenan was chosen. The Kolmogorov-Smirnoff test was used to check for Gaussianity. Stationarity was tested by breaking up the data into segments of equal length and checking whether the variance was the same for each segment. The data for each of the test courses were broken up into segments of equal length and the statistical tests run on each segment. The results show that the Belgian Block course is linear and Gaussian but not stationary, and Perryman3 is neither linear, nor Gaussian, nor stationary. The outcome of this study leads to the model building for Belgian Block and Perryman3 profiles.