DETERMINING WHERE INDIVIDUAL VEHICLES SHOULD NOT DRIVE IN SEMIARID TERRAIN IN VIRGINIA CITY, NV
AdvisorTaranik, James V
StatisticsView Usage Statistics
This thesis explored elements involved in determining and mapping where a vehicle should not drive off-road in semiarid areas. Obstacles are anything which slows or obstructs progress (Meyer et al., 1977) or limits the space available for maneuvering (Spenko et al., 2006). This study identified the major factors relevant in determining which terrain features should be considered obstacles when off-road driving and thus should be avoided. These are elements relating to the vehicle itself and how it is driven as well as terrain factors of slope, vegetation, water, and soil. Identification of these in the terrain was done using inferential methods of Terrain Pattern Recognition (TPR), analyzing of remotely sensing data, and Digital Elevation Map (DEM) data analysis. Analysis was further refined using other reference information about the area. Other factors such as weather, driving angle, and environmental impact are discussed. This information was applied to a section of Virginia City, Nevada as a case-study. Analysis and mapping was done purposely without field work prior to mapping to determine what could be assessed using only remote means. Not all findings from the literature review could be implemented in this trafficability study. Some methods and trafficability knowledge could not be implemented and were omitted due to data being unavailable, un-acquirable, or being too coarsely mapped to be useful. Examples of these are Lidar mapping of the area, soil profiling of the terrain, and assessment of plant species present in the area for driven-over traction and tire punctures. The Virginia City section was analyzed and mapped utilizing hyperspectral remotely sensed image data, remote-sensor-derived DEM data was used in a Geographical Information Systems (GIS). Stereo-paired air photos of the study site were used in TPR. Other information on flora, historical weather, and a previous soil survey map were used in a Geographical Information System (GIS). Field validation was used to check findings.The case study's trafficability assessment demonstrated methodologies of terrain analysis which successfully classified many materials present and identified major areas where a vehicle should not drive. The methods used were: Manual TPR of the stereo-paired air photo using a stereo photo viewer to conduct drainage-tracing and slope analysis of the DEM was done using automated methods in ArcMap. The SpecTIR hyperspectral data was analyzed using the manual Environment for Visualizing Images (ENVI) software hourglass procedure. Visual analysis of the hyperspectral data and air photos along with known soil and vegetation characteristics were used to refine analyses. Processed data was georectified using SpecTIR Geographic Lookup Table (GLT) input geometry, and exported to and analyzed in ArcMap with the other data previously listed. Features were identified based on their spectral attributes, spatial properties, and through visual analysis. Inaccuracies in mapping were attributable largely to spatial resolution of Digital Elevation Maps (DEMs) which averaged out some non-drivable obstacles and parts of a drivable road, subjective human and computer decisions during the processing of the data, and grouping of spectral end-members during hyperspectral data analysis. Further refinements to the mapping process could have been made if fieldwork was done during the mapping process.Mapping and field validation found: several manmade and natural obstacles were visible from the ground, but these obstacles were too fine, thin, or small to be identified from the remote sensing data. . Examples are fences and some natural terrain surface roughness - where the terrain's surface deviated from being a smooth surface, exhibiting micro- variations in surface elevation and/or textures. Slope analysis using the 10-meter and 30-meter resolution DEMs did not accurately depict some manmade features [eg. some of the buildings, portions of roads, and fences], evident with a well-trafficked paved road showing in DEM analysis as having too steep a slope [beyond 15°] to be drivable. Some features had been spectrally grouped together during analysis, due to similar spectral properties. Spectral grouping is a process where the spectral class's pixel areas are reviewed and classes which have too few occurrences are averaged into similar classes or dropped entirely. This is done to reduce the number of spectrally unique material classes to those that are most relevant to the terrain mapped. These decisions are subjective and in one case two similar spectral material classes were combined. In later evaluation should have remained as two separate material classes. In field sample collection, some of the determined features; free-standing water and liquid tanks, were found to be inaccessible due to being on private land and/or fence secured. These had to be visually verified - photos were also taken. Further refinements to the mapping could have been made if fieldwork was done during the mapping process. Determining and mapping where a vehicle should not drive in semiarid areas is a complex task which involves many variables and reference data types. Processing, analyzing, and fusing these different references entails subjective manual and automated decisions which are subject to errors and/or inaccuracies at multiple levels that can individually or collectively skew results, causing terrain trafficability to be depicted incorrectly. That said, a usable reference map is creatable which can assist decision makers when determining their route(s).