Methods for assessing the consistency of the New National Height Model

Abstract: Digital Elevation Models (DEM) are a simple representation of the Earth’s surface. DEMs play an important role in the field of remote sensing and GIS and are used as basis for mapping and analysis for a vest majority of scientific applications. There are many ways of producing DEMs, however the direct geo-referencing technology has made Airborne Laser Scanning (ALS) a preferred technology for the acquisition of accurate surface models over broad areas. ALS uses LiDAR (Light Detection and Ranging) which uses light in a form of pulsed laser to measure distances. Before the introduction of the DEM called Ny Nationell Höjdmodell (NNH), the highest level of height data over Sweden was the GSD-altitude data (Geographical Sweden Data). The NNH was a project by Lantmäteriet, where between 2009-2019 the entire Sweden was laser scanned. The product was a new height model called Laser Data NH with positional accuracy of 0,1 m in height and relative accuracy of 0,15 m. This project focuses on testing few methods for consistency assessment between the overlapping strips using linear features. Linear features are extracted for each overlapping area, based on intersection between planar patches extracted from gable rooftops. The first method of this study computes the distance between the overlapping areas without linear features, using two approaches: cloud-to-cloud distance and mesh-to-cloud distance. The second method computes the transformation shifts and rotations needed for the linear features to align by registering the strips with both levelled and not levelled registration. In the third method, distances and angles are measured between the lines, to further analyze how well the strips fit together. The distances are measured as distance between a mid-point of one line in the first LiDAR strip and the line on the second LiDAR strip, for all linear features. The distances were measures both as 3D distances and separately as horizontal and vertical distances. As a final step a hypothesis testing was performed to determine whether the distances and angles between the lines are significant or whether any systematic error is present in the point cloud. Based on the results obtained from the first method, significant distance between the point clouds was obtained. The results from the mesh-to-cloud distance yielded better result with higher uncertainty. According to the second method significant distances between the linear features were obtained based on the registration. The mean absolute error of the registrations showed an error at a dm level, with a minimal rotation in the vertical plane for the coalignment for the levelled registration. The third method showed a mean distance between the linear features of 20 cm. Moreover, this method showed a significant inconsistence between the linear features in the vertical plane based on the high standard uncertainty.