Uncertainty handling in multibeam bathymetric mapping

The purpose of this research is to develop a rational uncertainty management strategy in multibeam bathymetric mapping. Three approaches to handling random errors, outliers, and systematic errors are developed. Each of these is tested using data generated by Reson Seabat 9001 and Simrad EM1000 multibeam systems in an inter-tidal area. The ground truth of the test area is mapped by photogrammetric surveys at low tide.;To assess random errors, a new neighborhood structure is proposed by defining it as the horizontal probability distribution function of the spatial data. A griding algorithm is developed on the basis of this new neighborhood structure. Guidelines for choosing cell size and horizontal probability distribution function are given. Tests using real data demonstrate that the algorithm produces the same beam accuracy pattern with two existing methods. The standard deviations of this algorithm are 2-3 cm smaller than results of one of the existing methods because the latter include sea-floor roughness between two neighboring soundings. One by-product of the algorithm, termed reliability in this dissertation, can be used to explain the extremely small standard deviations.;An automatic outlier detection algorithm is developed. It has three properties. First, its working window is defined by both data-oriented and space-oriented approaches. Second, it employs discordancy tests based on a uniform distribution by recognizing that the distribution or multibeam data in the vertical dimension is dominated by the sea floor topography provided. The sea floor is uniformly sampled and a fairly small working window is used. Third, no a priori knowledge of sea-floor topography is required. Tests using real data show that the algorithm matches the results of interactive editing with a Rand index of 0.95, and that no distortion in the bathymetry is generated.;To evaluate systematic errors and to assess external accuracy, a set of software tools for quantitative comparison and evaluation of multibeam data is developed. Real data comparison demonstrates that the software tools can detect roll offset and rate errors, and real time heave error present in the data. Between-array consistency of Seabat 9001 data, and external comparison between EM1000 data and the ground truth are presented.