On the statistical tests over Fennoscandian GNSS/levelling networks

Sammanfattning: In Geodesy and Surveying we work with a large amount of observations which alwayscontains different types of errors. The errors decrease the quality of the observations andpropagate to the results. Therefore, detection and removing the gross errors are of vitalimportance. The Global Navigation Satellite Systems (GNSS) can be used to measure theellipsoidal height and by subtracting an existing geoid height from that, the orthometric heightcan be determined. There is a simple linear relation amongst these triple heights, whichcannot be fulfilled due to the presence of different types of error. One of the important sort ofthese errors is gross errors.This study concerns about investigation and detection of blunders or gross errors on the 4346GNSS/levelling points over Fennoscandia. Each country has its own data set with specificprecision. The well-known gravity model EGM08 is used to compute the geoid heights withrespect to WGS84 reference ellipsoid. We have a large amount of data and we expect thattheir errors follow the normal distribution. The main aim of this thesis is to apply some datascreening methods both before and after adjustment process in such a way that the normaldistribution of the data set is achieved by eliminating the erroneous data. This will be done byperforming the pre- and post-adjustment data screening. For the pre-adjustment we performeddata filtering, test of normality of observations and test of their variances for theGNSS/levelling data over Sweden, Denmark, Norway and Finland. We used the 4-, 5- and 7-parameter corrective surfaces for modelling the systematic trends of the differences betweenthe EGM08 geoid model and the ellipsoidal and orthometric height differences. The test ofnormality of residuals, global test of variance, Baarda’s data snooping and Tau test will beperformed after the removal of the trends. Numerical studies show that the GNSS/levellingdata of Sweden, Denmark and Finland are of good qualities, but the claimed errors for thedata are rather optimistic. The situation was complicated for the data of Norway and we couldnot see the normality of the data and even the claimed accuracies seem to be optimistic.