Analysis of the accuracy of different commercial GPS network adjustment software packages
Abstract
The surveying and geomatics community is currently faced with a wide range of commercial
proprietary GPS network adjustment software to choose from. Each of the software package is
however unique in some functions, with different data processing algorithms applied. The
difference in mathematical models adopted by the software packages influences the adjusted
coordinates and therefore leads to differences in final coordinates. Surveyors however continue to
use these software without knowing their comparative efficacy hence the need for congruity
analysis. This research therefore aims at assessing the accuracy of four commercial GPS network
adjustment software packages in the user market namely; Carlson SurveyGNSS 2016, CHC
Geomatics Office (CGO) version 1.0.1, GNSS Solutions (3.80.8) and Trimble -Business Center
(TBC) version 3.50.
Using 6 control stations, the evaluation method involves the comparison of coordinate residuals
obtained as deviations of the final (adjusted) coordinates, from the software, from the “known”
station coordinates. The station coordinates obtained from the Surveys and Mapping Department,
Entebbe, were considered as the ‘known coordinates’ in the residual computation.
The average residual values in the Northing, Easting and Orthometric height were below 10 cm in
CGO, GNSS Solutions and TBC in both Cases 1 and 2 while SurveyGNSS produced average
residual values ranging between 10 cm to 20 cm in the North, East and Height components. Since
the adjusted coordinates were compared with their ‘known’ values, the software package that
produced the least deviations was considered the most accurate amongst the four software
packages. The one that produced the highest deviations was however the least accurate. Further
analysis was conducted basing on the mean residual values, 2D and 3D residual errors, Standard
Deviations and Root Mean Square Errors (RMSE).
The RMSE of 0.043 m in Case 1 confirmed TBC as the most accurate amongst the test software
packages while SurveyGNSS produced the highest RMSE of 0.170 m implying that it is the least
accurate. CGO however performed slightly better that TBC in Case 2 with RMSE of 0.071 m
compared to 0.093 m of TBC. This could be partly attributed to the proprietary CHC N7 GNSS
receivers used at the Survnet CORS base stations. A similar research should be conducted to incorporate the effect of varying the proprietary GNSS receivers during observations on the
accuracy of the software packages.