In order to be able to use the operational Moldavian GNSS Positioning System MOLDPOS efficiently for the determination of normal heights in surveying engineering, e.g. during the construction of a road, an accurate quasigeoid model is needed. The main goal of this thesis is to present a new gravimetric quasigeoid model for Moldova (Mold2012), which has been determined by applying the Least Squares Modification of Stokes’ formula with Additive corrections (LSMSA), also called the KTH method. Due to limited coverage of gravity data, the integration area is often limited to a small spherical cap around the computation point, which leads to a truncation error for geoid height. Molodensky et al. (1962) showed that the truncation error can be reduced by the modification of Stokes’ formula, where the measured gravity data are combined with the low-frequency component of the geoid from a Global Gravitational Model (GGM). The LSMSA technique combines the GGM and the terrestrial data in an optimum way. In order to find the most suitable modification approach or cap size it is necessary to compare the gravimetric height anomalies with the GPS/levelling derived height anomalies, and for this purpose we use a GPS/levelling dataset that consists of 1042 points with geodetic coordinates in the MOLDREF99 reference system and normal heights at the same points given in the height system Baltic 77. The magnitude of the additive corrections varies within an interval from -0.6 cm to -4.3 cm over the area of Moldova. The quasigeoid model which results from combining the ITG-Grace02s solution (with n = M = 170, ψ0 = 3° and σΔg = 10 mGal) and the solution obtained from the modified Stokes’ formula together with the additive correction gives the best fit for the GPS/levelling data with a standard deviation (STD) of ±7.8 cm. The evaluation of the computed gravimetric quasigeoid is performed by comparing the gravimetric height anomalies with the GPS/levelling derived height anomalies for 1042 points. However, the above heterogeneous data include outliers, and in order to find and eliminate these, a corrector surface model is used. This surface provides a connection to the local vertical when the GNSS technique is used. After the elimination of the suspicious outliers (170 points) according to a 2-RMS test, a new corrective surface was computed based on the remaining 872 GPS/levelling points, and the STD of residuals became ±4.9 cm. The STD value for the residuals according to the order of the levelling network for the Mold2012 fitted to the local vertical datum is 3.8 cm for the I-order, 4.3 cm for the II-order, 4.5 cm for the III-order and 5.0 cm for the IV-order levelling network. But the STD of the residuals for the 18 control points indicates a better result where the STD is 3.6 cm and RMS is 3.9 cm and the min and max value of residuals is -5.3 cm and 9.0 cm, respectively. As the STD of the differences in height anomaly are not just the standard error of the height anomalies (quasigeoid model), but it contains also the standard errors of GPS heights and of normal heights. Assuming that the latter STDs are 3 cm and 3.5 cm, respectively, the STD of Mold2012 is estimated to 1.7 cm. / <p>QC 20121127</p>