The continuous gravity observation with a superconducting gravimeter (SG) #016 was started on March 22, 1993 at Syowa Station, Antarctica (69.0 N, 39.5 E). Since then, the observation is continued by Japanese Antarctica Research Expedition (JARE). This thesis mainly describes the analysis results for tidal gravity changes using the Syowa SG data and discusses the various oceanic effects on its observation. The tides treated here are those for the 12 hours to 1 year in the period. In connection with the analysis for long-period tides, the gravity effects at Syowa Station due to the polar motion were reanalyzed and compared with other two SG sites in the mid latitude (i.e. Esashi, Japan and Canberra, Australia). The effect of Sea Surface Height (SSH) variations are also discussed mainly focusing into the annual gravity changes.<br /> The thesis consists of nine chapters. In Chapter 1 (Introduction), first, the importance of gravity observation made at high latitude is described. The remaining part of Chapter 1 explains the outline of contents of this thesis. Chapter 2 reviews the tidal phenomena and the definition of several quantities appeared in the thesis. Chapter 3 describes the method to estimate the ocean effects and the computer program used in this thesis. The characteristics of SG is described in the first part of Chapter 4 in connection with the observation results shown in the later Chapters. The locality of Syowa Station, the procedures for setting up of the SG and the data acquisition system used in the observation are also introduced in Chapter 4. We used the computer program called BAYTAP-G and -L for the tidal analysis. The method and some problems on the actual analysis used this analysis method are mentioned in Chapter 5. The observed results are discussed in Chapters 6, 7 and 8 for the short-period tides, the long-period tides and the polar motion effect especially focusing into the annual component, respectively. Finally, the concluding remarks are given in Chapter 9.<br /> In Chapter 6, we reexamined the gravity tidal factor (δ-factor) of the diurnal and semi-diurnal tides at Syowa Station. The 2-year SG data obtained in the period from March 1993 to March 1995 were used in the analysis. The ocean tide effects (the effects of the attraction and loading due to the ocean mass) were estimated using a new global ocean tide model by Matsumoto et al. (1995). As the δ-factors corrected for the ocean tide effects, we obtained the values of 1.144, 1.127, 1.157 and 1.111 for O<SUB>1</SUB>,K<SUB>1</SUB>,M<SUB>2</SUB>,and S<SUB>2</SUB> waves, respectively. We compared the observed mean δ-factors with the two theoretical values inferred from Wahr's (1980) theory and the Dehant's (1987) theory. The discrepancies between our values and the Wahr's theory are at about 0.2% for both the diurnal and semi-diurnal tides and those with the Dehant's are at about 0.6% and 2% for the diurnal and semi-diurnal tides, respectively. The Showa SG data indicate that the Wahr's theory is much consistent with the observation than the Dehant's theory. Judging from the consistency among the three observation results obtained by the three different gravimeters (Ogawa et al., 1991; Kanao and Sato, 1995; this study), and from the results for the computations of the ocean tide effects, it is highly probable that the large discrepancy exceeding 10% in the δ-factors for the semi-diurnal tides at Syowa Station, which has been pointed out by Ogawa et al. (1991), is mainly caused by their inaccurate estimation of the ocean tide effects.<br /> In Chapter 7, we examined the long-period tides (M<SUB>f</SUB> and M<SUB>m</SUB> waves) based on the 2 years SG data which are the same data as those used in the analysis for the short period tides (Chapter 6). The obtained amplitudes, phase lags and amplitude factors (δ-factors) are 11.642±0.035μGal, -0.12°±0.17°and 1.1218±0.0034 for the M<SUB>f</SUB> wave, and 6.143±0.058μGal, 0.33°±0.54°and 1.1205±0.0106 for the M<SUB>m</SUB> wave, respectively (1 μGal＝10<SUP>-8</SUP> ms<SUP>-2</SUP>). The ocean tide effects at the observation site were estimated using the five global ocean tide models: equilibrium ocean tide model, Schwiderski (1980) model, Dickman (1989) model, CSR model (Eanes, 1995), and Desai & Wahr (1995) model. The averages of the five estimates are 0.433μGal and 0.243μGal in amplitude and 192.9°and 179.5°in phase for the M<SUB>f</SUB> and M<SUB>m</SUB> waves, respectively. The five estimates differ by a maximum of 0.057 μGal in amplitude and 18.7°in phase for the M<SUB>f</SUB> wave, and by 0.033 μGal and 6.4°for the M<SUB>m</SUB> wave. The estimated M<SUB>m</SUB> phases are nearly 180°for the five models, and the variation of their values among the models is relatively small compared with that of the M<SUB>f</SUB> phases. These indicate that the M<SUB>m</SUB> wave is much close to an equilibrium tide than the M<SUB>f</SUB> wave. Due to the variation of the ocean tide corrections, the corrected δ-factors were scattered within the ranges of 1.157 to 1.169 for the M<SUB>f</SUB> wave and of 1.161 to 1.169 for the M<SUB>m</SUB> wave. However, it is noted that the mean δ-factors of the five ocean models, i.e. 1.162±0.023 for the M<SUB>f</SUB> wave and 1.165±0.014 for the M<SUB>m</SUB> wave, prefer slightly larger value rather than those estimated from the theory of the elastic tide.<br /> In Chapter 8, the results for the polar motion effect are described. First, the previous analysis results and the problems on the analysis for this effect, which were obtained from the analysis by Sato et al. (1997) using the two years Syowa SG data, are summarized. They discussed the two problems, i.e. on an interference problem between the annual and Chandler components and on an effect of the inaccurate estimation of step-like changes including the observed data. Based on their experiences, a revised analysis model is applied here and the analysis is carried out using the Syowa SG data much longer than those used in the previous analysis. Thus, it is revised so that (1) the annual component of the polar motion data was excluded from the IERS EOP (International Earth Rotation Service Earth Orientation Parameter) data before fitting and (2) the term to estimate the step-like changes using the Heviside's function was added to the previous model. It was shown that, by using the revised model, the analysis error for the annual component is improved by about 15% in the case of Esashi, for example. The reliability of the analysis results is also affected by the stability of the period of Chandler component in time. We, therefore, examined this using the 22 years IERS EOP data, and we recognized that the period of Chandler component is stable within ±1.3 days during the observation period of the 17 years from 1983 to 1999. We obtained a value of 435.4 as the mean Chandler period averaged over the 17 years. This value was used for the fitting though this study.<br /> In Chapter 8, we discussed also the results for the estimation of annual gravity changes by calculating the four effects of the solid tide, ocean tide, polar motion and SSH variations, In order to pull out the effect of mass changes in the SSH variations, it is needed to estimate the thermal steric changes in SSH variations, and to correct its effect. We evaluated the steric coefficient based on mainly the POCM (Parallel Ocean Climate Model, Stammer, 1996) SSH (Sea Surface Height) data and the SST (Sea Surface Temperature) data, and we obtained a value of 0.60×10<SUP>-2</SUP>m/℃. The predicted annual effect at the three observation sites (i.e. Esashi in Japan, Canberra in Australia and Syowa in Antarctica) were compared with the actual data obtained from the superconducting gravimeters respectively installed at these three sites. The results of the comparison indicate that the predictions agree with the observations within 20% in amplitude (i.e. within 0.2 μGals, where 1 μGal＝1×10<SUP>-8</SUP> ms<SUP>-2</SUP>) and 10°in phase at each observation site, when we use the above steric coefficient. We have also tested other values for the steric coefficient, i.e. 0.0×10<SUP>-2</SUP>m/℃ and 1.0×10<SUP>-2</SUP>m/℃, but find that the fit to observations is clearly better at 0.60×10<SUP>-2</SUP>m/℃. For comparison, we have also evaluated the SSH effect using TOPEX/POSEIDON data (T/P data). The results from the T/P data indicate a very similar dependence on magnitude of the steric coefficients with that obtained from POCM data, although there exist some systematic differences in amplitude and phase between the two SSH data of the POCM and T/P. It is worth noting that the gravity observations favor a steric coefficient of 0.60×10<SUP>-2</SUP>m/℃. This means that the gravity observations support the steric coefficient which was independently estimated from the SSH and SST data. This may be important on future applications of the gravity observation as a data to monitor the mass changes in the oceans which could not be detected from the precise satellite altimetry such as the T/P altimeter. We consider that the agreement between the observation and prediction shown here gives us a base to study the Earth's response to the Chandler motion and/or the excitation problem on it.