Although satellite gravity gradient data plays a great role in determining short-wavelength part of static gravity field model, accuracy of the long-wavelength part of gravity field model recovered by them are poor, which leads to only a few applications in time-variable gravity investigation. The reason is that some factors limit the accuracy of the gravity field recovered using gradient data, including accuracy of the gravity gradient observations, measurement bandwidth (MBW) of gradiometer, satellite inclination, etc. This paper aims at analyzing the influence of these limitations on gravity field recovery and discusses the possibility of time-variable gravity field detection by using gravity gradient observations. Firstly, for arbitrary satellite orbit inclination, we give the frequency distributions of all the components of gravity gradients (i.e. Txx,Tyy,Tzz,Txy,TxzandTyz, ). The results show that the maximum frequency of each component of the gravity gradients is the same, i.e. l/Ts (l is degree of the gravity field model, Ts is the orbital periods), and it is not influenced by the inclination of the satellite orbits. Secondly, the paper gives a theory proof to explain why only the low orders of the coefficients are influenced by polar gaps. Big polar gaps are experimented by a numerical test with inclination of 45°. Finally, considering that the measurement bandwidth can be expanded and accuracy of gradient observations can be improved by superconducting gravity gradiometer (SGG) compared to gradiometer used in Gravity field and steady-state Ocean Circulation Explorer (GOCE), the possibility of detecting time-variable gravity using gravity gradient observations is discussed. The results show that the SGG creates errors in MBW with magnitude of 0.014 mE, which is smaller than the magnitude of the time-variable gravity gradient signals (i.e., 0.02 mE) derived from Gravity Recovery and Climate Experiment (GRACE) gravity field models. This indicates the potential of SGG in time-variable gravity detection.