Investigation on monitoring of the earth's gravity field by monthly GRACE solutions and its sensitivity

Since 2002, GRACE (Gravity Recovery and Climate Experiment) satellite system has provided important information on the temporal variation of the Earth?s gravity field. Gravity field data deduced from this system, which is composed of two low Earth orbit artificial satellites, is published as harmonic coefficients monthly by some data processing centers to the users. In this study, having used the coefficient solutions belonging to 92 months between 2003 January and 2010 September, geopotential changes (equivalent water thickness, geoid height and gravity changes) as well as GRACE sensitivity against those changes are investigated.The geopotential changes are examined depending on 1) ordinary difference of annual GRACE models created as weighted arithmetic means and 2) time series including whole monthly solutions, separately. Former differencing method is generally used for some special aims (for example, monitoring of some tectonic events at regional scale) whereas latter is the mostly chosen method in which the annual changes are estimated using a proper velocity model. This study examines these changes and their standard deviations at global and regional scales for both methods. Furthermore, the solutions for different smoothing radii (halfwidths) of Gaussian filter and the decorrelation method, the so-called C-P7M8 here, as well as DDK data type are realized and the deduced changes and their standard deviations from different solutions are compared. The monitored geopotential changes are affected by the filtering and filtered data type. Also the standard deviations of the corresponding changes depend on the point position and the filter halfwidth. Moreover, while the standard deviations from the first method depend on the standard deviations of the harmonic coefficients, the standard deviations of the annual changes estimated from a fitted velocity model, which is assumed as deterministically correct, to the point geopotential changes are independent from the ones of harmonic coefficients. Thus, the sensitivity of GRACE in determination of the corresponding changes is examined depending on these second kind of standard deviations having compared the above-mentioned different solutions. According to the numerical example results, the detectable changes by the GRACE solutions with the probability of more than 80% power of the test for the worst cases are as follows: i) At global scales; 2.1 cm/year EWT velocity, 0.05 cm/year annual geoid height change and 0.84 ? Gal/year annual gravity change, ii) for the region covering Turkey ([30o ??? 45o; 0o ??? 50o]), 0.7 cm/year EWT velocity, 0.024 cm/year annual geoid height change and 0.30 ? Gal/year annual gravity change.In addition, it may be concluded that the geopotential changes from GRACE solutions are influenced by numerous effects. Therefore, it is natural to expect some differences in the magnitude and distribution of the changes in spatial domain as well as in their standard deviations depending on, for instance, different processing strategies of the data processing centers, the filtering method applied to the coefficients, reference model/GRACE model used in the computations, data span in the analysis and the assumed velocity model. Thus, accordingly, the confidence interval estimation and their representation in spatial domain is examined to show the quality of the obtained annual changes. It is desired to measure the accuracy of the corresponding solution having representing the confidence interval limits with 95% confidence level in the spatial domain for the estimated annual changes and their standard deviations from the data span of 92 months belonging to the studied points. According to the numerical examples, it is observed that some unremoved systematic errors in the spatial domain decreases and increases in the lower limit and upper limit maps, respectively. Therefore, it is deduced that some systematic errors observed in the geopotential changes may actually have random characters. Moreover, from the comparison of the confidence interval maps of different filtered solutions, it is clearly seen that some anomalies in the spatial domain are actually be the effect of random errors. Accordingly, it is suggested that the confidence interval estimation should be realized to show the accuracy of the corresponding geopotential change solution.