Calibration of the superconducting gravimeter T011 by parallel observation with the absolute gravimeter FG5 #210--a Bayesian approach

Abstract

SUMMARY The superconducting gravimeter (SG) is a relative gravimeter and its sensitivity must be calibrated. One of the possible methods of calibration is parallel observation with an absolute gravimeter. Since the precision of the calibration is roughly proportional to N −1/2 , where N is the number of data, a long time span of observation consisting of a large number of data will be helpful for a better result. In this work, a 27 day long experiment of absolute gravity measurements with the absolute gravimeter FG5 #210 was made for calibration of the SG T011 at Matsushiro, Japan. The effective number of measurements was about 100 000. The standard deviation of single measurements varied from 50 to 90 nm s −2 , depending on the weather condition. A regression analysis of the data from both gravimeters with the usual least-squares method yielded (−928.01 ± 0.34) nm s −2 V −1 for the (negative) scale factor of the SG T011. Although at first this result appeared satisfactory, a closer look at the data revealed existence of an irregular ‘drift’ in the absolute gravity data and possibly a systematic bias in the estimated scale factor. For unbiased estimation of the scale factor, two additional methods of data analysis were employed. One is the method based on tidal analysis of gravity data from both absolute and relative gravimeters, and the other is a modification of the simple least-squares method. Both of these methods separate the drift from real gravity changes based on the Bayesian statistics, so that estimation of the scale factor is free from the effect of the drift. The scalefactors obtained from these methods were found to be slightly larger (in absolute values) than those from the simple least-squares method. Although the tidal analysis method was found to also be potentially useful for phase calibration, it had the largest estimation error of the scale factor among the three methods. Considering the accuracy as well as the precision of calibration, the scale factor from the modified least-squares method was regarded as being most reliable. Thus, the final result is (−928.79 ± 0.36) nm s −2 V −1 . The relative precision of this result is 0.039 per cent. This is one of the most precise calibration results for an SG obtained from a single absolute gravity experiment. The cause of the drift in the absolute gravity measurements is unknown. The change in the sensitivity of about 0.5 per cent associated with the replacement of the electronics for the SG has been verified quantitatively.