Since the 1st International Comparison of Absolute Gravimeters (ICAG) and accompanying Relative Gravity Campaign (RGC) held at the BIPM in 1981, repeated ICAG-RGCs have been organized every four years. A total of 19 absolute gravimeters (AG) and 15 relative gravimeters (RG) participated in the 7th ICAG-RGC, which took place in 2005. Co-located absolute and relative gravity measurements as well as precision levelling measurements were carried out.The final version of the absolute g values of the 7th ICAG has been officially released recently. This paper is the final report of the 7th RGC and replaces the preliminary results published earlier. It covers the organization of the RGC and the data processing, analyses RG behaviour, computes g, δg and OAG (offset of AG) and discusses their uncertainties. In preparation for the BIPM key comparison ICAG-2009, a standard data-processing procedure has been developed and installed in the BIPM ICAG-RGC software package, GraviSoft. This was used for the final data processing.

We present a method dedicated to the interpretation of full tensor ͑gravity͒ gradiometry ͑FTG͒ data called tensor deconvolution. It is especially designed to benefit from the simultaneous use of all the FTG components and of the gravity field. In particular, it uses tensor scalar invariants as a basis for source location. The invariant expressions involve all of the independent components of the tensor. This method is a tensor analog of Euler deconvolution, but has the following advantages compared to the conventional Euler deconvolution method: ͑1͒ It provides a solution at every observation point, without the use of a sliding window. ͑2͒ It determines the structural index automatically; as a consequence, the structural index follows the variations of the field morphology. ͑3͒ It uses all components of the measured full gradient tensor and gravity field, thus reducing errors caused by random noise. It is based on scalar invariants that are by nature insensitive to the orientation of the measuring device. We tested our method on both noise-free and noise-contaminated data. These tests show that tensor solutions cluster in the vicinity of the center of causative bodies, whereas Euler solutions better outline their edges. Hence, these methods should be combined for improved contouring and depth estimation. In addition, we use a clustering method to improve the selection of solutions, which proves advantageous when data are noisy or when signals from close causative bodies interfere.

In mineral and oil exploration, gravity gradient data can help to delineate small-scale features that cannot be retrieved from gravity measurements. Removing high-frequency noise while preserving the high-frequency real signal is one of the most challenging tasks associated with gravity gradiometry data processing. We present a method to reduce gravity and gravity gradient data noise when both are measured in the same area, based on a least-squares simultaneous inversion of observations and physical constraints, inferred from the gravity gradient tensor definition and its mathematical properties. Instead of handling profiles individually, our noise-reduction method uses simultaneously measured values of the tensor components and of gravity in the whole survey area, benefiting from all available information. Synthetic examples show that more than half of the random noise can be removed from all tensor components and nearly all the noise from the gravity anomaly without altering the high-frequency information. We apply our method to a set of marine gravity gradiometry data acquired by Bell Geospace in the Faroe-Shetland Basin to demonstrate its power to resolve small-scale features.