A new computational approach to reduce the signal from continuously recording gravimeters for the effect of atmospheric temperature

The experience of several authors has shown that continuous measurements of the gravity field, accomplished through spring devices, are strongly affected by changes of the ambient temperature. The apparent, temperature-driven, gravity changes can be up to one order of magnitude higher than the expected changes of the gravity field. Since these effects are frequency-dependent and instrument-related, they must be removed through non-linear techniques and in a case-by-case fashion. Past studies have demonstrated the effectiveness of a Neuro-Fuzzy algorithm as a tool to reduce continuous gravity sequences for the effect of external temperature changes. In the present work, an upgraded version of this previously employed algorithm is tested against the signal from a gravimeter, which was installed in two different sites over consecutive 96-day and 163-day periods. The better performance of the new algorithm with respect to the previous one is proven. Besides, inferences about the site and/or seasonal dependence of the model structure are reported.