A methodology for reducing the effect of meteorological parameters on a continuously recording gravity meter

Gravity changes in geodynamically active areas areusually monitored by using discrete measurements. The temporalresolution, which is only as good as the repeat rate of the observations(usually ranging between one month and one year), isthe main drawback of such measurements. To achieve a bettertemporal resolution and thus improve the possibilities of the investigationtask, continuous gravity measurements have to be performed.However, temperature and pressure fluctuations seriouslyaffect the output of continuously running spring gravity meters.In this paper, a methodology allowing the signal from a recordinggravity meter to be compensated for the effect of temperatureand pressure is discussed. To model how the meter output dependson the interfering signals, both polynomial (LMS) andnonlinear (neuro-fuzzy) forms of three different model structureshave been tested. The results obtained through the nonlinear algorithmare satisfactory and could constitute the background forthe implementation of a real-time correction system using a fuzzymicro-controller chip.