A robust approach to terrestrial relative gravity measurements and adjustment of gravity networks

Like many geophysical observations, relative gravity (RG) measurements are affected by random errors, systematic errors, and occasional blunders. When RG measurements are used to build large gravity networks in remote areas under adverse environmental or logistical conditions (such as extreme temperatures, heavy precipitation, rugged terrain, difficult or dangerous roads, and high altitudes), it is more likely for significant errors to occur and accumulate. Therefore, obtaining accurate gravity estimates at regional gravity networks largely depends on defensive data collection protocols and robust adjustment techniques. In this work, we present a measurement field protocol based on highly redundant observation patterns, and a two-step least squares adjustment scheme implemented as a MATLAB package. This software helps us identify blunders, mitigates the impact of random errors, and downweights or removes outlier observations. The methodology also guarantees that adjusted gravity values have well-constrained standard error estimates. We illustrate the capabilities of our approach through the case study of the Bolivian gravity network, where we determined the acceleration due to gravity at 2548 stations that spread over difficult and sometimes extreme environments, with a typical level of uncertainty of 0.10–0.15 mGal.