AbstractThe use of the gravity method to predict rock bursts in mines is based on the relationship between the development of a dilatancy process in the exploited rock mass and the time‐dependent gravity anomalies induced by this phenomenon. The differences between successive observations of anomalies and the time behaviour of their trend amplitudes as precursors of preceding changes of rock stability are interpreted. The centres of zones of induced rock density variation are determined by computing the position of singular points of the differences between anomalies.Two gravity surveys have been carried out in the Radbod coal mine (Germany). The first survey took place at the level of the Dickebank seam (depth 1030 m), the second in the Sonnenschein seam (depth 1090 m). The observations were made with Worden and LaCoste‐Romberg (D‐type) gravimeters. The differences between successive anomalies were less than 100 μGal.In the case of the Dickebank seam, the position of singular points demonstrates the effect of two approaching longwalls on a previously mined‐out seam and on the gallery in which the gravity observations were made. In the case of the Sonnenschein seam, the trend amplitudes show distinct variations in the formation of the approaching longwall below the edges of all previously mined‐out seams. In particular, the effect of a remnant pillar has caused the largest gravity gradients. This result corresponds to the existence of a zone of rock‐burst hazard known from test drilling. The computed singular points are grouped together under the remnant pillar indicating two local hazard zones.Both results, the observed development of rock instability with time and the information about the position of the disturbed rock mass relative to the mine workings, are of importance, subsurface gravity surveying can therefore be a valuable tool for predicting