During operational ash-cloud forecasting, prediction of ash concentration and total erupted mass directly depends on the determination of mass eruption rate (MER), which is typically inferred from plume height. Uncertainties for plume heights are large, especially for bent-over plumes in which the ascent dynamics are strongly affected by the surrounding wind field. Here we show how uncertainties can be reduced if MER is derived directly from geophysical observations of source dynamics. The combination of infrasound measurements and thermal camera imagery allows for the infrasonic type of source to be constrained (a dipole in this case) and for the plume exit velocity to be calculated (54-142m/s) based on the acoustic signal recorded during the 2010 Eyjafjallajökull eruption from 4 to 21 May. Exit velocities are converted into MER using additional information on vent diameter (50±10m) and mixture density (5.4±1.1kg/m3), resulting in an average ~9×105kg/s MER during the considered period of the eruption. We validate our acoustic-derived MER by using independent measurements of plume heights (Icelandic Meteorological Office radar observations). Acoustically derived MER are converted into plume heights using field-based relationships and a 1D radially averaged buoyant plume theory model using a reconstructed total grain size distribution. We conclude that the use of infrasonic monitoring may lead to important understanding of the plume dynamics and allows for real-time determination of eruption source parameters. This could improve substantially the forecasting of volcano-related hazards, with important implications for civil aviation safety.