Inter-comparison of plume model mass eruption rate predictions based on the 2010 Eyjafjallajökull eruption

During explosive eruptions, integral plume models are typically used to estimate real-time source parameters (e.g. mass eruption rate, plume height, and total grain size distribution) which are crucial for initiating tephra dispersal models and subsequently providing ash concentration forecasts for the aviation industry. However, uncertainties within integral models can significantly affect the accuracy of these estimates. Uncertainties may arise from the inclusion of additional parameters in more complex models, or from assumptions such as steady-state source conditions during inherently unsteady eruptions. This paper compares two integral plume models of differing complexity, FPLUME (by Folch et al., 2016) and REFIR (by Dürig et al., 2018; Dioguardi et al., 2020), to evaluate how well they replicate the mass eruption rate of the unsteady 2010 Eyjafjallajökull eruption in Iceland. In doing so, the aim is to evaluate whether the use of additional input parameters provides improved model accuracy when predicting mass eruption rates, and whether integral models can effectively capture unsteady source behaviour. Results highlight trade-offs between model complexity, parameter sensitivity, and the ability of integral frameworks to represent time-dependent processes in real time. Greater model complexity does not necessarily yield improved mass eruption rate predictions; instead, accurate parameterisation of the entrainment coefficient β is identified as the most crucial factor for improvement.