Abstract. Coastal population in the Western Mediterranean basin is frequently affected by high-impact weather events that produce loss of life and property for an invaluable amount. Among the wide spectrum of maritime severe weather events, tropical-like Mediterranean cyclones (a.k.a. medicanes) draw particular attention, specially due to their poor predictability. The accurate prediction of this kind of events still remains a key challenge to our community, mainly because of (i) the errors in the initial conditions, (ii) the lack of accuracy modeling micro-scale physic processes and (iii) the chaotic behavior inherent to current numerical weather prediction models. In particular, the 7th October 2014 Qendresa medicane, that took place over the Sicilian channel, affecting the Islands of Lampedusa, Pantelleria and Malta was selected for this study because of its extremely low predictability behavior in terms of its track and intensity. To enhance the prediction of Qendresa, a high-resolution (4-km) ensemble-based data assimilation technique, known as Ensemble Kalman Filter (EnKF) is used. In this study both in-situ conventional and satellite-derived observations are assimilated with the main objective of improving Qendresa's model initial conditions and thus, its subsequent forecast. The performance of the EnKF system and its impact on the Qendresa forecast is quantitatively assessed using different deterministic and probabilistic verification methods. A discussion in terms of the relevant physical mechanisms adjusted by the EnKF is also provided. Results reveal that the assimilation of both conventional and satellite-derived observations improves the short-range forecasts of the trajectory and intensity of Qendresa. In this context, it is shown the relevance of assimilating satellite-derived observations to improve the pre-convective estimation of Qendresa's upper-level dynamics, which is key to obtain a realistic track and intensity forecast of this event.