Abstract. New particle formation (NPF) is a key atmospheric process which may be responsible for a major fraction of the total aerosol number burden at the global scale, including in particular cloud condensation nuclei (CCN). NPF has been observed in various environments around the world, but some specific conditions, such as those encountered in volcanic plumes, remain poorly documented in the literature. Yet, understanding such natural processes is essential to better define pre-industrial conditions and their variability in climate model simulations. Here we report observations of NPF performed at the high-altitude observatory of Maïdo (2165 m a.s.l., La Réunion Island) between 1 January and 31 December 2015. During this time period, three effusive eruptions of the Piton de la Fournaise, located ∼39 km away from the station, were observed and documented, resulting in 29 d of measurement in volcanic plume conditions to be compared with 250 “non-plume days”. This dataset is, to our knowledge, the largest ever reported for the investigation of NPF in tropospheric volcanic plume conditions, and it allowed for the first time a statistical approach to characterize the process and also assessment of its relevance with respect to non-plume conditions. NPF was observed on 90 % of the plume days vs. 71 % of the non-plume days during the 4 months when the eruptions occurred. The events were on average detected earlier on plume days, most likely benefiting from larger amounts of precursors available at the site prior to nucleation hours. The overall effect of the plume conditions on the particle growth rate was limited. However, with the exception of September, particle formation rates were significantly higher on plume days. The signature of the volcanic plume on the aerosol spectra up to dp=600 nm was further investigated based on the analysis and fitting of the particle size distributions recorded under in-plume and off-plume conditions. The spectra recorded prior to nucleation hours, in the absence of freshly formed particles, featured a significant contribution of particles likely formed via heterogeneous processes at the vent of the volcano (and assimilated to volcanic primary particles) to the concentrations of the two accumulation modes on plume days. Later on in the morning, the concentrations of the nucleation and Aitken modes showed important variations on plume days compared to event days outside of plume conditions. The spectra recorded on event days, under in-plume and off-plume conditions, were further used to provide an average size distribution of the particles of volcanic origin, which clearly highlighted the dominant contribution of secondary over primary particles (93 %) to the total concentration measured on NPF event days within a volcanic plume. In a next step, particular attention was paid to the concentration of particles with dp>50 nm (N50), used as a proxy for potential CCN population. The contribution of secondary particles to the increase in N50 was the most frequent in plume conditions, and the magnitude of the increase was also more important on plume days compared to non-plume days. Finally, in order to further evaluate the effect of volcanic plume conditions on the occurrence of NPF, we analysed the variations of the condensation sink (CS) and [H2SO4], previously reported to play a key role in the process. Over the investigated months, higher CS (calculated prior to nucleation hours) were observed in plume conditions and coincided with high SO2 mixing ratios. Those most likely compensated for the strengthened loss rate of the vapours and favoured the occurrence of NPF, suggesting at the same time a key role of H2SO4 in the process. This last hypothesis was further supported by the correlation between the formation rate of 2 nm particles (J2) and [H2SO4], and by the fair approximation of J2 that was obtained by means of a recent parameterization of the binary nucleation of H2SO4–H2O. This last result demonstrates that in the absence of direct measurements of [H2SO4] and sub-3 nm particle concentrations, estimates of J2 could be fairly estimated from the knowledge of SO2 mixing ratios only. Finally, the use of the parameterization for ion-induced binary nucleation also highlighted the likely significant contribution of ion-induced nucleation for [H2SO4] below ∼8×108 cm−3.