During two growing seasons of a maritime pine stand, in 2014 and 2015, ozone (O-3) fluxes have been determined using the eddy covariance (EC) method and compared to the outputs of a big-leaf O-3 deposition model including stomatal, cuticular and soil pathways. The model developed in this study generally allowed to properly reproduce the measured ozone deposition. Ozone fluxes showed a strong reduction during two water stressed periods in September 2014 and July 2015. The model partly explain this fall due to the reduction of stomatal deposition. Despite this stomatal closure, measured O-3 fluxes presented systematically lower negative values than the model outputs, and sometimes even positive values around midday during periods marked by strong water stress. In other words, the difference between observed and modelled O-3 fluxes (hereinafter referred to as the residual O-3 flux) is systematically positive on daytime during these water-stressed periods. This positive residual flux traduced the existence of an O-3 source below the flux measurement level, responsible for positive fluxes that counterbalance deposition fluxes. We developed an O-3 production module based on a terpene emission algorithm and an OH concentration proxy, to try to explain the observed ozone production. As this parametrisation allowed us to reproduce well the daily and inter-daily dynamics of the residual O-3 flux, it confirms that the latter actually resulted from O-3 production processes. This ozone production is here highlighted for the first time using O-3 fluxes measurements by the EC method. The chemical reactions possibly involved in O-3 production processes in this maritime pine forest have been discussed and different mechanisms are proposed, based on peroxy radicals chemistry or stress-induced BVOCs.< Réduire