In order to improve the purification capacity of an aerobic granular sludge, two aeration strategies were designed and tested. The goal was to optimize the nitrogen removal performances via various oxygen patterns in the aeration phase of the reactor cycle. The main process driving theses patterns is that a difference between oxygen concentrations in the reactor will modify the size of the anoxic inner zone of a granule. The first aeration was based on a succession of oxygen peaks (50%DO) and semi anoxic valleys (5%DO). The second was based on a short pulse of oxygen followed by the hermetical isolation of the reactor and a continuous recirculation of the pulse. The objective was to get an n-removal superior to 82%, results of a previous article, and if possible better than 88%, results of a previous experiment in the lab. The succession of oxygenated peaks strategy worked well and gave good results, as the n-removal was increasing with the sludge bed height and time. However, due to laboratories problem and to the shortness of the time-slots available to experiment, the system could not have the time to stabilize and no constant removal value could be found. Therefore, the average n-removal was equal to 69% for 40 days of experimentation, with some peak over 80%. The highest value observed was 88% of n-removal. The recirculation strategy gave unsatisfactory results, actually unexplained. The first recirculation led to the disappearance of high nitrification capacity, giving an n-removal percentage of 48%. The second attempt was worse, as the oxygen concentration started to behave incoherently and over-oxygenated the system. This behavior still lacks a proper explanation and only hypotheses can actually be proposed. Besides these mains results, the surveillance of the sludge in the reactors allowed the discovery of a bias in the evaluation of the suspended biomass measurement. It was explained by the fact that heavier granules tend not to mix well, and therefore the system could not be considered homogenous. The bias was created because measurements were made at different height between the two reactors. In order to remove it, a linear transformation was developed and tested. The first results seem to be consistent, but due to the lack of experimenting time, the working hypothesis could not be deepened. Another bias that was spotted is the possible influence of the operator during the aeration cycle measurement. Three of four off-trend points in the first experiment originated from a cycle measurement. The data set was too small to draw strong conclusions upon the bias. An extended survey of the cycle measurement and their supposed off-trend consequences in previous and future experiments could clarify the situation.