 This paper proposes a novel approach to reduce non-stationarity in brain-computer interface, BCI, systems by selecting appropriate channels from the available data. The authors used a Riemannian geometric framework to analyze the data and identify which channels were most relevant to the user's behavior. This was done by comparing the data across sessions and identifying those channels that showed the greatest variation. The authors then developed a selection algorithm that would select the least variable channels, thus reducing the impact of non-stationarity on the BCI system. The results show that the proposed approach outperformed existing methods in terms of accuracy and robustness. This article was authored by Khadija Sadatnajed and Fadyan Latte.