Ocean dissolved oxygen (O₂) is an essential climate variable critical for sustaining marine ecosystems. The ongoing decline in oceanic O₂, known as ocean deoxygenation, not only threatens marine life but also perturbs global biogeochemical cycles. However, the accurate quantification of long-term O₂ trends has been persistently impeded by data quality issues in in-situ observations. For the oxygen data from BGC-Argo (2002-present), we utilize high-quality ocean station data (OSD) to correct for residual offsets. By creating a comprehensive matchup dataset of co-located Argo profiles and OSD data, we statistically analyzed the characteristics of these offsets, revealing a highly consistent negative bias in Argo data between 1000 and 2000 m depth. Furthermore, we identified distinct bias characteristics and distribution patterns associated with different Data Assembly Centres (DACs) and sensor models. Based on these findings, we developed and applied targeted correction schemes for each sensor.In addition to the Argo dataset, our research extends to historical OSD data from the pre-2000 era. We will present a newly recalculated historical O₂ time series and demonstrate the impact of these data corrections on the assessment of long-term oxygen changes. Ultimately, this work will help us to have a more accurate understanding of the global ocean oxygen inventory and its response to climate drivers.