Dissolved organic carbon (DOC) is a critical component of the oceanic carbon pool, significantly influencing the global carbon cycle and climate change. The sources and transformations of DOC are generally affected by various processes, resulting in pronounced spatiotemporal variations. However, traditional ship-based sampling methods possess considerable limitations, hindering our understanding of DOC distribution and its responses to climate change. Our study aims to integrate global DOC datasets and employ machine learning algorithms, utilizing variables such as latitude, longitude, sampling time, temperature, salinity, depth, and dissolved oxygen, to infer the spatiotemporal distribution of DOC in the North Pacific. We systematically analyze vertical structural, monthly and interannual fluctuations, and their relationships with climate events like El Niño and La Niña. Furthermore, our study incorporates variations in dissolved inorganic carbon (DIC) and particulate organic carbon (POC) to assess the long-term trends of the three marine carbon pools and the contribution of DOC to carbon export. This study provides critical insights into the dynamic characteristics of DOC in the North Pacific and its importance in the carbon cycle, thereby enhancing our understanding of the regional carbon budget.
 

dissolved organic carbon, spatiotemporal variations, machine learning algorithms, carbon pools