Development of a marine environmental baseline database with enhanced spatiotemporal completeness: a case study of the Bohai Sea, China

Long-term marine environmental assessment requires baseline data with consistent spatiotemporal coverage; yet, inadequate standardization and integration of monitoring data often hinder oceanographic assessment and management. This study developed a gap-filling framework based on spatiotemporal distribution baseline paradigms and a baseline database for dissolved inorganic nitrogen (DIN) and phosphate (PO 4 -P) in the Bohai Sea (BS), China's only semi-enclosed inland sea. Multi-source raw data from 1982 to 2023 were standardized and assessed using station coverage (SC), granularity of station distribution (GS), and seasonal balance (SB). Datasets classified as mid-high quality or higher were used to identify spatial, seasonal, and interannual distribution paradigms. Incomplete data were matched to suitable paradigms using a back-propagation neural network (BPNN), adjusted with available raw data, gap-filled, and integrated into a database. DIN and PO 4 -P data were highly uneven spatially and temporally, and only 25.17% and 23.13% of the datasets were suitable for paradigm identification. The 0.1° × 0.1° grid scheme provided the best balance between spatial consistency and numerical deviation. The BPNN-based method outperformed reference methods, with mean root mean square errors (RMSEs) of 0.25 and 0.07 μmol/L for DIN and PO 4 -P, respectively. Gap-filling increased spatial and seasonal completeness by 41% and 52%, respectively, and interannual reconstruction achieved an RMSE of 0.032. The reconstructed baseline data provide a consistent basis for describing nutrient patterns and may support eutrophication assessment in the BS. Their application to other environmental parameters or coastal seas requires further testing.