The Xiamen Symposium on Marine Environmental Sciences (XMAS 2027)

Secchi Disk Depth (Z_SD) is a crucial indicator characterizing marine ecological environment, playing a key role in monitoring water clarity and understanding climate-driven processes. Despite recent advances in satellite remote sensing retrieval of Z_SD, existing products remain constrained by deficiencies in reliability, temporal continuity, and spatial completeness in coastal waters. To address these limitations, we propose a coupled reconstruction framework (DCT-XGB) to generate a global, daily, high-accuracy, and seamless Z_SD dataset (65°N-65°S, 4 km resolution) from multi-satellite merged OC-CCI remote sensing reflectance data (1998–2023). Building upon reliable retrieval results (APDm = 28.9%), we developed the DCT-XGB model (APDm = 23.5%) to reconstruct missing regions by integrating traditional statistical learning (Discrete Cosine Transform-Penalized Least Squares, DCT-PLS) and machine learning (Extreme Gradient Boosting, XGBoost). Specifically, it fuses the DCT-PLS-derived Z_SD background field with environmental and temporal drivers (Mixed Layer Depth, Sea Surface Temperature, Month, Day of Year) to produce the final gap-free product. Independent evaluation with in situ observation reveals that this product has substantially higher overall accuracy (APDm = 27.5%) than the existing GlobColour product (APDm = 56.1%). Notably, in low-Z_SD waters (Z_SD≤5m), the new product improves accuracy by 37%. Furthermore, the new product reliably reproduces the time-series characteristics of in situ Z_SD at validation stations, with higher correlation (R² = 0.59±0.15) and lower error (APDm = 16.3%±5.3%) compared to GlobColour (R² = 0.22±0.19, APDm = 36.0%±16.6%). Trend analysis based on the new product reveals a significant contracting trend (–1.9×10⁴ km² yr⁻¹, P < 0.01) in low-Z_SD coastal waters across the global ocean over the past 26 years, a feature that cannot be captured by the GlobColour product due to its limited accuracy.  In summary, the proposed framework and the resulting global daily Z_SD product represent a significant breakthrough in spatial completeness, temporal continuity, and coastal accuracy. By effectively overcoming the spatiotemporal discontinuities of existing satellite records, this work provides high-quality, long-term data essential for marine ecological monitoring and global climate change research.