SCSDCT: South China Sea Full-coverage Daily 4-km Surface Chlorophyll-a Remote Sensing Reconstruction Dataset from Discrete Cosine Transform 2005-2019

A discrete cosine transform approach was applied to the OCCCI ocean color chlorophyll products, yielding a 15-year full-coverage daily 4-km CHL product over the South China Sea. Against the cross-validation set (5% of original data preserved) and an independent observational dataset collected from 34 cruises, evaluations suggest that DCT‑PLS has outperformed the widely applied classical data-interpolating empirical orthogonal function method. Besides, the DCT‑PLS method is characterized by more efficient computation. For the cross-validation set, the R2 is larger than 0.98, while root-mean-square error is 0.09 mg m-3. For the in-situ observation, root-mean-square error is 0.15 mg m-3. This dataset was successfully applied to study the intra-seasonal variability in the adjacent Luzon Strait (https://ieeexplore.ieee.org/document/9393603), suggesting its stronger capability compared with original gapped OCCCI data.

对OCCCI海洋色叶绿素产品应用离散余弦变换方法，得到覆盖南海的15年完整覆盖每日4公里CHL产品。经交叉验证集（保留原始数据的5%）及从34次航行中收集的独立观测数据集验证，结果表明，DCT-PLS方法在经典数据插值经验正交函数方法中得到广泛应用，且表现更为卓越。此外，DCT-PLS方法具有计算效率更高的特点。对于交叉验证集，R²值大于0.98，均方根误差为0.09 mg m-3。对于现场观测数据，均方根误差为0.15 mg m-3。该数据集成功应用于研究邻近的吕宋海峡的年内变化（https://ieeexplore.ieee.org/document/9393603），相较于原始的OCCCI缺失数据，显示出其更强的能力。

搜集汇总

背景与挑战

背景概述

该数据集是一个2005-2019年南海全覆盖的每日4公里表面叶绿素a遥感重建数据集，采用离散余弦变换方法处理OCCCI海洋颜色产品生成。评估显示该方法具有高精度（交叉验证R2>0.98，RMSE为0.09 mg m-3）和高效计算特点，已成功应用于研究吕宋海峡的季节内变化，优于传统插值方法。