Data Availability

Data Availability Statement:

The datasets and resources supporting this study are publicly available under the CC-BY 4.0 license via the following repositories:

Primary Data & Figures:

Raw in-situ measurements, satellite-derived CDOM products, and original figures (e.g., spatial maps, calibration plots) are hosted on Figshare.

Reproducibility Resources:

Python scripts for reproducing figures, preprocessing data, and training machine learning models (SVM, MLP, XGB, BRR, KRR).

python scripts documenting the implementation of the Mixture Density Network (MDN) algorithm, including hyperparameter tuning and uncertainty quantification.

Model Outputs & Validation:

Pre-trained MDN model weights and architecture files.

Validation metrics for all algorithms, alongside cross-validation results.

This comprehensive archive ensures full transparency, reproducibility, and reuse potential for coastal CDOM retrieval studies.