Adaptive Hybrid Algorithm for Retrieving Particulate Organic Carbon in Coastal Waters

Three Python scripts for training and applying an adaptive hybrid algorithm to retrieve particulate organic carbon (POC) in optically complex coastal waters. The code uses KMeans-based water-type classification and cluster-specific models trained with station-level leave-one-station-out cross-validation. The software package is 33.8 KB and was last updated on June 3, 2026.

Use Cases

• Classifying coastal water types based on optical features like bbp442, Apg443, and Cpg443.
• Training cluster-specific POC retrieval models using algorithms such as Random Forest, Support Vector Regression, and XGBoost.
• Applying trained models to new Excel data to predict particulate organic carbon concentrations.
• Developing reproducible workflows for remote sensing of carbon in variable coastal waters.

Strengths

• Includes a complete workflow with three distinct Python scripts for training, validation, and application.
• Exports trained models in reusable formats (.pkl, .joblib) for direct application.
• Implements station-level leave-one-station-out cross-validation, which may help reduce spatial overfitting.

Limitations

• Description metadata is limited; actual data quality and required input feature structure require manual inspection after download.
• Row count and specific dataset size for model training are unknown, which may limit suitability assessment.
• The 33.8 KB size indicates the package contains only code and model files, not the underlying training data.

Provenance

Source

figshare

Collection Method

Software package containing algorithms for POC retrieval, likely developed from research data.

Freshness

Last updated 2026-06-03 09:28:56; freshness should be verified.

Geography

Complex coastal waters (region unspecified).