Repository: A machine learning approach to driver attribution of dissolved organic matter dynamics in two contrasting freshwater systems

Repository containing data and machine learning workflows used to identify environmental drivers controlling fluorescent dissolved organic matter (fDOM) dynamics in freshwater systems: https://github.com/danielmerbet/driver_attribution_fdom/ The repository accompanies the preprint (the paper was accepted, updated link will be added soon): A machine learning approach to driver attribution of dissolved organic matter dynamics in two contrasting freshwater systems https://doi.org/10.5194/egusphere-2025-4049 Overview Dissolved organic matter (DOM) dynamics are influenced by multiple environmental drivers including hydrology, meteorology, and seasonal cycles. This repository provides: Data used in the study Machine learning workflows Feature importance analysis SHAP interpretation of models Scripts to reproduce all results The workflow combines multiple machine learning algorithms: Random Forest XGBoost LightGBM CatBoost Kernel methods k-nearest neighbors These models are used to identify the most influential drivers controlling fDOM variability across two contrasting freshwater systems. Study sites Two study sites were analyzed:Lough Feeagh (Ireland): humic oligotrophic lake with a peatland-dominated catchment and temperate oceanic climateSau Reservoir (Spain): eutrophic reservoir with a human-influenced catchment and Mediterranean climate Repository structure driver_attribution_fdom/│├── README.md│├── 1_hyperparameter_tuning.R│ Hyperparameter optimization for all ML models│├── 2_MLrun_most-influential-features.R│ ML simulations using selected drivers│├── 3_MLrun_reanalysis-julianday.R│ ML simulations using reanalysis meteorology + seasonal predictors│├── 4_extract_importance.R│ Extraction of feature importance across ML models│├── 5_shap_analysis.py│ SHAP analysis for model interpretation│├── feeagh/│ ├── data/│ └── output/│├── sau/│ ├── data/│ └── output/│├── figures/│ Figures used in the manuscript│├── codes_supplementary/│ Additional scripts used in supplementary analyses│└── old_codes/ Archived development scripts