Paired 10Be – 26Al Global Catchment Dataset for Single-Nuclide Reliability (v1.0)

This dataset provides both Random Forest (RF) and XGBoost (XGB) classification and mapping scripts. In the mapping module, users can input predictors from their own new dataset to obtain a robustness prediction of whether single-nuclide measurements are representative at the catchment scale. A predicted label of 0 indicates that the model considers the catchment suitable for single-nuclide analysis. We recommend using the RF model as the primary basis for interpretation.Paired_nuclides_basin_features.xlsx is the paired-nuclide training dataset used to build the models. Predictalldrain_clean.xlsx is the dataset used for mapping/predicting across the global single-nuclide sample set; “clean” indicates that zero values and other anomalous entries in the predictors have been removed. The remaining folders contain the output files generated after running the classification and mapping workflows.