This GRID project output combines USGS river chemistry and watershed information to examine relationships between land cover and rock-weathering fluxes. The release includes derived data tables, watershed predictor objects, and R code linking geochemical inverse modeling with XGBoost, model interpretation, and uncertainty analysis.
The workflow estimates silicate- and carbonate-derived alkalinity fluxes and generates predictions across the contiguous United States. It provides an example of how compiled river observations and environmental context can support AI-enabled research.
Access and contents
View and download the archived data and code · Explore the GitHub code repository
| Release detail | Information |
|---|---|
| Repository creator | Xiying Sun |
| Release date | August 8, 2026 |
| Release DOI | 10.5281/zenodo.21846324 |
| Contents | R scripts and derived CSV/RData files |
| Download | ZIP, approximately 26.8 MB |
| Repository resource type | Software |
| Repository license | CC BY 4.0 |
The archive contains geochemical flux estimates, watershed predictor objects, and merged analysis tables. Its README documents 27 study-specific watershed predictors and code for InveR modeling, nested cross-validation, SHAP interpretation, spatial prediction, and bootstrap uncertainty estimation. Follow the archived documentation for required inputs, local path configuration, R packages, and cluster-based steps.
The archived flux table contains 1,153 unique sites; the merged flux-and-predictor table contains 1,084. These are table-specific counts, not necessarily the sample size of each fitted model or the total number of GRID sites. The release’s derived annual and seasonal summaries should be distinguished from individual observations and from predictions generated by the modeling code.
Reuse and citation
Use the release DOI to identify this archived package. The underlying study workflow is described in the manuscript titled Human modification of land cover reshapes rock weathering fluxes across the contiguous United States.
Data and code: Sun, X. (2026). XiyingSun/weathering-ml-project: Initial release for manuscript (Zenodo) [Software]. Zenodo. https://doi.org/10.5281/zenodo.21846324.
Explore other releases · Methods and interpretation · Data-use guidance