Skip to content

About

Large Sample Emulator (LSE) based calibration for conceptual models

Resources

Stars

0 stars

Watchers

2 watching

Forks

Latest commit

 

History

5 Commits

Folders and files

Repository files navigation

LSE-conMod

LSE-conMod uses large-sample emulators (LSEs) to calibrate and regionalize parameters for conceptual hydrological models. The companion study evaluates Global Optimization And Localized Selection (GOALS) with the HBV model across 531 CAMELS basins.

Code and data for the paper

Use the versioned Zenodo release, v1.0.0 for the study workflows. Download LSE_conMod_public_release.zip, extract it, and follow its README.md and NOTEBOOK_GUIDE.md. The archive includes five-fold spatial cross-validation, generation and local selection of regionalized parameter candidates, the single-site emulator (SSE) comparison, selected analysis notebooks, and the processed camels_data/ inputs.

This GitHub repository contains the earlier HBV calibration experiments. Its scripts and settings differ from the archived implementation, and its data/ directory contains parameter bounds, basin metadata, and preparation notebooks. Use the code and configurations from the same snapshot when running the study workflows.

GitHub repository layout

Directory Contents
HBV/ HBV model implementations
LSEexp/ LSE calibration scripts, experiment TOML files, and PBS examples
SSEexp/ SSE calibration and optimization scripts and PBS examples
auxiliary/ Data loading, model execution, evaluation, and optimization helpers
data/ HBV parameter bounds, CAMELS basin metadata, and preparation notebooks

Running the GitHub experiments

Install the direct Python dependencies with python -m pip install -r requirements.txt. The workflows also require the MO-ASMO modules sampling, gp, and NSGA2 on the Python import path. The SSE workflow additionally imports mo_evaluation from the research support code included in the Zenodo archive. See NOTICE for provenance.

The TOML files and PBS scripts retain the original NCAR HPC settings. Point inpath_data to the extracted camels_data/ directory and check the parameter bounds, output, external support module, and PBS settings for your environment. Run the scripts from their own experiment directories so the relative imports resolve.

For example, from the repository root, initialize the historical HBV-KM LSE experiment with the following command:

cd LSEexp
python main_LSEworkflow_HBVKM.py config_LSEexp7.toml 0 1

The last two arguments specify the iteration and CPU count. Subsequent iterations use results from earlier iterations. The historical SSEexp1 configuration reuses the LSE iteration-zero simulations via external_iter0_path.

Citation

Tang, G., Wood, A. W., Song, Y., Farahani, M. A., & Shen, C. (in press). Calibration, Regionalization, and Locally-Informed Selection of Hydrological Model Parameters Using AI-Based Large-Sample Emulators. Water Resources Research.

For the archived code and processed inputs, cite Tang, G., & Wood, A. (2026). LSE-conMod: Code and Data for Hydrological Parameter Calibration and Regionalization (v1.0.0). Zenodo. 10.5281/zenodo.23873715.

When using the CAMELS data, also cite the time series and basin attributes. Machine-readable citation metadata are in CITATION.cff.

Licensing and attribution

See LICENSE for the proposed license for original project software and NOTICE for the scope and third-party material.

Acknowledgement

We thank Rhys Pulling for his contribution to integrating the CFE model into the LSE-conMod package.

About

Large Sample Emulator (LSE) based calibration for conceptual models

Resources

Stars

0 stars

Watchers

2 watching

Forks

Releases

Packages

Used by

Contributors

Languages