Observation-constrained transfer learning on global runoff projections
Year: 2026
Type: article
Abstract: Observation-constrained transfer learning on global runoff projections This repository contains the code, processed data, trained model examples, and representative input data used for the analyses presented in the manuscript. Folders Fig_1–Fig_8: Contain the processed data and scripts used to generate the corresponding figures in the manuscript. Folder Supplementary_Figures: Contains the processed data and scripts used to generate Supplementary Figures S1–S12. Folder model: Contains scripts for the transfer learning models (TLMs), including: Pre_Training.py: Pre-trains the TLMs using CMIP6 Earth System Model data.Fine_tuning_ESMs.py: Fine-tunes the TLMs using a perfect-model evaluation framework.Fine_tuning_observations.py: Fine-tunes the TLMs using ERA5 observational runoff data. Folder attribution: Contains scripts and processed outputs for controlled attribution and intermodel regression analyses. Folder example_data: Provides representative CMIP6 and ERA5 data, trained model examples, prediction outputs, and processed inputs used to demonstrate the pre-training and fine-tuning workflows. Detailed information on the repository structure and reproduction procedures is provided in README.md.
Authors: Lu, Ziyang, Zhang, Quan
Institution: Hebei GEO University
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: green
OpenAlex ID: W7210285142
FWCI:
Cites:
Cited by:
Related to: