OneScience-Group/Delta-HBV

Delta-HBV

# Model Introduction Delta-HBV is a differentiable, mass-conserving, physics-informed hydrological model with dynamic parameter learning.

Paper: The suitability of differentiable, physics-informed machine learning hydrologic models for ungauged regions and climate change impact assessment
https://doi.org/10.5194/hess-27-2357-2023

Model Description

The model was proposed by Pennsylvania State University and KAUST. It was trained with CAMELS daily forcing, 35 basin attributes, and streamflow. An LSTM drives 16 parallel HBV components and dynamic beta and gamma parameters for ungauged-basin and historical-trend assessment.

Use Cases

Use Case	Description
Ungauged hydrology	Evaluate spatially held-out streamflow.
Physical variables	Predict discharge and evapotranspiration.
Ungauged regions	Test extrapolation to contiguous held-out regions.
Historical trends	Assess annual and high/low-flow trend preservation.
ModelScope/OneCode execution	Validate data, training, inference, metrics, and visualization.
Multi-GPU training	Start multi-process training through torchrun.

Usage Instructions

Use a GPU or DCU when available; CPU supports the default smoke configuration.

hf download OneScience-Group/Delta-HBV --local-dir ./Delta-HBV
cd Delta-HBV
python scripts/fake_data.py

For single-process training, use:

python scripts/train.py

For multi-process training, use:

torchrun --standalone --nproc_per_node=2 scripts/train.py

Run inference and evaluation with:

python scripts/inference.py
python scripts/result.py

Training optimizes streamflow after a 365-day warm-up. Inference produces finite 730-day discharge and ET sequences; evaluation reports NSE, ET RMSE, and trend error.

Trained Weights

No weights are bundled under weight/. The paper does not provide a directly loadable official pretrained checkpoint, so no weight link is listed; the authors' code archive is not presented as model weights.

Citation and License

This repository is an independent engineering reproduction of the public Delta-HBV specifications.

The original paper is licensed under CC BY 4.0; the original paper, official code, CAMELS, and MODIS data remain subject to their respective licenses and terms.