WRF-Hydro Parameter Calibration Using PSO

This project applies Particle Swarm Optimization (PSO) to calibrate parameters of the WRF-Hydro hydrological model, aiming to improve simulation accuracy and hydrologic response.

📁 Directory Structure

.
├── core/               # Core modules including objective function, optimization process, etc.
│   ├── NSGA2.py
│   ├── ObjFun.py
│   ├── Optimize.py
│   ├── SenAna.py
│   ├── SurrModel.py
│   └── runner/         # Simulation runner and job scheduler
├── jobs/               # Generated simulation configuration files (.yaml)
├── models/             # WRF-Hydro model files
├── params/             # Parameter configuration files
│   ├── calib_params.tbl
│   ├── run_params.yaml
│   └── sens_params.tbl
├── pic/                # Output figures (e.g., sensitivity plots, optimization results)
├── util/               # Utility scripts
│   ├── jobs.py         # Job configuration tools
│   ├── sample.py       # Parameter sampling
│   ├── visual.py       # Visualization tools
├── run_pso.py          # Main PSO calibration script
├── runhydro.sh         # Shell script to launch the workflow

🚀 Getting Started

1. Environment Setup

Activate the Anaconda environment:

source ~/anaconda3/etc/profile.d/conda.sh
conda activate pso

Ensure the following dependencies are installed:

• pyswarms
• SALib
• numpy
• pandas
• xarray
• logging
• seaborn
• pandas
• xesmf # creat precip forcing files
• joblib
• yaml
• Custom modules: core, util

2. Run the Calibration

Shell Script: runhydro.sh

#!/bin/bash

cp -r ../run_source ./run_source
source ~/anaconda3/etc/profile.d/conda.sh
conda activate pso
echo "The environment is activated and script is running."
nohup python3 run_pso.py > ./run_pso.log 2>&1 &
echo "Python script is running in the background. Check the log file for output."
ps aux | grep run_pso.py

This script performs:

• Copying source templates
• Activating conda environment
• Launching run_pso.py and saving logs
• Checking background process

3. PSO Calibration Logic (run_pso.py)

Workflow overview:

1. Define calibration parameters (e.g., BEXP, DKSAT, MannN, etc.)
2. Generate initial samples
3. Run model simulations and manage jobs
4. Read results and calculate objective metrics (KGE + NSE)
5. Iteratively optimize via PSO
6. Save optimal parameters and objective value

Results will be saved in:

• pso_results.txt: readable text format
• pso_results.npz: NumPy format for analysis

⚙️ Parameter Notes

• Parameter configuration located in params/
• obsdir specifies observed streamflow directory (e.g., ./Qobs)
• sim_info sets WRF-Hydro working directory and meta-info
• Objective function used: 0.5 * KGE(-1) + 0.5 * NSE(-1)

📞 Contact

Author: shihx2003
Email: [email protected]

📌 Notes

• This project relies on the PBS job scheduler and a complete configuration of the WRF-Hydro model.
• Recommended to run on HPC or server environment for performance.

Documentation Acknowledgement

Parts of this README documentation were generated with the help of ChatGPT.