run2-glm-calibration

name	run2_glm_calibration
description	Instructions for modifying GLM parameters using regex and systematically optimizing them within physical limits.

GLM Calibration (Improved)

This skill covers programmatic configuration of GLM and automated parameter tuning.

Modifying Parameters

GLM uses a .nml configuration file. A robust way to edit it without dedicated Fortran namelist parsers is using Python's re module. This ensures parameters with scientific notation or decimals are reliably replaced.

import re
def update_nml(file_path, params):
    with open(file_path, 'r') as f:
        content = f.read()
    for k, v in params.items():
        # \b ensures exact keyword match, avoiding partial matches like 'catchrain'
        content = re.sub(rf"(\b{k}\s*=\s*)[0-9\.eE+-]+", rf"\g<1>{v}", content)
    with open(file_path, 'w') as f:
        f.write(content)

Running the Model

Call the glm executable in the directory containing glm3.nml:

subprocess.run(['glm'], cwd='/path/to/project', stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)

Automated Tuning

To quickly find parameters satisfying specific RMSE constraints without manual trial-and-error, you can employ scipy.optimize.minimize (like Nelder-Mead) over a custom objective function. Provide boundary lists to keep values within physically reasonable calibration ranges.