Add hurs and hursmin squeezing script

[Joshdpaul]

This branch adds bias_adjust/squeeze_hurs.py, a post-processing script that clips bias-adjusted hurs (mean daily relative humidity) and hursmin (minimum daily relative humidity) zarr files to the physically valid [0, 100] range.

What it does

• Scans an input directory for all hurs*_adjusted.zarr files
• Counts and logs values below 0 and above 100 with percentages before clipping (Jeremy requested this info for the manuscript)
• Clips out-of-range values to the [0, 100] bounds while preserving NaNs
• Writes squeezed outputs to hurs_squeezed/ and hursmin_squeezed/ subdirectories created within the user supplied input directory
• Saves a hurs_hursmin_squeeze_summary.csv with per-file stats (counts and percentages of clipped values)
• Supports --skip-existing flag to resume interrupted runs

How to test:

• Copy some of the 12km or 4km hurs and hursmin zarr outputs from the directories below into any directory of your choice:

/beegfs/CMIP6/jdpaul3/cmip6_downscaled_llm_fixes_4km_new_vars/sfcWind_hurs_hursmin/adjusted/
/beegfs/CMIP6/jdpaul3/cmip6_downscaled_llm_fixes_12km_new_vars/sfcWind_hurs_hursmin/adjusted/

• Run the new script like so:

srun --partition=analysis --mem=250GB --pty /bin/bash
conda activate cmip6-utils
cd /beegfs/CMIP6/jdpaul3/cmip6-utils/bias_adjust/
python squeeze_hurs.py /path/to/your/copied/outputs/

• When it's finished, inspect the squeezed outputs and the CSVs. Values should match the CSVs attached here:
  hurs_hursmin_squeeze_summary_4km.csv
  hurs_hursmin_squeeze_summary_12km.csv

[Copilot]

Pull request overview

Adds a new post-processing utility to the bias-adjustment workflow to “squeeze” (clip) bias-adjusted relative humidity outputs (hurs, hursmin) into the physically valid [0, 100] range, while generating per-file out-of-range statistics and a combined summary CSV.

Changes:

• Introduces bias_adjust/squeeze_hurs.py to scan for adjusted hurs/hursmin Zarr stores, compute out-of-range counts/percentages, clip values to bounds (NaN-preserving), and write “squeezed” outputs.
• Writes outputs into hurs_squeezed/ and hursmin_squeezed/ subdirectories and emits hurs_hursmin_squeeze_summary.csv.
• Adds --skip-existing to allow resuming runs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[cstephen]

I copied some 4km and 12km hurs and hursmin outputs (for the MIROC6 model), ran the squeeze_hurs.py against them, and verified that the squeeze counts matched the CSVs mentioned in the PR description. I also verified that the squeeze counts matched my earlier counts_outside_thresholds.py outputs from a ~month ago, and then re-used the counts_outside_thresholds.py script to confirm that no values existed below 0 or above 100 in the post-squeeze Zarr outputs. Everything looks good here!