Grad Analytics: Pandas groupby + Datetime on Climate Data (with GHCN-from-Cloud)

This repository is a self-contained in-class activity showing idiomatic pandas groupby + datetime patterns on climate data, now powered by a cloud dataset (NOAA GHCN Daily, public S3).

What you'll do

1. (Notebook 02) Fetch Illinois GHCN daily data from NOAA's public S3 for 4 stations with ≥30 years of record and write a local Parquet file.
2. (Notebook 01) Load that Parquet (or read via a cloud HTTPS raw URL once you push the repo) and perform groupby+datetime analytics:
   • Monthly means (per station), annual totals & rankings
   • Station-by-month climatology
   • Warm-season aggregations
   • Rolling/windowed stats per station
   • groupby vs resample

Quick start

# conda/mamba
mamba env create -f environment.yml
mamba activate pandas-datetime-climate

# or pip
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

# Launch
jupyter lab

Open notebooks in this order:

• notebooks/02_fetch_ghcn_il_to_parquet.ipynb (creates data/ghcn_il_top4_daily.parquet)
• notebooks/01_groupby_datetime.ipynb (loads that Parquet and runs the analytics)

Cloud-style access

After pushing this repo to GitHub, you can read the Parquet over HTTPS by setting:

CLOUD_PARQUET = "https://raw.githubusercontent.com/USER/REPO/main/data/ghcn_il_top4_daily.parquet"

in 01_groupby_datetime.ipynb.

Repo layout

grad-analytics-pandas-datetime-climate_v2/
├── data/
│   └── ghcn_il_top4_daily.parquet        # created by Notebook 02 (not committed by default)
├── notebooks/
│   ├── 01_groupby_datetime.ipynb          # the analysis exercise
│   └── 02_fetch_ghcn_il_to_parquet.ipynb  # cloud → parquet tutorial
├── src/
│   └── io_helpers.py
├── environment.yml
├── requirements.txt
└── README.md

License

MIT