scorvec.com

Source of scorvec.com, a weather and climate diagnostics site: real-time products for the mid-latitudes, the tropics and the stratosphere, sub-seasonal and seasonal outlooks, and the studies behind them. Every product is computed from open data by the pipelines in this repository and published as a static site on GitHub Pages.

What is on the site

Theme	Products (examples)
Mid-latitude weather	short-range model viewer (HRRR, RRFS, RDPS), South America model viewer (IFS control, GDPS), dynamic tropopause, North Pacific jet, atmospheric rivers, wave-activity flux
Tropics	ENSO monitor and forecasts (ERSSTv6, official ONI/RONI), MJO, equatorial waves and Kelvin waves, Walker circulation by basin, tropical-cyclone shear and environment
Stratosphere	polar vortex and E–P flux (AIFS-ENS, IFS-ENS, GEFS), 100 hPa heat flux, QBO, Brewer–Dobson circulation, sudden-warming history
Weeks to seasons	GEPS and GEFS days 1–35, SEAS5 and the C3S multi-system seasonal maps, NOAA SFS, regimes and teleconnections, Claude's month-ahead forecast (ERA5 statistics + GEFS)
Studies and verification	AIFS verification, angular-momentum budgets of AI and physics models, El Niño and angular momentum
Tools	searchable plot catalogue, sounding explorer, five-minute airport observations, US climate trends, US credit conditions

Methods, data sources and citations are documented on each page; topics/ holds the explainers.

How it is built

open data ──► GitHub Actions workflows ──► scripts/<product>/ (Python) ──► figures + JSON ──► GitHub Pages
                     (cron + dispatch)                                       │
                                                                             └► animation frames ──► object store (CDN)

• Workflows (.github/workflows/): one per product family, on cron schedules, most with a concurrency group so a product never builds twice at once. Each run fetches its inputs, computes the diagnostics, and commits the outputs with a snapshot-and-resync step (outputs are copied aside, the checkout is reset to the current main, and the outputs are restored before committing), so concurrent runs never overwrite each other's results.
• Pipelines (scripts/<product>/): Python 3.11/3.12 with xarray, numpy/scipy, matplotlib/cartopy and eccodes. Shared numerical methods come from quantmet, pinned to a release tag.
• Animation frames go to an S3-compatible object store served from frames.scorvec.com (scripts/lib/frames_store.py); large state and archives live on dedicated data branches (frames, verify-data, skewt-data, skewt-archive, …) rather than on main.
• Pages: hand-written HTML with shared chrome stamped in by scripts/site/apply_chrome.py (menus, header and footer from one product table). The multi-product "stage" pages are generated from partials in scripts/sst/pages/ by scripts/sst/enso_site.py.

Data

Inputs are open datasets. Real-time inputs are fetched by the workflows; static reference climatologies are built once from the same archives. Sources include ECMWF open data (IFS, AIFS; Google Cloud mirror), NOAA (GFS/GEFS/HRRR/RRFS on AWS, CPC, PSL, NCEI), ECCC (GDPS/GEPS/RDPS), Copernicus C3S/CDS (ERA5, SEAS5 and the C3S seasonal systems), NASA (MERRA-2, GEOS FP, IMERG) and others. Each page credits its sources and their licences.

Statistical conventions

Anomalies are against 1991–2020 unless a page says otherwise. Fields and composites are shown only where they pass significance testing (Benjamini–Hochberg false-discovery-rate control for maps); results that do not are not published.

Repository layout

*.html               site pages
assets/              site JS/CSS, generated figures and data (written by the workflows)
scripts/<product>/   the pipeline of each product (sst, mjo, strat, geps, gefs, sfs, models, skewt, …)
scripts/lib/         shared helpers (frame store, locking, publishing)
scripts/site/        site chrome, catalogue and topic-page generators
scripts/watch/       upstream-change watches (GFS v17 upgrade notices; opens an issue when something is new)
topics/  studies/    explainers and study pages
skewt/   qbo/        self-contained apps
.github/workflows/   the schedules

Running a pipeline

Each product's workflow is the reference for how it runs: it lists the Python version, the requirements file (scripts/*/requirements*.txt or inline pip install), the environment variables and the exact commands. To run one outside Actions, install the same requirements and run the same commands from the repository root.

Security

Report security problems privately; see SECURITY.md.

License

Code: MIT. Site text and figures © Shawn Corvec unless noted; third-party data remain under their providers' terms, as credited on each page.