Estimated CO₂e methodology

The footer line Estimated CO₂e for this page is a build-time estimate for the URL you are viewing — not a site-wide average, and not a measurement of any individual visitor’s actual emissions.

What the number means

• Model: Sustainable Web Design Model version 4, via @tgwf/co2.
• Visit type: First visit (full transfer). The figure assumes a cold cache so CSS, JavaScript, fonts, and images are downloaded. A repeat visit with a warm cache would usually transfer less.
• Hosting: Calculated with green hosting enabled, reflecting this site’s deployment on Cloudflare.
• Unit: Grams of CO₂ equivalent (CO₂e) attributed to the measured bytes for that page view under the model above.

How it is measured

1. A production-like build of the site is served locally with Wrangler (wrangler pages dev _site), so Brotli (and the Pages Function) match edge behaviour as closely as practical.
2. Playwright (Chromium) loads the page once with an empty cache (load, then a short networkidle wait) and records encoded network transfer sizes for the document, stylesheets, scripts, fonts, images, and other responses that actually load. Deferred assets that start after that window may be under-counted.
3. Those bytes are passed to CO2.js (perByte on SWD v4) to produce the grams shown in the footer.
4. Results are stored in committed Eleventy data (_data/co2eAudit.json) and rendered at build time. Normal deploys do not launch a browser; they only read the last committed audit.

When numbers change

Historic posts rarely change. The opt-in script npm run measure:co2e skips URLs whose source content fingerprint is unchanged and reuses the cached row. Run it after publishing a new post (or with --force / --url=/path/ when you want a refresh). Then rebuild or redeploy so the footer picks up the updated data.

Limits

• The estimate follows a lab methodology (SWD v4 + measured transfer). Real-world emissions vary with device, network path, grid intensity, caching, and what the visitor does next.
• Shared asset churn (for example fingerprinted CSS or JS) does not by itself invalidate historic rows until the next measure run.
• Pages without a cached row omit the footer line entirely rather than inventing a number.