about · methodology + honest disclosures

About

About this project, the data, and how it was built.

What this is

A research-grade visualization of how the climate has warmed at 11,608 specific locations on Earth — cities, towns, and ocean points — between 1850 and 2024, plus an interactive investigation of the data-quality story behind the wide window of current AMOC collapse predictions.

The project is part of a portfolio of small, deeply-researched visualizations published at hermanity.dev, each focused on a single empirical question.

How it was built

  • Data acquisition: Python with xarray, netCDF4, pandas, numpy, scipy — pulled directly from each provider’s URL with no proprietary tooling
  • Rendering: 9,685 SVG stripes (annual resolution, 175×30px each) + 470 monthly-detail SVGs (175×12×60px each) — pure SVG, no canvas
  • Interactive browser: Vanilla JavaScript (no React, no Vue, no framework) — 200 lines total for the atlas + 200 for the AMOC chart
  • Site: Hugo static-site generator with a custom dark theme (the hermanity theme, shared with other family sites)
  • Hosting: GitLab Pages with custom subdomain climate.hermanity.dev

Total compute: ~5 minutes on a Linux VM with 4 cores. Total site size: ~180 MB including all SVG assets.

Honest disclosures

  1. All 11,608 locations are warming in the Berkeley Earth record. The minimum is +0.047 °C/decade (Norwegian coast). The maximum is +0.229 °C/decade (Southern Ocean). The median is +0.085 °C/decade. None of this is surprising.

  2. The RAPID AMOC trend is statistically significant at 26°N (-0.95 Sv/decade, p=0.031). The subpolar OSNAP trend is not significant. Both are consistent with Fox et al. 2026 (SCOTIA).

  3. The Ditlevsen 2023 prediction is not falsified by our RAPID-only analysis — the 20-year record is too short and too noisy to detect the pre-tipping variance rise that the prediction rests on. We describe this honestly rather than overclaiming “the prediction is wrong.”

  4. We do not claim this is the definitive AMOC paper. The 28-paper dossier and our 20-year RAPID analysis are one snapshot of the current literature. The honest assessment is that the community is genuinely divided and the data is genuinely insufficient.

  5. The climate stripes visualization is Ed Hawkins’s contribution (Reading, 2018). We applied it at scale (11,608 locations vs. his original ~150 cities) but the technique is his.

Future work

  • Add annually-resolved coral/sclerosponge reconstructions of subpolar SST to the EWS analysis
  • Add real-time RAPID updates as new monthly data is released (the dataset is updated annually)
  • Add AMOC paleoclimate reconstructions from sortable silt and Pa/Th proxies
  • Cross-validate the per-location trends against the Berkeley Earth Land-Only and Combined products

Contact

Built by Herman. Reach out on Discord or via the hermanity.dev family site.