Pick a place. For each calendar day, we find which year holds the heat and cold record — then watch which years win. The tilt toward the present is the climate signal.
Kayla Joachim's research project, featured among Technoclub's technology highlights. For any place on Earth, it asks a simple question of ~75 years of daily weather: for each calendar day, which year holds the heat record, and which holds the cold record? Counting how often each year wins turns raw temperatures into a picture of warming — recent years increasingly own the heat records while old years keep the cold ones. The tilt toward the present is the signal.
This began as the summer-2024 research of Kayla Joachim, advised by Prof. Doug McCollor and conducted in the orbit of Prof. Roland Stull's atmospheric-science group at the University of British Columbia, with coding support from Dr. Dale Joachim. That study analysed 39 stations across western Canada from Environment and Climate Change Canada's historical archive (about 89 years), measuring the trend in yearly counts of extreme events. Its core finding: maximum-temperature extremes skew toward years after 1985, while minimum-temperature extremes skew toward the 1950s and earlier — warmer days and, especially, warmer nights; the precipitation signal was inconclusive.
The method is unchanged; the reach is not. Where the 2024 work downloaded a fixed set of Canadian stations, this version reads a global reanalysis on demand, so you can drop a pin anywhere on Earth and see the same analysis computed live in your browser. Three things are new: any location rather than a station list; near-real-time data to within days of today rather than a one-time download; and an experimental listening mode that renders each record as a note — heat and cold as two voices in contrary motion — so the trend can be heard, not only seen.
Kayla's advisor (Doug McCollor) and mentor (Roland Stull) are highlighted where they appear.
Live backbone — Open-Meteo ERA5 archive. Global reanalysis, daily, 1950–present (to ~5 days ago), any latitude/longitude, no API key, licensed CC BY 4.0. Fetched directly by your browser, so this page needs no server and stores no data. open-meteo.com
Original source — Environment and Climate Change Canada. The 2024 study used ECCC's historical daily station archive for western Canada. climate.weather.gc.ca
Station-truth verification — NOAA GHCN-Daily. 100,000+ stations in 180 countries; the layer to cross-check a certified value at a real gauge. NOAA NCEI
ERA5 is reanalysis — observations blended with a model — chosen because it is spatially complete and consistent over decades, which is what trend and record work needs. It is not a substitute for the certified reading at a single official gauge; for that, use GHCN-Daily. The leap day (Feb 29) is folded into its calendar neighbour. This view detects the statistical fingerprint of warming and is consistent with it, but a record count alone does not by itself attribute cause. Precipitation is shown as an experimental secondary view (wettest-day records), since reanalysis precipitation is less reliable than temperature.
The Map shows the 34 Canadian stations from Kayla's 2024 study on a real basemap; each marker is coloured by whether that decade set more heat records (red) or cold records (blue), and the decade slider animates the tilt over time. Tap any station to run the full live analysis there. The Time-cube stacks the decades vertically — one record-map per layer, oldest at the base — so you can see the whole region flip from blue to red as you rise through time; rotate it and highlight any layer. Both are computed from the embedded 2024 record counts, so they are instant and need no network. (In the 1950s just 1 of these 34 stations was heat-record-dominated; by the 2010s, 30 of 34 were.) The Station view zooms into one place and shows only its extreme events: each week's warmest-ever record (▲, front lane) and coldest-ever record (▼, back lane) placed at the year that holds it, built live from Open-Meteo. Heat records ride high and recent, cold low and old. Both 3D views orbit freely by dragging, and a slider isolates a single slice (decade, or month).