Loading the climate atlas…
Behind the comparisonsSources, coverage & methodology

Four fixed periods

The historical reference period is 1956–1985. Recent climate is 2016–2025, a ten-year average, not an official 30-year climate normal. Projections cover 2041–2060 and 2081–2100. Recent weather offers individual months, seasons, years and year-to-date comparisons from 2010 onward.

Like for like

The same calendar months are compared on both sides. Winter is labeled by its ending year: winter 2026 is December 2025–February 2026. Annual means give equal weight to the twelve monthly climatologies. YTD always uses January through the latest completed month in this snapshot, including when another year is selected.

Temperature analogs

These matches compare monthly average daily highs, lows and their mean. They do not establish equivalence in precipitation, snow, humidity, extremes, ecology or every other aspect of climate.

Ranking the matches

For each selected month, the score weights squared mean-temperature differences by 50%, high differences by 25%, and low differences by 25%, then takes the square root of the average. Smaller scores indicate closer matches. The same city and grouped airport/downtown records are excluded; ties use station IDs.

All different-city stations that meet the coverage requirements are ranked by temperature similarity, whether their mean is warmer, cooler or equal to the reference. This applies to historical, recent and projected comparisons. A nearest match can still be a poor match; the score is always shown.

Data coverage

Each selected historical calendar month needs at least 18 usable years out of 30. A usable historical or individual-year month needs at least ten daily highs and ten daily lows. Recent 2016–2025 profiles require nine years per month, with at most three missing daily observations per variable. High and low observations need not fall on the same dates. Partial-month means can be unrepresentative. Missing months are never silently skipped.

Every year has equal weight within each historical calendar month. Individual weather observations use the separately retrieved monthly archive; small differences from the stricter recent-climate dataset are possible. Insufficient coverage is displayed explicitly.

Future estimates

Seven CMIP6 models provide monthly changes in daily highs and lows, interpolated at the station coordinates and added to observed 2016–2025 values. SSP2-4.5, SSP3-7.0 and SSP5-8.5 each use their own simulated 2016–2025 baseline and actual future output. They are alternative scenarios, without assigned probabilities.

Models: ACCESS-CM2, FGOALS-g3, INM-CM4-8, IPSL-CM6A-LR, MIROC6, MPI-ESM1-2-LR and MRI-ESM2-0. One simulation from each model is used for each scenario, with all seven models weighted equally. This common subset is not the entire CMIP6 ensemble. All ten model baseline years and all twenty future years are required, along with nine usable observed baseline years for each selected month.

The displayed model range is the minimum and maximum of seven individual estimates, not a confidence interval. Regional grids do not resolve station moves, local topography or future urban development. Natural variability and model selection can change the nearest analog.

Source: Pangeo / ESGF CMIP6 archive ↗ · Model credits and licenses ↗

Station histories matter

The catalog contains 413 U.S. ThreadEx city series, plus Death Valley and Toronto (City). It excludes U.S. territories, Baltimore Downtown and Philadelphia Center City. ThreadEx joins successive stations serving a city. Site moves, instruments and urban development can influence the unadjusted observations. These descriptive comparisons do not isolate the cause of temperature changes.

Toronto follows ECCC’s city thread, separate from its airports, and includes projections. Death Valley (Furnace Creek) is NOAA COOP USC00042319 and has no model projections in this dataset. Map positions represent city areas rather than historical instrument sites.

Observations: ThreadEx ↗ · ACIS ↗ · ECCC ↗. Map: U.S. Census via us-atlas ↗.