US Extreme Weather and Climate Change Dashboard

This site's headline detection metric for heat waves: ETCCDI/CLIMDEX standard WSDI (Warm Spell Duration Index), run on NOAA's homogenized nClimGrid-Daily national series -- the same metric and homogenized-data standard the IPCC/NCA detection literature itself uses. See note for why, and for a real, disclosed raw-vs-homogenized dataset dependency. Unit: days per year, CONUS daily spatial mean.

ETCCDI/CLIMDEX standard definition (Zhang et al. 2005): annual count of days that are part of a run of 6+ consecutive days with a high temperature above the 90th percentile for that calendar day, against a fixed 1961-1990 base period -- the same definition this project's raw-station WSDI panel below uses, run here on NOAA's homogenized, gridded nClimGrid-Daily series instead of raw per-station data. This is the headline metric on this page for two reasons: it's the ETCCDI-standard metric the detection literature (AR6, NCA5) itself uses, and trend detection specifically calls for homogenized data -- raw station records carry known, per-station inhomogeneities (instrument changes, station moves, time-of-observation changes) a trend test can't distinguish from real climate signal. Real, disclosed dataset dependency: this project's own raw, non-homogenized WSDI panel below shows a small but statistically real decline over its full 1895-2025 record (Mann-Kendall p=0.011), too small relative to natural variability to count as a detected change by this site's own criteria -- a different picture from this panel's clear detected increase (p<0.001) since nClimGrid-Daily's record begins in 1951. An independent, peer-reviewed analysis using similarly raw, unadjusted CONUS daily station data reaches a similar raw-data conclusion: Christy (2026, Theoretical and Applied Climatology) finds hot-extreme metrics show "modest negative trends" across the conterminous US since 1899, with the 1930s Dust Bowl era remaining the most extreme on record. See the Methodology page and the raw-station panels below for the fuller picture -- both the homogenized and raw results are real, computed on the same underlying stations, and the gap between them is itself the finding worth disclosing, not resolved by picking one.

Source: NOAA NCEI nClimGrid-Daily (homogenized)

Computing trend…

National heat wave index, NOAA HCN reference stations -- an unweighted mean across raw, non-homogenized stations (see note). This site's own original, bespoke construction -- see the WSDI (nClimGrid, homogenized) panel above for this page's headline detection metric. Unit: events per station per year.

Reliable trend since 1895 — the HCN reference network doesn't reach enough long-record stations to be nationally meaningful before then. This index is an unweighted mean across raw GHCN-Daily stations, with no homogenization or area-weighting for uneven station density -- see the "Homogenized" TXx/WSDI panels on this page for a NOAA-adjusted, gridded comparison, and the Methodology page's Known Caveats.

Source: NOAA GHCN-Daily, HCN reference stations

Computing trend…

IPCC AR6/ETCCDI standard TXx -- the single hottest daily high temperature observed each year, no percentile threshold involved. Unweighted mean across raw, non-homogenized stations -- see the "Homogenized" TXx panel below for a NOAA-adjusted comparison. Unit: °C (annual max daily high).

The single hottest daily high temperature observed each year at each qualifying station, averaged across stations. No percentile threshold or base period involved -- the simplest standard IPCC AR6/ETCCDI extreme-value metric. This is an unweighted mean across raw GHCN-Daily stations -- no homogenization for instrument/station-move inhomogeneities, and no area-weighting for uneven HCN station density. See the "Homogenized" TXx panel below, which runs the same metric on NOAA's adjusted, gridded nClimGrid-Daily product -- and shows a detected increase (since 1951) where this raw panel shows a detected decrease over its full 1895-2025 record, an even starker raw-vs-homogenized disagreement than WSDI's on this page (opposite detected directions, not just detected vs. not).

Source: NOAA GHCN-Daily, HCN reference stations

Computing trend…

ETCCDI/CLIMDEX standard WSDI -- annual count of days in a run of 6+ consecutive days above a station's fixed-base-period 90th percentile. Unweighted mean across raw, non-homogenized stations -- see the "Homogenized" WSDI panel at the top of this page (this site's headline heat-wave metric) for a NOAA-adjusted comparison. Unit: days per year.

ETCCDI/CLIMDEX standard definition (Zhang et al. 2005): annual count of days that are part of a run of 6+ consecutive days with a high temperature above the station's calendar-day 90th percentile, computed against a fixed base period. This is an unweighted mean across raw GHCN-Daily stations -- no homogenization for instrument/station-move inhomogeneities, and no area-weighting for uneven HCN station density. See the "Homogenized" WSDI panel at the top of this page, which runs the same metric on NOAA's adjusted, gridded nClimGrid-Daily product and is this site's headline detection metric for heat waves.

Source: NOAA GHCN-Daily, HCN reference stations

Computing trend…

ETCCDI/CLIMDEX standard TN90p -- percent of days with an overnight low above a station's fixed-base-period 90th percentile. Unweighted mean across raw, non-homogenized stations (see note). Unit: % of valid days.

ETCCDI/CLIMDEX standard definition: annual percentage of days with an overnight low above the station's calendar-day 90th percentile, computed against a fixed base period. NCA5 flags nighttime-low warming as a distinct, high-confidence signal (less overnight relief), not a redundant restating of TXx. This is an unweighted mean across raw GHCN-Daily stations -- no homogenization for instrument/station-move inhomogeneities, and no area-weighting for uneven HCN station density (unlike TXx/WSDI above, this metric doesn't yet have a homogenized nClimGrid counterpart on this page).

Source: NOAA GHCN-Daily, HCN reference stations

Computing trend…

A homogenized-daily-resolution comparison for the TXx panel above: same definition (annual hottest day), run on NOAA's nClimGrid-Daily national series instead of raw per-station GHCN-Daily data. Spatially averages across the whole CONUS before taking the annual max (unlike the per-station TXx above, which takes each station's max first) -- expect this series to read more moderate in magnitude; the fair comparison is trend direction/detection, not matching values. Unit: °C, CONUS daily spatial mean.

Same math as the per-station TXx/WSDI above (90th percentile, fixed 1961-1990 base period, 5-day window, WSDI's 6-consecutive-day rule) run on a different, homogenized, daily-resolution data source -- NOAA's pre-computed CONUS-wide spatial daily average, not a per-station reconstruction. Disclosed methodological difference: spatially averaging before taking the temporal extreme (this series) mechanically smooths day-to-day extremes relative to averaging each station's extreme afterward (the per-station approach) -- expect this series to read more moderate in magnitude even before any homogenization effect. The fair comparison is trend direction/detection, not matching magnitudes -- and even on direction, the two disagree here: this homogenized series shows a detected increase (since 1951) while the raw per-station TXx panel above shows a detected decrease over its full 1895-2025 record, a real, disclosed raw-vs-homogenized dependency, not an error in either.

Source: NOAA NCEI nClimGrid-Daily (homogenized)

Computing trend…