US Extreme Weather and Climate Change Dashboard
Detection & Attribution: IPCC AR6 vs. This Site
IPCC AR6 (2021) is the most current, most rigorously reviewed formal climate assessment with a dedicated extreme-events chapter (WG1 Chapter 11, supplemented by Chapter 12's regional detail) -- see the Methodology page for the underlying IPCC detection/attribution definitions behind this whole site. The table below summarizes AR6's North-America-focused finding for each hazard alongside this site's reliable-window verdicts, computed live from the exact same data as each hazard's phenomena page.
Read the "Consistency" column as a qualitative judgment call rather than a formal statistical test. AR6's findings are global or continental-regional syntheses of many studies and multiple lines of evidence; this site's findings are single-country, mostly-national-scale metrics. Where this site tracks several alternative metrics for one hazard, this page shows every metric's verdict, including the ones that disagree with AR6 -- a real mismatch in this page reports a real mismatch in scope or metric definition as "not directly comparable" instead of smoothing it away.
What the IPCC means by detection, by emergence and by time of emergence, where the two come apart when read backwards, AR6 Chapter 12's Table 12.12 reproduced in full, and how long this record would need to detect the warming AR6 projects for US heat extremes.
Read backwards, detection and time of emergence separate in four places.AR6 puts emergence at the date a signal first exceeds a natural-variability reference, so a date in the past means the signal exceeds that reference today. Detection and emergence share that much, and four differences in AR6's definitions separate them. AR6 defines a time of emergence around "a specific anthropogenic signal" and defines detection "without providing a reason for that change", so a past time of emergence names human influence where a verdict on this page names none. The two apply different statistics: a likelihood test against a signal-to-noise ratio. AR6 Chapter 12 measures from a pre-industrial baseline no instrument recorded, and the verdicts here rest on the observational record. Detection returns a verdict over a window; emergence returns a date. The answers also disagree in both directions: a long, quiet record can clear the significance bar on a change smaller than the year-to-year variation, and a shift measured from a pre-industrial baseline can have emerged while the trend inside the observed window misses that bar. The full treatment sits on the Detection and Time of Emergence page.
| Hazard | IPCC AR6 finding | This site's findings | Consistency |
|---|---|---|---|
| Heat Waves | "It is virtually certain that hot extremes (including heatwaves) have become more frequent and more intense across most land regions since the 1950s." |
| Consistent This site's headline heat-wave metric -- WSDI on NOAA's homogenized nClimGrid-Daily record -- shows a clean detected increase since 1951 (p=0.002), and TXx on the same homogenized record agrees (p=0.022). nClimGrid-Daily's record happens to start in 1951, matching AR6's stated 'since the 1950s' window almost exactly, so this row lands as a direct match: same window, same conclusion. Warm nights (TN90p, raw stations) move the same direction but sit just above the bar (p=0.102), so this site renders no detected change there. This site's raw, non-homogenized station panels are more mixed -- the bespoke heat-wave index and raw WSDI don't clear this site's detected-change bar over their full 1895-2025 record, and raw TXx shows a detected decrease, because the 1930s Dust Bowl anchors an unusually hot early benchmark for the single hottest day each year. That raw/homogenized gap is a genuine, disclosed dataset dependency, not something swept under the 'Consistent' label -- an independent, peer-reviewed analysis using similarly raw, unadjusted station data (Christy 2026, Theoretical and Applied Climatology) reaches a comparable raw-data conclusion, finding modest declines in hot extremes across the same conterminous US since 1899. Read together: on the homogenized data this site treats as authoritative for detection, and over the same window AR6 itself cites, the two agree; on raw station data, they don't -- both real findings on the same underlying stations, and not a contradiction once the data source becomes explicit. |
| Cold Extremes | "...cold extremes (including cold waves) have become less frequent and less severe [across most land regions since the 1950s], with high confidence that human-induced climate change is the main driver of these changes." |
| Partially consistent On the raw US Historical Climatology Network stations, 2 of 4 cold metrics match AR6 directly: the cold spell duration index and cold nights (TN10p) both show a detected decrease. Frost days show no detected change (p=0.408), and the annual coldest night (TNn) shows a detected INCREASE -- that is, coldest nights getting colder, the opposite of AR6's direction. The homogenized nClimGrid TNn shows no detected change (p=0.264) and does not reproduce the raw increase, the same raw-versus-homogenized split this site discloses for heat waves, and the reason this row reads partially consistent, not consistent. AR6's statement is explicitly about most land regions globally rather than the contiguous US specifically; the companion global dashboard, which tests the same four metrics on a global station network, finds a detected decrease in the two duration-and-frequency metrics and no detected change in the other two -- closer to AR6, and at the scale AR6 addresses. |
| Severe Convective Storms | "There is low confidence in past trends in characteristics of severe convective storms, such as hail and severe winds, beyond an increase in precipitation rates." |
| Consistent Every metric on this page -- tornado count, tornado days, outbreak size, hail (excluded from trend assessment by design, matching AR6's data-insufficiency framing), and three independent CAPE-based instability proxies from reanalysis -- shows no detected change. Matching AR6's 'low confidence, data insufficient' position across metrics built from entirely different data sources (human storm reports vs. atmospheric reanalysis) offers stronger confirmation than any single metric alone. |
| Flooding | "In North America, peak flow has increased in the northeast US and decreased in the southwest US." "Confidence about peak flow trends over past decades on the global scale is low, but there are regions experiencing increases, including parts of Asia, southern South America, the northeast USA, northwestern Europe, and the Amazon, and regions experiencing decreases, including parts of the Mediterranean, Australia, Africa, and the southwestern USA." |
| Consistent This site's own river-flow metrics -- the headline high-flow index (a single national CONUS-wide index) and annual peak flow (AMS) -- show no detected change over their reliable windows. AR6's US-specific finding explains why a national index would show no clean signal in either direction: peak flow has moved in opposite directions within the country, increasing in the northeast and decreasing in the southwest -- exactly the kind of regionally offsetting pattern that would wash out in a single national index without contradicting AR6's finding. The separate heavy-precipitation-extremes metric on this page (Rx1day/Rx5day) sits just under the significance bar (p=0.086) but does not clear this site's magnitude-vs-variability check, so it too renders as no detected change. AR6 itself draws the distinction that would matter if it did: AR6 assigns meaningfully higher confidence to precipitation-extreme detection than to river-flood detection specifically, because dams, diversions, land-use change, and water management confound the streamflow signal in a way they don't for a single storm's rainfall total. An undetected national river-flow trend, when AR6's regional evidence shows opposing US trends that would cancel out nationally, matches the pattern AR6's literature would predict, and does not contradict it. |
| Drought | "Human-induced climate change has contributed to increases in agricultural and ecological droughts in some regions due to increased land evapotranspiration (medium confidence)." On the other drought types this page also tracks: "Few AR6 regions show observed increases in meteorological drought, mostly in Africa and South America; a few others show a decrease (WSB, ESB, NAU, CAU, NEU, CNA: medium confidence)" -- Central North America (CNA), which covers a large share of the continental US, is among the few regions with an observed decrease. On hydrological drought: "The more limited availability of datasets makes it more difficult to assess historical trends in hydrological drought at regional scale. Increasing (MED: high confidence; WAF, EAS, SAU: medium confidence) and decreasing (NEU, SES: medium confidence) trends in hydrological droughts have only been observed in a few regions" -- none of them in North America. |
| Not directly comparable This page now tracks five measures, one for each of AR6's three drought types plus a precipitation-only index and the operational composite: soil moisture (added September 2026, a direct agricultural measure -- the quantity AR6's agricultural/ecological finding actually addresses, where this page previously had only the USDM composite standing in for it), SPEI (a precipitation-evapotranspiration index, the same construction AR6 uses for meteorological drought), SPI (added 15 September 2026, the same index with the evaporative-demand term set aside), the low-flow index (an explicitly hydrological, runoff-based signal), and the USDM D2+ composite itself, whose 26-year record is under this site's 30-year minimum and so renders no verdict. None of the four with a long enough record shows a detected change nationally: SPI p=0.292 over 131 years, SPEI p=0.238 over 131, soil moisture p=0.863 over 78, low flow p=0.121 over 69. Adding a direct agricultural measure makes the agricultural row a real test, not a proxy one, and it comes back negative at the national scale. <strong>SPI closes the one gap a reader could press on.</strong> PDSI, SPEI and the CPC soil-moisture model each depend on a potential-evapotranspiration formula, and Xu et al. (2026, Commun. Earth Environ. 7:726) report that formulas without a land-atmosphere coupling constraint inflate drying trends at least sixfold. SPI uses no such formula, so the national no-detected-change result no longer rests on any of them. The scale mismatch remains the reason this row stays 'not directly comparable': AR6's agricultural and hydrological findings are regional (Western North America; and no North American region at all for hydrological drought) while this site presents national indices, which would dilute a real regional signal instead of confirming or refuting it. For meteorological drought specifically, AR6's Central North America finding is a genuine, if partial, match: a reported decrease, not an increase, aligns directionally with SPEI showing no detected increase here. |
| Winter Storms | "There is low confidence in past changes of maximum wind speeds and other measures of dynamical intensity of extratropical cyclones." Separately: "Mid-latitude storm tracks have likely shifted poleward in both hemispheres since the 1980s, with marked seasonality in trends (medium confidence)." |
| Not directly comparable Fritzen, Lang & Gensini (2021) -- an independent, peer-reviewed NARR-based climatology -- found no significant East Coast bomb-cyclone trend. RSI's storm catalog is intensity-pre-filtered (64% of its tracks are bomb cyclones, versus ~7% in Fritzen et al.'s unfiltered data), which invalidates a raw count from it as a frequency measure. The peak-wind metric on this page (same pre-filtered catalog, but measuring severity rather than frequency) still shows a detected increase; minimum pressure and the RSI storm count show no detected change. Without a valid frequency metric, this row falls short of a clean test of AR6's intensity-trend finding than as confirming or exceeding it. |
| Tropical Cyclones | "A subset of the best-track data corresponding to hurricanes that have directly impacted the USA since 1900 is considered to be reliable, and shows no trend in the frequency of USA landfall events" (Knutson et al. 2019, cited by AR6) -- a direct, US-specific finding, not only a global proportion-of-intense-storms statement. |
| Consistent The US-landfalling hurricane count on this site shows no detected change (p=0.335) over the same since-1900 window AR6's cited reliable subset uses -- a direct, clean match. ACE, rapid intensification, and peak rainfall also show no detected change, none of which AR6's landfall-frequency statement specifically addresses but none of which contradict it either. |
| Wildfire | "Fire weather conditions (compound hot, dry and windy events) have become more probable in some regions (medium confidence)." AR6 separately cites Abatzoglou & Williams (2016), which attributes approximately 49% of the cumulative 1984-2015 burned-area increase in the western United States to anthropogenic climate change -- one of the report's strongest, most quantified attribution statements for any hazard on this site. |
| Consistent MTBS burned area and both fire-weather (VPD) metrics from an independently-sourced reanalysis dataset show a detected increase -- the most consistently confirmed finding on this site across independently-sourced metrics, directly matching AR6's characterization of Western US wildfire as one of its strongest findings. MTBS large-fire count sits exactly on the 10% bar (p=0.100) and so renders as no detected change; NIFC burned area and season length show none either, both drawn from a shorter record than MTBS. This site does not attempt AR6's ~49% attribution fraction itself -- that requires the model-based attribution step this site's detection-only methodology doesn't undertake. |