Global Extreme Weather and Climate Change Dashboard

A project by Roger Pielke Jr., author of The Honest Broker · Data last updated September 20, 2026

Check out the US Dashboard
What's New

What's New

Everything added to this dashboard since it launched on August 19, 2026, newest first. Routine data refreshes and the daily Earth image run on a schedule, so this page skips them.

  1. New variable
    • The site now counts tropical cyclone landfalls globally, and the answer depends sharply on where the record starts. Global Landfalls plots landfalls per season at hurricane strength or above across all basins, detected from each storm's 6-hourly fix positions by the Weinkle, Maue and Pielke Jr. (2012) method. Over 1980–2025 the headline Cat 1+ level shows no detected change (p=0.170), while Cat 3+, Cat 4+ and Cat 5 landfalls each show a detected increase (p=0.036, 0.043, 0.021). Start the window in 1990 instead and every one of those four levels falls to no detected change.
    • Two checks point at the record rather than at the storms. The share of hurricanes that reach land at all climbs +3.40 percentage points per decade over 1980–2025 (p=0.013) — a fixed population of storms has no reason to find coastlines more often, so improving best-track position accuracy fits that better than a change in the weather. And the NATL+WPAC record reaching back to 1945, selectable on the same chart, detects nothing at any intensity over its full span. The Methodology page sets out both checks with the full start-date table.
    • An independent compilation reproduces all of it. Ryan Maue's global landfall record at climatlas.com, built from a different source chain and a different landfall convention, agrees with this site at r=0.93 on major landfalls and returns the same verdicts over both windows. The site now fetches his table on every pipeline run, so that comparison re-runs instead of sitting here as a dated claim. His aggregated best-track feed also supplements this site's current-season figures whenever IBTrACS stalls on a live storm, and his power-dissipation work sets a limit on every peak-wind metric here: storm size carries more of the variance than intensity, and no variable on this site measures it.
  2. New series
    • Observed precipitation now drives the meteorological drought index alongside reanalysis, and the two records give different answers. Selecting a GPCP level in the Drought chart runs the identical index — same 5° blocks, land mask, 60°S–90°N scope, three-month window, percentile ladder and area weights — on GPCP v2.3, which merges satellite retrievals onto the GPCC gauge analysis, instead of on ERA5. Over the full record the ERA5 series gives a detected increase, the GPCP series no detected change.
    • The disagreement has a geography: it opens where the rain gauges thin out. Across Europe, North America and Oceania the two records agree, and neither detects a rise. Africa and South America supply 85% of the ERA5 increase, and only South America's appears in the observations. GPCC's gauge-only analysis, which covers 1979–2019, returns a detected decrease over that window where ERA5 returns a detected increase. ERA5's global land precipitation also steps down about 5% at 2000–2001, a step absent from both observational records. This site publishes both series rather than choosing between them, and the Methodology page now says to read the meteorological drought trend as a property of ERA5 until an observational record corroborates it.
    • The three drought types share a parent. ERA5 precipitation forces the GloFAS river-discharge model, and the soil moisture on the Drought page comes from the ERA5 land surface, so all three indices descend from one precipitation field rather than corroborating one another. The Drought page and the Methodology page both now state that dependence.
    • Flooding and hydrological drought now stand at July and June. Both read GloFAS river discharge, which had stopped updating in the spring and left Flooding at April and hydrological drought at March. Both now sit within the one-to-two-month publication lag of the GloFAS product, and the added months moved no verdict.
  3. Measurement
    • Measurement now settles how much the choice of evaporative-demand formula matters, and one published figure does not survive it. This audit built a full FAO-56 Penman-Monteith series and the energy-only Milly & Dunne formulation from ERA5, then ran each through this index over the 571 months all three share. Swapping only that term, Penman-Monteith returns 0.91 times the AED-inclusive period change and the energy-only formulation 0.47 times. Every formulation still detects an increase at every severity level, so the choice moves the size of the trend and not the verdict.
    • Xu et al.'s sixfold gap does not appear on this index. The Methodology page has quoted their finding that Penman-type evaporative-demand formulations inflate drying trends at least sixfold over energy-constrained ones, and drought-area trends 7.8-fold, since 15 September. Here the ratio comes to 1.95, holding between 1.84 and 1.95 across severity levels. Their figures come from a different index over a different window on gauge-based precipitation, so this measures the gap on the construction used here rather than contradicting theirs — and the page now distinguishes the two. The "upper bound" label on the Hargreaves series holds, and the margin runs small: 11% over Penman-Monteith at D0, falling to 1% at D4.
    • The zero floor on the Hargreaves estimate touches 7.6% of land cell-months within the 60°S–90°N scope, and it moved the AED-inclusive period change by about 0.3%. The Methodology page had described this pipeline as applying the equation without a floor, which stopped holding on 15 September.
  4. Correction
    • Every index on the 5° grid now weights its cells by area. The Drought page's three indices averaged their grid cells equally until today, and that average tracks cells rather than land: a 5° cell at 60° of latitude covers about half the area of an equatorial one. Of the 867 land cells, 263 sat south of 60°S — 30.3% of the cell count against 9.8% of the land area — so Antarctica took 3.08 times the weight its area warrants, Africa 0.55 times, South America 0.57, Asia 0.66 and Oceania 0.61.
    • Three verdicts changed, all in the same direction. Global meteorological drought (the precipitation-only series) moved from no detected change to a detected increase, at every severity level rather than only D2+. Global root-zone soil moisture did the same, and so did North America's. The unweighted count held these trends below the bar rather than lifting others above it.
    • "Global" now means 60°S–90°N, so the global figures and the six continental ones describe the same land. Antarctica's cells entered every global figure and no continental one, which put the homepage's global drought tile beyond the reach of any continent page.
    • A zero floor now bounds the Hargreaves evaporative-demand estimate. Its (Tmean + 17.8) term turns negative below −17.8 °C, and 21.9% of land cell-months had returned a negative demand, which pushed the water balance above precipitation and let warming move a very cold cell toward drought by shrinking a negative term.
    • The Gebrechorkos et al. 2025 citation now names its retraction. Nature retracted "Warming accelerates global drought severity" on 2 September 2026 (doi:10.1038/s41586-026-11027-z). The Methodology page and the Library now name the retraction and its four listed problems, and both have dropped the word "independent". This audit tested two of those problems here: grid-cell averages taken without area weighting, and an omitted arid-region mask.
    • Smaller corrections alongside: the chart unit now reads "% of land area" rather than "% of land cells", the superseded +0.93/+3.35 period-change figures become +3.01/+6.04, and the Benjamini-Hochberg table names which of a variable's series each row tested (drought offers ten).
    • A new paper sharpens the evaporative-demand caveat, and it cuts against this site. Xu et al. (2026, Communications Earth & Environment 7:726) report that Penman-type evaporative-demand formulations inflate drying trends at least sixfold over energy-constrained ones, and that the share of the drought trend attributable to evaporative demand falls from 47.5% to 25% under the constraint. The Hargreaves method behind the AED-inclusive series here applies no such constraint, so the Methodology page now says to read that series as an upper bound, and the Library lists the paper. Xu and coauthors write against the Gebrechorkos study retracted on 2 September 2026.
    • This site no longer reproduces the source paper behind its AED-inclusive series, and the Methodology page now says so. Vicente-Serrano et al. (2022) report no substantial global change in meteorological drought over at least 120 years, and a statistically significant decline in 12-month-SPI drought area over 1950–2020. The precipitation-only series here now shows a detected global increase — the opposite sign. The two agreed until today, because the unweighted grid-cell average held this series below the detection bar. Gauge-based precipitation against ERA5, a 12-month index against three months, and 1950–2020 against 1979–2026 all bear on the comparison, which stands unresolved rather than settled by preferring one dataset.
    • The full audit — reproduction, five weighting variants, an aridity-class decomposition, construction sensitivities and four figures — sits in the repository under audit/drought/, with every number traceable to a CSV.
    • Drought went from one measure to three, on the global Drought page and all six continent pages: meteorological (a rainfall deficit), agricultural (a root-zone soil-moisture deficit) and hydrological (a river low-flow deficit).
    • Africa and South America show a detected increase on all three. Europe, North America and Oceania detect on none; Asia detects only in soil moisture. Globally, only hydrological drought detects — the regional signals do not aggregate to a global one.
    • Hydrological drought reads the low tail of the GloFAS river-discharge record the Flooding page already uses for the high tail. Agricultural drought adds a new ERA5 soil-moisture backfill, root zone 0–100 cm plus the surface layer, 573 months.
    • The AR6 comparison now tests the report's "agricultural drought in some regions" finding directly, and the Methodology page and PDF cover both new indices.
    • Two corrections to the Detection & Attribution page. It had described Africa's soil moisture as a detected increase; at p=0.079 against a false-discovery-adjusted bar of 0.05, it does not qualify, which leaves South America as the only continent detecting on all three drought measures. It had also counted Frost Days among the cold extremes showing a detected decrease; at p=0.053 against a 0.046 bar, Frost Days misses that bar too, so the count there reads 2 of 4 rather than 3 of 4.
    • A record of 30 years or fewer now suppresses the trend line — the boundary previously admitted an exactly-30-year record. The Every Variable table marks each affected series "time series too short" rather than rendering a detection judgment.
    • The Methodology page gained a Data Sources — Attribution & Licensing section naming every provider, the citation each one asks for, and its license terms.
    • A chart that hides its trend line for a too-short record now gives the reason in its legend. The Every Variable table names those series.
  8. Launch
    • The dashboard went live at globalextremeweather.thehonestbroker.org with eight phenomena pages — tropical cyclones, heat waves, cold extremes, convective instability, wildfire, extratropical cyclone intensity, drought and flooding — each drawn from a publicly available official source.
    • Alongside them at launch: six continent pages, from Africa to South America, breaking every phenomenon down by region; the Library; the reworked Loss Normalization page; and the AR6 consistency layer that sets each result against what the IPCC concluded for that hazard and region.
    • The header links to the companion US dashboard, and a link to this site unfurls with a title, description and image on social platforms.