US Extreme Weather and Climate Change Dashboard

US Drought Monitor, weekly national area by severity category. Unit: % of CONUS.

The reliable window spans USDM's whole record — 2000 is the record start rather than a sub-window found within the data, so there's no earlier period to exclude here. USDM's native categories (D0 abnormally dry through D4 exceptional drought) are already cumulative ("this severity or worse") in the source data rather than raw per-band values — D2+ is the level selector's default, chosen as the closest USDM analogue to a "1-in-10 or worse" severity cutoff (full rationale on the Methodology page), but the selector offers every category.

Source: US Drought Monitor (USDM)

Computing trend…

SPEI computation (Hargreaves PET, log-logistic standardization) -- a second, independently-derived drought index alongside PDSI, using a different PET formula specifically so a drought finding meets a documented critique that PDSI's formula overstates warming-driven drying. Unitless; negative values are dry, positive are wet. Unit: SPEI.

Uses the same GHCN-Daily HCN station network that shows a reliable trend from ~1895; the pipeline exports station count per month so a thin-network period shows rather than hides. Uses a different evaporative-demand formula from PDSI's (Hargreaves rather than the Thornthwaite family PDSI builds in), the specific formula Milly & Dunne (2017) show overstates warming-driven drying -- the two indices meaningfully diverge for 2021-2025 (PDSI strongly negative, SPEI only mildly so), a real finding disclosed on the Methodology page. That claim now reads narrower than it once did. Xu et al. (2026, Communications Earth & Environment 7:726) group Hargreaves with the formulas that lack a land-atmosphere coupling constraint, and report such formulas inflating drying trends at least sixfold over energy-constrained ones. So this series escapes PDSI's particular formula without escaping the wider critique, and it reads as an upper bound on the drying signal rather than as a settled answer.

Source: NOAA GHCN-Daily (temperature + precipitation), SPEI computed in-house

Computing trend…

SPI from the same HCN station precipitation the SPEI series uses, standardized the same way, with the evaporative-demand term set aside. Unitless; negative values are dry, positive are wet. Unit: SPI.

The one drought measure here that depends on no potential-evapotranspiration formula. PDSI uses a Thornthwaite-family formula, SPEI uses Hargreaves and the CPC soil-moisture series comes from a model driven by one -- and Xu et al. (2026, Communications Earth & Environment 7:726) report that formulas lacking a land-atmosphere coupling constraint inflate drying trends at least sixfold, with precipitation supplying 75% of the drought trend under the constraint. So this series answers a question the other three cannot, and the SPI-minus-SPEI gap measures what the evaporative-demand term contributes. The standardization deliberately keeps SPEI's log-logistic fit rather than McKee, Doesken & Kleist's (1993) gamma, leaving the evaporative-demand term as the single difference between the two series; a gamma fit would confound a change of distribution with a change of water-balance definition. Read this as a log-logistic SPI variant paired with the SPEI above rather than as a published-product SPI. See Methodology.

Source: NOAA GHCN-Daily (precipitation), SPI computed in-house

Computing trend…

% of 3,308 CONUS land cells (0.5° grid) whose 3-month mean soil moisture falls at a selectable USDM-equivalent severity level or worse against that cell's same-calendar-month history — the same modeled-soil-moisture-percentile approach USDM uses as one of its standing inputs, broken out as its own series to expose one USDM input. Unit: % of CONUS land area.

A rainfall deficit becomes an agricultural drought only once the soil store draws down, so this series lags and damps precipitation rather than tracking it. CPC calculates soil moisture with a one-layer "leaky bucket" model driven by observed precipitation and temperature, rather than measuring it. That gives a single column of soil water instead of separate surface and root-zone depths, and makes the series a model estimate constrained by observations rather than an observation. Read the trend with the record's start in mind: the two driest years here are 1956 and 1954, so Sen's slope measures +0.04 points/decade over the full 1948 record but +1.96 starting from 1990 — cutting the 1950s Southern Plains drought produces the drying trend, and so this series keeps the full record. See Methodology.

Source: NOAA CPC Monthly Soil Moisture v2 (Fan & van den Dool 2004), NOAA PSL

Computing trend…

The mirror image of the Flooding page's high-flow index: % of HCDN-2009 reference gauges currently at or below the historical 10th percentile for the day of year -- a runoff-based drought signal, distinct from PDSI/USDM/SPEI's precipitation/soil-moisture basis. Unit: % of reference gauges below threshold.

Shares the high-flow index's 1957 reliable-trend window (the same underlying gauge-network coverage finding applies to either tail). 1934/1931 rank as the two driest years in the full record, and 1988/2012 both land in the reliable window's top 10.

Source: USGS NWIS daily discharge, HCDN-2009 reference gauges

Computing trend…