
(© Feng Yu - stock.adobe.com)
Study Reveals ‘Warming Hole’ Where No Detectable Increases Found At Any Temperature Level
In A Nutshell
- Only 44% of U.S. states show warming in average temperatures since 1950, but 84% show warming when scientists examine the full temperature distribution
- Eight Southern states (Alabama, Arkansas, Illinois, Kansas, Mississippi, Missouri, Oklahoma, Texas) form a “Warming Hole” showing no detectable warming at any temperature level
- States ranking higher in warming measurements show about 78% correlation with how those states vote in presidential elections
- Three warming types exist: states heating more at coldest temperatures (Type W2), states heating more at hottest temperatures (Type W3), and states warming uniformly (Type W1)
Talk about climate change in Oklahoma or Kansas, and you’ll likely get pushback. The numbers seem to support the skeptics: neither state shows statistically significant warming in average temperatures since 1950. Case closed, right?
Not quite. A new study analyzing 72 years of daily temperature data reveals that focusing on averages hides most of the warming story. Researchers at the University of Zaragoza and Universidad Carlos III de Madrid found that while only 44% of states show rising average temperatures, 84% are warming when scientists look at the full temperature picture, including the coldest nights and hottest days.
The gap matters because it means millions of Americans are experiencing climate change even where average temperatures look stable. Warming often shows up first in extremes: brutal summer heat waves in Florida, disappearing subzero nights in Minnesota, record-breaking temperatures in California. By the time averages catch up, the changes are already disrupting lives.
The researchers published their findings in PLOS Climate, introducing what they call “Warming Dominance.” This is a way to compare not just whether states are warming, but how and where in the temperature distribution that warming appears.

Tracking Temperature Change Beyond the Average
The research team analyzed daily mean temperatures for every county in the contiguous United States from 1950 through 2021. Rather than simply calculating annual averages, they examined temperature trends at multiple points across the distribution—tracking changes in the coldest days (5th percentile), median temperatures (50th percentile), and the hottest days (95th percentile), along with several points in between.
They tested each temperature measure for statistically significant trends using methods designed to account for natural year-to-year variation. A state was classified as “warming” if researchers could detect an upward trend at any point in its temperature distribution with 95% confidence. The dataset comprised 26,298 daily observations per state, aggregated from county-level measurements weighted by land area.
To compare warming intensity across states, the team developed a “Warming Dominance” index that ranks which states are heating up faster than others at various temperature thresholds.
The Warming Hole: Eight States Defy the Trend
The most unusual geographical pattern appears in the South and Midwest: eight states showing no detectable temperature trends at any level.
Alabama, Arkansas, Illinois, Kansas, Mississippi, Missouri, Oklahoma, and Texas all fall into this category, mostly concentrated in the Southeast climate zone. While much of the planet warms, this region has stayed relatively stable since 1950.
Scientists have proposed several explanations. Aerosol pollution might reflect sunlight and mask warming. Complex ocean interactions between the Atlantic and Pacific could create cooling patterns. Land use and agriculture changes may also contribute.
The new study doesn’t solve the Warming Hole mystery but confirms its existence using rigorous statistical methods. Researchers found these eight states showed no significant trends even when examining temperature extremes, setting them apart from nearly every other state in the nation.
A 78% Correlation With Voting Patterns
The researchers noticed a strong relationship in their data: states ranking higher in their warming measurements show about 78% correlation with how those states vote in presidential elections.
States with the strongest warming signals cluster along the Pacific and Atlantic coasts, regions that lean Democratic. Many states in the Warming Hole and surrounding areas lean Republican.
The pattern raises an obvious question: Does experiencing climate change firsthand shape opinions about it? Or does the correlation reflect other factors like coastal geography, urbanization, or economic structures? The researchers didn’t establish causation, but the connection adds another layer to debates about climate politics. As warming becomes more visible in some regions than others, the partisan divide over climate policy could widen.
Different States, Different Warming Stories
The study identified three distinct types of warming across the United States, each with different implications for how residents experience climate change.
Type W2 states like Iowa, Minnesota, Montana, Nebraska, North Dakota, and South Dakota are warming more at their coldest temperatures. The harshest winter days aren’t as harsh. This can bring longer growing seasons and lower heating bills, though it also disrupts ecosystems adapted to deep cold.
Type W3 states like California, Idaho, Nevada, Oregon, Washington, and Wyoming show the opposite. These states are heating up more during their hottest days. Record-breaking heat events are becoming both more frequent and more intense, creating ideal conditions for wildfires, drought, and heat emergencies.
Type W1 states in the Upper Midwest, parts of the Northwest, and the Northeast show uniform warming. Cold days and hot days are warming at roughly the same rate.
California faces particular challenges. The state’s hottest days are warming faster than the hottest days in most other states. Combined with vulnerability to wildfires, drought, and agricultural disruption, this makes California a test case for Type W3 adaptation.
The Coasts Are Heating Fastest
Rhode Island, Arizona, Connecticut, California, Delaware, Massachusetts, New Jersey, and Maryland rank highest for warming intensity.
All but Arizona sit on a coast, mostly in the Northeast and Southwest. Coastal states may need more planning for sea level rise, storm surge, and heat waves. Western states face escalating wildfire risk and water shortages. States in the Warming Hole may experience delayed impacts, potentially affecting public support for climate action.
Why Averages Miss the Story
Florida shows no significant trend in average temperature but clear warming in its hottest days. Someone citing the average might conclude Florida isn’t experiencing climate change. But residents sweating through increasingly brutal summer heat waves know better.
The reverse happens in northern states where winters are milder even as summers stay stable. The lived experience involves dramatically different winters that averages don’t capture.
People notice extreme temperatures more than gradual shifts. That’s why someone experiencing sweltering heat or disappearing cold snaps is more likely to accept climate change than someone in a place where temperatures shift uniformly and subtly.
The contiguous United States as a whole shows Type W1 warming: relatively uniform increases across temperatures. But this national average obscures wild variation from state to state. From 1990 to 2021, the proportion of warming states jumped to 87%, with more Type W3 states appearing.
The Warming Hole may not last. Oklahoma shows no warming from 1950 to 2021 but has started showing modest trends in recent decades. As the planet heats, even resistant regions may eventually show clear signals.
America’s warming map is a patchwork of different warming types, rates, and experiences. Understanding that complexity matters for climate policy, because Americans aren’t all experiencing the same changes.
Paper Notes
Limitations
The study has several limitations worth noting. The analysis focuses exclusively on temperature trends and doesn’t examine other climate variables such as precipitation, humidity, or extreme weather events. The temperature data, while comprehensive, relies on county-level aggregations that may smooth out localized variations within states. The study period of 1950-2021 captures substantial warming but may not reflect the full scope of changes that began with industrialization in the late 19th century.
The statistical methods used are designed to detect trends but can’t definitively attribute causes. While the researchers discuss possible physical mechanisms for regional warming patterns, the study doesn’t perform formal attribution analysis to separate human-caused warming from natural climate variability.
The Warming Dominance comparisons assume linear trends, which may not capture nonlinear acceleration or deceleration in warming rates. The correlation with political voting patterns is observational and doesn’t establish any causal relationship in either direction.
Funding and Disclosures
This research was funded by Gobierno de Aragón and European Regional Development Fund (ERDF, EU), grant LMP71-18; Agencia Española de Investigación and ERDF, grants PID2020-114646RB-C44, PID2023-147593NB-I00, PID2023-150095NB-C44, and RED2022-134122-T; MCIN/AEI and European Union NextGenerationEU/PRTR, grant TED2021-129784B-I00; MCIN/AEI grant CEX2021-001181 (María de Maeztu); and Comunidad de Madrid grants EPUC3M11 and V PRICIT. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors declared no competing interests.
Publication Details
Authors: María Dolores Gadea Rivas (Department of Applied Economics, University of Zaragoza, Spain) and Jesús Gonzalo (Department of Economics, Universidad Carlos III, Getafe, Spain) | Journal: PLOS Climate | Title: Regional heterogeneity and warming dominance in the United States | DOI: 10.1371/journal.pclm.0000808 | Published: February 4, 2026
Data Availability: Temperature data were obtained from PRISM-based gridded datasets (1950-2019) compiled by Wolfram Schlenker and extended to 2021 using recent PRISM data. The processed datasets and analysis code are available from the corresponding author upon request.







