2026 was not the “hottest summer on record” in the Conterminous United States. This framing by CNN and most other legacy media outlets is highly misleading, even if the information that they're reporting is correct [at face value]. The term “hottest” implies that the average maximum temperatures (daytime highs) were the highest on record. But NOAA NCEI's nClimDiv dataset shows that they were 4th highest on record behind 1936, 1934, and 2012, respectively. What they should say is that it was the “warmest on record” (but even that might not be true for reasons I'll mention in a later post). Even NOAA does not claim 2026 was the “hottest summer on record” in U.S. They carefully said “warmest.” 🔗ncei.noaa.gov/news/national-… The reason why summer 2026 was the “warmest” on record is due to very high overnight low temperatures, not so much daytime highs. Most of the extreme heat in terms of daytime highs this past summer was also confined to the Desert Southwest / Four Corners area of the Lower 48. I plotted a map of the average temperature rankings for all 344 U.S. climate districts NOAA uses below. Notice that there are very few No. 1s in the U.S. this year, and that all are confined to the Southwest (as well as the majority of the other top fives). But if you look at the summer of 1936, the No. 1s cover about half of the map and it was a top five warmest summer for most of the country. So, spatially speaking, 2026 came nowhere close to 1936, and anyone claiming otherwise is being highly dishonest. It is an apples-to-oranges comparison.

There is no dumber metric than the “average mean temperature” (Tavg). How is “average mean temperature” calculated? Well, you take a station's daily maximum temperature (Tmax) and daily minimum temperature (Tmin) and then divide it by two. That is just an arithmetic mean. This is standard operating procedure. But there are TWO major problems with Tavg. 1⃣ Tavg does not take into consideration fluctuations in temperatures throughout the course of a day. For example, if a strong cold front comes through at 1:00 a.m. in January, the high temperature at a station may occur shortly after midnight, but the rest of the day will be much colder due to cold air advection. But by using only the Tmax and Tmin values, the average gets skewed high because it was not calculated using 24 separate hourly observations or 288 separate five-minute observations. To illustrate this better, suppose that the temperature observations (in °F) at the top of the hour at DCA for a random day in January (from midnight to 11:00 p.m.) are as follows: 60°, 62°, 50°, 43°, 32°, 26°, 22°, 18°, 19°, 20°, 21°, 23°, 22°, 22°, 23°, 24°, 21°, 20°, 19°, 18°, 16°, 15°, 14°, and 12°. If you calculate Tavg using the mean of those 24 separate hourly observations, Tavg = 21.9°F. However, if Tavg is calculated using only Tmax and Tmin, then Tavg = 37°F, some 15.1°F higher. Similarly, the Tavg can skew low due to the northward advance of a warm front, especially during the winter (although in my experience, this is less common than the previous scenario). 2⃣ Tavg (no matter how it's calculated) is going to be heavily contaminated by the urban heat island effect (UHIE) and other land use forcings, which significantly affect overnight Tmin and morning readings. Surfaces such as asphalt, concrete, and brick have a high heat capacity, allowing them to store more solar energy during the day than vegetation or bare soil do. After sunset, that stored heat is released gradually, slowing nighttime cooling of the urban environment. Buildings also inhibit a lot of that infrared radiation (IR) from making it back into outer space. Cities also tend to have more mechanical turbulence in the lower atmosphere because buildings increase surface roughness. In rural settings, the surface cools rapidly after sunset, but the air just a few hundred feet up does not. This causes what's called boundary layer decoupling, and it is what causes temperature inversions, especially during the fall and winter. But in urban environments, buildings enhance turbulence and therefore mechanical mixing, which mixes the warmer air aloft down to the surface. Together, these two processes act to prevent the air just above the ground (where temperatures are measured) from cooling as quickly at night, which the figure on the left shows. This effect is strongest when a site is first developed beyond wilderness conditions (so UHIE affects rural and suburban environments too), but it has grown larger spatially across the global land surface because more land is being developed with time. This significantly affects Tmin trends, which is a significant (and perhaps the main) reason why Tmin (and by extension, Tavg) has gone up, at least in the mid-latitudes. In rural environments, Tmin is also affected by corn sweat (evapotranspiration off of corn that increases the atmospheric vapor pressure; that is, the humidity, which means higher dew points and higher Tmin). Case in point, on the right, I plotted the annual mean Tmax (red), Tavg (green), and mean Tmin (blue) for Huntsville, Alabama from 1920 to 2025. As you can see, the annual average Tmax has not increased at all. The statistically significant increase in Tavg, though, comes entirely from the even larger increase in Tmin, which is driven by UHIE. And this is not just exclusive to Huntsville; you can see it in Fresno or Boston and virtually anywhere else that isn't purely rural and/or removed from cornfields. I like using average Tmax because it represents how warm (or hot) it actually gets (which is relevant to global warming) and, more importantly, it represents a well-mixed planetary boundary layer (PBL) where UHIE is minimized without having to adjust data that much using hypotheticals.