NOAA keeps a network of 1,218 long-running weather stations in the 48 contiguous states, with daytime high temperatures going back to 1895. A common way to summarize a record like this is to draw a straight line through the yearly averages and report how steeply it rises or falls, in degrees per century. This chart draws that line between every start year and every end year at least 30 years later, the period the World Meteorological Organization uses to define normal climate, 5,253 lines in all. It does this first for the raw thermometer readings and then for NOAA’s adjusted record. Each square is one line. Pick a start year along the bottom and go up to an end year. The color at that spot shows the line’s slope: blue means it falls, which is cooling; red means it rises, which is warming; pale means it is close to flat. The solid black contour marks lines that are exactly flat. The dashed contour marks lines rising 0.9 degree Fahrenheit per century, about half the rate in NOAA’s adjusted record from 1895 through 2025. In the raw readings, on the left, the answer depends on which years you pick. American daytime highs peaked in the 1930s, sagged through the 1960s and 1970s, and rose again after 1980. A line that takes in the 1930s peak and stops before the recent warm years have built up slopes downward, and those lines make up the large blue region. Of all 5,253 lines, 46 percent are flat or falling, and 70 percent rise 0.9 degree per century or less. Starting in the same year and stopping two years later can turn cooling into warming. A line through the raw readings from 1931 to 2023 shows cooling of 0.11 degree per century. The same line from 1931 to 2025 shows warming of 0.11 degree. Every start year from 1895 through 1946 gives a raw trend through 2025 of 0.9 degree per century or less. A single degrees-per-century figure tells you as much about the years someone chose as about the climate. In NOAA’s adjusted record, on the right, only 21 percent of the lines are flat or falling, and most of the rest are clearly red. The blue shrinks because NOAA’s adjustments lower every year from 1902 through 1979 and raise every year since 1980, measured against the 1961 to 1990 average. That cuts down the 1930s peak and lifts the recent years. The blue that remains comes from the same peak and dip, now smaller. The years are the same in both panels. Most of the difference comes from NOAA’s adjustments to the readings. A small part comes from the estimates NOAA fills in for missing months and closed stations: with them, the adjusted trend from 1895 through 2025 is 1.74 degrees per century, and without them, 1.64. From 1930 through 2025, the raw trend is 0.10 degree per century and the adjusted trend is 1.86. NOAA’s scientists have given two main reasons for the adjustments. Volunteer observers gradually shifted from afternoon to morning readings. In the 1980s, many stations switched to electronic sensors that read daytime highs lower than the thermometers they replaced. NOAA’s scientists say both changes made the raw record show less warming than actually occurred. The chart cannot show whether NOAA’s corrections are the right size. It does show where the warming comes from. Counting only stations and years that have both a raw reading and an adjusted value, nearly two-thirds of the warming in NOAA’s record for these stations since 1895 comes from NOAA’s own corrections, not from what the thermometers recorded. Measured from 1925, nearly all of it does. My upcoming article features the work of John Shewchuk, a meteorologist and retired Air Force lieutenant colonel. Shewchuk has documented 232 weather stations that NOAA still carries on estimated temperatures after they closed, and he argues that NOAA’s adjustments exaggerate warming.

NOAA’s official USHCN page is outdated; it claims the network actively measures national temperature trends while omitting its 2014 replacement by the nClimDiv dataset, its heavy reliance on estimated data, and the true number of inactive stations. A forthcoming article next week will feature the work of John Shewchuk, a meteorologist and retired Air Force lieutenant colonel, who has investigated these structural dataset discrepancies.