ChescoWx Climate Studies · Station siting · Contiguous U.S.
Airport Stations Have Warmed Faster Than Other Stations Since 1950
985 NOAA stations at airports and air bases set beside 9,858 other stations across the contiguous 48 states, 1950 to 2025, using NOAA's raw readings
In our PHL study, Philadelphia International Airport pulled away from the 23 stations around it by about 0.2°F per decade. Airports everywhere have grown since the 1950s, with longer runways, more pavement, bigger terminals and parking, and development around them. That raises a broader question: do airport thermometers as a group warm faster than the other stations in NOAA's network? This study takes the station records from our Raw vs NCEI study and splits them in two: airport stations and everything else.
Airports vs other stations, raw+0.23°F
per decade faster at airport stations since 1950: +0.26 vs +0.03. Statistically significant (p < 0.001).
Same test, anomaly method+0.13°F
per decade faster at airports: +0.34 vs +0.21. Just short of statistical significance (p = 0.07).
State by state42 of 48
states where the airport stations warmed faster than the state's other stations since 1950 (42 of 47 under the anomaly method).
1 · The results
Faster under both methods, significant in the raw readings
| 1950–2025, °F per decade | Other stations | Airport stations | Difference | NCEI adjusted, contiguous U.S. |
Raw readings averaged as recorded | +0.03 not significant | +0.26 significant | +0.23 ±0.06 p < 0.001 | +0.35 significant |
Anomaly method each station vs its own 1951–1980 average | +0.21 significant | +0.34 significant | +0.13 ±0.15 p = 0.07 |
"Significant" means the trend is real at the 95% level, allowing for runs of warm or cool years. The difference is tested on the year-by-year gap between the two series.

Annual average temperature, contiguous 48 states, each station measured against its own 1951–1980 average. Dashed lines are linear trends; * = statistically significant. The airport-only line runs almost on top of NCEI's adjusted national series (+0.34 vs +0.35 °F per decade).

The same comparison with the readings averaged as recorded. Each line is shown against its own 1951–1980 average so the three can share one axis.
One thing stands out: under the anomaly method, the airport stations alone warm at almost exactly the rate of NCEI's adjusted national series. The other 9,858 stations warm more slowly in both methods.
2 · When it happened
The gap opened from the 1970s on

Airport stations minus other stations, decade averages, relative to their 1951–1980 difference. Through the 1960s the two groups moved together. From the 1970s the airports pulled ahead, to about 1.3°F (raw) and 0.9°F (anomaly method) in the 2020s.
The gap grows across five decades rather than in a single jump. A one-time equipment change would show up as a step; a gradual drift fits growth around the sensors, but also any other change that spread through the network over the same decades (see section 5).
3 · State by state
It shows up in most states, not just a few big hubs

Each dot is one state: the airport stations' 1950–2025 trend minus the state's other stations' trend. Orange = airports warmed faster. Raw: 42 of 48 states, median +0.18°F per decade. Anomaly method: 42 of 47 states (one state's airport series is too short for this method), median +0.15°F per decade.
4 · Does this change the Raw vs NCEI results?
Very little: airports are about 9% of the stations
| Contiguous U.S., 1895–2025, °F per decade | All stations | Without airports |
| Raw | +0.01 | −0.01 |
| Anomaly method | +0.12 significant | +0.11 significant |
| NCEI adjusted | +0.18 significant |
Our Raw vs NCEI study uses every qualifying station, airports included. Each state's record is averaged county by county, so the airports carry no more weight than the co-op stations around them, and taking them out moves the national raw trend by only 0.02°F per decade.
5 · What could explain it
Airport growth is one candidate, and not the only one
This study shows that airport stations warmed faster than other stations since 1950. It does not show why. Several changes over the same decades could each contribute, and this test cannot separate them:
- Growth around the sensors. More runways, taxiways, aprons, buildings, parking and traffic at and around airports, and the cities growing out to them. This is the question raised by our PHL study.
- Instrument changes at airports. Airport stations moved to new electronic sensors in the 1980s and to the automated ASOS system in the 1990s, and some early units are known to have read warm. Instrument changes usually cause steps rather than a steady drift, but they happened at different times at different airports.
- Reading time at the other stations. Airports record midnight to midnight. Many cooperative observers switched from afternoon to morning readings over these decades, which tends to pull their raw trends down. Neither the raw nor the anomaly method corrects this. In our Chester County work, correcting reading times added about +0.11°F per decade to the county's raw trend since 1941, similar in size to the anomaly-method gap here.
- Station moves. Airport stations have moved within airport grounds, and co-op stations have moved between observers' homes.
Next steps: repeat the comparison against co-op stations corrected only for reading time; compare each airport with the rural stations nearest it, as we did for PHL; and check individual airports against their documented site and equipment histories.
6 · How we tested it
Method
- Readings: NOAA's raw daily highs and lows (GHCN-Daily v3.35) for all 10,843 stations in our Raw vs NCEI study, with the same rules: a month counts only with 3 or fewer missing days, a year only with all 12 months, and a station needs at least 10 complete years.
- Airport stations: stations whose NOAA name marks them as an airport, airfield or air base (for example "AP", "AIRPORT", "INTL", "RGNL", "MUNI", "AFB", "NAS", "ASOS"). That gives 985 airport stations and 9,858 others. The list is in our data files.
- Combining: exactly as in the Raw vs NCEI study. Stations are averaged by county, counties are weighted by area into each state, and states are weighted by area into the contiguous U.S. Run on each group separately. For airports only, each year uses the states that have an airport station that year (46 to 48 states in every year since 1950).
- Anomaly method: each station is measured against its own 1951–1980 average, using only stations with at least 15 complete years in 1951–1980 (478 airport stations, 3,763 others). Stations that opened later, including most newer automated airport sites, do not enter it.
- Why 1950: before about 1935 fewer than 30 states have any airport station, so the airport-only series is too thin. From 1950 on, airports cover the country.
- Check: run on all stations together, the method reproduces the published Raw vs NCEI national trends exactly (raw +0.01, anomaly +0.12, NCEI +0.18 °F per decade since 1895).
7 · Limits
What this study does and does not do
- Airport stations are identified by name, so a few will be misclassified. Some airport stations with plain names are counted as other stations, which would likely make the measured gap slightly smaller than the true one.
- The number of airport stations roughly doubled in the 2000s as automated sites were added. That can shift the raw (as-recorded) airport average. The anomaly method avoids this, since new sites do not enter it.
- Some airport records may continue a station that began in town before moving to the airport.
- The study describes what the stations measured. It does not separate airport growth from equipment changes, reading-time changes or station moves (section 5).
- All methods involve choices made by people, including ours. That is why the raw and anomaly results are both shown, next to NCEI's.