■ CRITICAL RISK ■ Science & Research
For routine surface observations, the takeover is decades old — ASOS and AWOS stations do the observing, and the human role has contracted to augmenting the machines where aviation safety demands it. The dedicated observer is now one of meteorology's rarest job titles.
“Automated weather stations observe 24/7 without needing a coffee break.”
Our AI replacement risk score — how we score jobs
A weather observer's craft is systematic noticing: measuring temperature, pressure, wind, and precipitation; estimating cloud height, type, and coverage; judging visibility against known landmarks; and encoding it all into METAR reports on a strict schedule, every hour, forever. At airports this feeds directly into flight safety; historically, networks of human observers built the entire climate record your weather app is trained on.
Automation arrived early and thoroughly. The FAA and National Weather Service deployed ASOS (Automated Surface Observing Systems) across the US through the 1990s — ceilometers measure cloud height by laser, visibility sensors replace the landmark squint, precipitation discriminators tell rain from snow, and the whole package transmits without a human within miles. Most of the world's surface observation network is now automated stations, buoys, and satellites. The residual human role is 'augmentation': at larger airports, contract weather observers or trained tower personnel correct the machine when it misjudges — a thunderstorm ASOS can't see approaching, a tornado, volcanic ash, patchy fog that fools a single-point visibility sensor. That's a real safety function, and it employs a small number of people.
The interesting resistance isn't technical, it's about edge cases with lives attached: aviation regulators keep humans in the loop precisely at the busiest airports because sensor blind spots have consequences. But sensor suites, camera networks, and AI weather-recognition keep narrowing those blind spots, and each contract renewal cycle trims augmentation sites. Observing as a career path has effectively folded into meteorology, aviation operations, and instrument-technician work. Our risk score of 91 describes an occupation already reduced to its final, small, regulated remnant.
Automatability: our editorial assessment of current and near-term AI capability
The main displacement finished last century — ASOS/AWOS rollouts automated routine surface observation across the 1990s and 2000s. What remains is a slow endgame: contract augmentation positions at major airports get reviewed and occasionally trimmed as sensor and camera coverage improves, and AI cloud/visibility recognition targets the last human advantages. Expect the remnant roles to persist at the busiest hubs into the 2030s, shrinking with each technology refresh.
Yes, but barely. Automated ASOS/AWOS stations handle routine observations almost everywhere; humans remain mainly as 'augmenters' at larger airports, correcting the sensors during thunderstorms, ash, or tricky visibility, plus a few research and polar stations. It's a small, specialized remnant of what was once a large observational workforce.
Single-point sensors have blind spots: a visibility sensor measures one spot while fog sits in patches, a ceilometer stares straight up while a thunderstorm approaches sideways, and no ASOS discriminator recognizes a funnel cloud. At busy airports those gaps carry safety consequences, so regulators keep trained humans augmenting the machines — though cameras and AI recognition keep shrinking the gap list.
The natural moves are aviation operations and dispatch, meteorological technician work maintaining the automated networks, forecasting (with added credentials), and emergency management. Observers' real asset is disciplined weather judgment under safety pressure — that transfers well; the manual observation routine itself does not.
Forecasting is being transformed — AI weather models now rival traditional physics models at a fraction of the compute — but that changes the forecaster's job rather than deleting it, since interpretation, communication, and warnings still need humans. Observation is the part that automated almost completely. If you love weather, aim at the analysis-and-decision end, not the measurement end.