Skip to sources
Time Machine

Atmospheric and Space Scientists

Scrub through 166years of this role's history, from when it first emerged, through every wave of technology that reshaped it, to the cited projections for where it's heading next.

2026drag to travel through time
187519001925195019752000now
Country
2026
Known today as Atmospheric and Space Scientists (BLS SOC 19-2021)
Latest actual · 2024
9K
BLS OEWS May 2024 employment for SOC 19-2021 (Atmospheric and Space Scientists), confirmed via O*NET and the BLS national employment matrix for the 2024-2034 projections cycle. Sector distribution: federal government 34.1%, professional/scientific/technical services 29.4%, information sector 16.7%, educational services 16.6%. The federal government share reflects roughly 3,200 staff at NOAA, NWS, Air Force Weather, and NASA atmospheric science positions. Median annual wage: $97,450.
Latest actual · 2024
$97,450
BLS OEWS May 2024 median annual wage for SOC 19-2021 (Atmospheric and Space Scientists). The lowest 10 percent earned less than $49,990; the highest 10 percent earned more than $160,710. Private-sector climate risk firms (ClimateAi, Jupiter Intelligence, Salient Predictions) paid AI-fluent atmospheric scientists in the $100k-$160k range as of 2025-2026, above the NWS median.
Each dot is a cited figure over time; the dotted line only links them (values between aren't measured). Hollow dots are estimates.
Tools of the era

The tools that defined the work

Select an era to see how it reshaped the work.

  • Telegraph network and synoptic chart (Weather Bureau founding era)

    The foundational tool of professional meteorology was the electric telegraph combined with hand-drawn synoptic weather maps. The Signal Service established a nationwide network of observation stations connected by telegraph; readings were transmitted at fixed hours to the central office in Washington, where staff plotted surface observations onto charts and identified storm systems. This was the entire technical apparatus of the profession for its first fifty years: systematic observation, transmission, and human pattern recognition on paper maps. The Smithsonian Institution had pioneered the observation network in the 1850s, but it was the telegraph that turned weather observation from a correspondence exercise into near-real-time operational forecasting.

    Effect on the work

    The telegraph network created the occupation itself: without a communications infrastructure fast enough to outrun storms, professional weather forecasting was impossible. The first 24 stations in 1870 grew to roughly 500 stations by 1900, each requiring trained observers and creating a growing federal scientific workforce.

    Work toolChanging equipment
  • Radiosonde and upper-air sounding network (1937 US launch, operational by 1940s)

    The radiosonde, a small instrument package carried aloft by weather balloon to measure temperature, pressure, humidity, and wind at altitude, gave meteorologists their first systematic view of the three-dimensional atmosphere. The United States launched its first radiosonde in 1937; by the early 1940s a nationwide upper-air sounding network was operational, with ascents twice daily at dozens of stations. For forecasters, the radiosonde transformed the craft: upper-level charts showing jet stream position, temperature advection, and moisture flux became the foundation of synoptic analysis. The ability to see the atmosphere in three dimensions was the prerequisite for every subsequent development in numerical weather prediction.

    Effect on the work

    Upper-air analysis was cognitively demanding and created a new specialty within the profession. Radiosonde data processing, tephigram construction, and upper-air chart analysis required specialized training, contributing to the formalization of meteorological education and the establishment of university programs in atmospheric science during the 1940s and 1950s.

    Work toolChanging equipment
  • Numerical weather prediction on mainframe computers (ENIAC 1950, Joint NWP Unit 1954)

    The first numerical weather prediction was computed on the ENIAC digital computer at the Institute for Advanced Study in Princeton in April 1950, by a team led by Jule Charney and including John von Neumann. The calculation required round-the-clock operation of the machine and took more than 24 hours to produce a single one-day forecast, but it worked: the equations of atmospheric dynamics could be solved on a computer. The Joint Numerical Weather Prediction Unit was established in July 1954 as a collaboration of the Weather Bureau, Air Weather Service, and Navy, and began issuing experimental numerical forecasts operationally by 1958. NWP on mainframes transformed the forecasting workflow: instead of pattern-matching from memory and charts, forecasters began interpreting model output, checking it against observations, and applying local knowledge to adjust the guidance. The meteorologist's role shifted from primary analyst to skilled interpreter.

    Effect on the work

    NWP did not reduce forecaster employment but changed the skill mix required. Proficiency in atmospheric dynamics and thermodynamics became more important; hand-drawing skills became less so. The transition drove the expansion of university atmospheric science programs: by 1970 every major research university offered graduate meteorology, and the profession had a credentialed academic pipeline for the first time.

    Mainframe processingComputerized records
  • Geostationary weather satellite (TIROS 1960, GOES series from 1975)

    TIROS-1, launched April 1, 1960, returned the first satellite images of Earth's cloud cover and immediately transformed tropical forecasting: for the first time, atmospheric scientists could see the structure of hurricanes and large-scale cloud systems over data-sparse ocean regions. The GOES (Geostationary Operational Environmental Satellite) series, beginning with GOES-1 in 1975, provided continuous coverage of the Western Hemisphere from geostationary orbit, enabling forecasters to track storm development, monitor cloud-top temperatures as a proxy for convective intensity, and analyze mesoscale weather features in near real-time. Satellite imagery became the primary situational awareness tool for severe weather forecasting and a defining feature of the television meteorologist's presentation.

    Effect on the work

    Satellite data processing created a new specialty (satellite meteorologist) and a large civilian government workforce at NOAA's satellite operations centers. The visual clarity of satellite imagery also democratized weather awareness for the public, fueling demand for broadcast meteorologists who could translate satellite imagery into accessible forecasts.

    Work toolChanging equipment
  • WSR-88D Doppler radar network (NEXRAD) and workstation-based forecast tools

    The WSR-88D (Weather Surveillance Radar, 1988 Doppler) network, deployed to 160 sites across the United States beginning in the late 1980s and early 1990s, gave every NWS office the ability to see wind speed and direction inside precipitation systems. Doppler radar transformed severe weather operations: tornado lead times improved from near-zero (or negative, where the warning came after the tornado) to an average of 13 minutes by the 2010s. The simultaneous rollout of AWIPS (Advanced Weather Interactive Processing System) workstations gave forecasters integrated access to model output, satellite, radar, and surface observations on a single screen. This represented the modern forecaster's workstation paradigm that persisted until AI NWP tools began to supplant parts of it after 2023.

    Effect on the work

    NEXRAD deployment required additional trained forecasters at newly opened 24-hour Weather Forecast Offices, increasing NWS employment modestly in the early 1990s. The improved warning skill built public trust in NWS forecasts and reinforced the political case for maintaining a well-staffed federal forecasting workforce.

    Work toolChanging equipment
  • High-resolution ensemble NWP (GFS, ECMWF IFS, HRRR) on supercomputers

    The period from roughly 2000 to 2023 was the golden age of physics-based numerical weather prediction at scale. NOAA's Global Forecast System (GFS), ECMWF's Integrated Forecasting System (IFS), and the High-Resolution Rapid Refresh (HRRR) for convective-scale forecasting ran on supercomputers with ever-increasing horizontal resolution and ensemble size, delivering progressively more accurate 1-10 day forecasts. By the mid-2010s, day-5 forecast accuracy equaled what day-3 accuracy had been in the 1990s, and the seven-day forecast had become operationally reliable for planning purposes. The operational meteorologist's job became managing a rich ensemble of model guidance products: knowing which models performed best for specific weather patterns in their region, applying local MOS (Model Output Statistics) corrections, and communicating probabilistic uncertainty to the public.

    Effect on the work

    The supercomputer NWP era raised the technical floor for operational meteorology: a working knowledge of GFS vs. ECMWF model characteristics, ensemble probability interpretation, and MOS calibration became standard professional expectations. Private-sector weather companies (IBM Weather Company, DTN, Maxar) built businesses on value-added NWP post-processing, creating new employment outside the federal sector.

    Work toolChanging equipment
  • AI foundation model NWP: GraphCast, ECMWF AIFS, GenCast, Aurora, FourCastNet

    The publication of Google DeepMind's GraphCast in Science (December 2023) marked a qualitative break in the history of weather prediction. GraphCast, trained on 40 years of ERA5 reanalysis data, produced 10-day global forecasts in under one minute on a single TPU, outperforming ECMWF IFS on 90% of 1,380 test variables. DeepMind's GenCast (Nature, December 2024) generated 50-member probabilistic ensemble forecasts in 8 minutes, outperforming ECMWF ENS on 97.4% of 1,320 targets. ECMWF launched its AIFS (Artificial Intelligence Integrated Forecasting System) into operational use in 2024, providing near-IFS skill at less than 1% of IFS computational cost. Microsoft's Aurora foundation model (arXiv 2405.13063, May 2024) demonstrated skill across weather, air quality, space weather, and ocean wave prediction from a single 1.3-billion-parameter model. NVIDIA's FourCastNet and Earth-2 platform enabled km-scale regional downscaling at GPU speed. The paradigm shift is not incremental: the computational bottleneck that defined operational meteorology for 70 years has been removed. The forecaster who understands how AI NWP fails, and can translate its probabilistic output into accountable public guidance for high-impact events, is now the central value proposition of the profession.

    Effect on the work

    The rote production layer of medium-range global NWP guidance has been automated. The private-sector market for AI-fluent atmospheric scientists (climate risk firms, energy sector, agricultural weather intelligence) is growing rapidly, with salaries in the $100k-$160k range for scientists who combine domain expertise with AI tool proficiency. NWS forecasters face pressure to adapt their role from global-model interpretation toward high-impact event judgment, communication, and AI forecast quality control.

    Work toolChanging equipment
Projection cone · present → 2035

What credible sources project

Scrub the slider past now to anchor each scenario on the scrubber. The spread is the range of futures credible sources project for this role.

Employment outlook
Projected change in the number of people doing this work.
AMS / WMO assessment — human forecaster role in the AI NWP era (2025)
2035
+5%
Qualitative scenario synthesized from the American Meteorological Society AMS 2025 Annual Meeting AI sessions and the WMO State of Climate Services 2024 report, which both conclude that human atmospheric scientists remain essential for high-impact event interpretation, probabilistic guidance communication, and extreme-event quality control of AI NWP systems. The +5% estimate reflects a professional consensus view that demand for AI-fluent atmospheric scientists in climate risk, energy, agriculture, and insurance sectors will modestly outpace any reduction in traditional NWS operational forecasting positions. This is a qualitative synthesis, not a model-derived forecast; uncertainty is high in both directions.
BLS National Employment Matrix 2024-34
2034
+0.7%
BLS occupational employment matrix for 19-2021. Baseline employment 9,400 (2024); projected 9,500 (2034); net change approximately +100 positions. Classified as "slower than average" growth (all-occupations average is +4% over the same decade). The BLS methodology models continued federal government demand (NWS staffing is constrained by budget, not technology demand), growing private-sector climate risk services, and AI-driven productivity improvement that limits net headcount growth despite expanding application scope. The projection does not explicitly model the AI NWP disruption (GraphCast/AIFS/GenCast) or the potential for AI tools to reduce the number of operational forecasters needed per Weather Forecast Office, both of which could shift the projection toward flat or slightly negative growth at the occupation level.
AI task exposure
Share of the role’s tasks that researchers estimate AI can do. This is a measure of task exposure, not a forecast of jobs lost.
Eloundou et al. — "GPTs are GPTs" (2023/2024)
2028
45%
of tasks
GPT-4 task-by-task LLM exposure labeling on O*NET task descriptions for Atmospheric and Space Scientists (19-2021). The occupational profile scores moderate-to-high LLM exposure on knowledge-intensive tasks (literature review, writing reports and briefings, interpreting model output, climate policy communication) but low exposure on tasks requiring physical presence, instrument operation, safety-critical accountability, and spatially grounded local knowledge. The 45% task-exposure estimate reflects that roughly half the tasks in the O*NET profile are meaningful candidates for AI augmentation, while the other half (real-time severe weather monitoring, operational warning issuance, instrument QC, satellite image interpretation with local knowledge) remain substantially human. Rendered here as a task-exposure strip, not an employment decline projection.
Today, in this role

What's shifting in the work right now

The historical view above shows how this role has moved. This is the present-day detail: which AI tools are picking up which tasks, where the edge still is, and the natural directions this work can grow.

What's changing in your day

Three parts of your work where AI is already doing real lifting, and what stays yours.

AI is sitting alongside you hereEvaluate and integrate AI-based NWP guidance (GraphCast, ECMWF AIFS, FourCastNet, GenCast) alongside traditional physics-based model output (GFS, NAM, ECMWF IFS, HRRR) to produce daily zone forecasts — examining where AI and physics-based models agree or diverge at critical lead times

Evaluate and integrate AI-based NWP guidance (GraphCast, ECMWF AIFS, FourCastNet, GenCast) alongside traditional physics-based model output (GFS, NAM, ECMWF IFS, HRRR) to produce daily zone forecasts — examining where AI and physics-based models agree or diverge at critical lead times; diagnosing AI model failure signatures on extreme or out-of-distribution synoptic patterns; applying model output statistics (MOS) and AI postprocessing calibration to reduce AI-specific biases for your region; and issuing National Weather Service zone forecast products and Area Forecast Discussions that explicitly document model reasoning for the public record.[3],[5],[1]

Where your edge is

GraphCast and ECMWF AIFS have automated the majority of rote medium-range deterministic forecast production — global guidance that once required 1,000 CPU-hours now runs on a single GPU in under a minute. Your irreplaceable contribution is interpreting where AI guidance is wrong: current ML NWP systems fail systematically on extreme events (record-breaking ridge amplitudes, novel recurving cyclone tracks, deep-layer moisture advection events outside the training climatology) — exactly the events where forecast error has the highest consequence. Build a personal verification library of cases where GraphCast diverged from GFS/IFS and the observed truth, categorized by synoptic regime. This case knowledge lets you weight AI guidance appropriately when the current pattern resembles a historical failure mode for ML models.

AI is sitting alongside you hereDevelop and validate AI weather model postprocessing and bias correction systems — training statistical and ML-based Model Output Statistics (MOS) systems that adjust raw AI NWP guidance for local biases specific to a station or region

Develop and validate AI weather model postprocessing and bias correction systems — training statistical and ML-based Model Output Statistics (MOS) systems that adjust raw AI NWP guidance for local biases specific to a station or region; evaluating AI model performance against archived observations using standard NWS verification metrics (Brier score, CRPS, track error MAE); designing hybrid forecast systems that blend physics-based and ML-based model output optimally for a specific weather element and geographic domain; and contributing to ECMWF/WMO interoperability standards for AI forecast product exchange.[10],[1]

Where your edge is

The operationalization of AI NWP guidance requires systematic bias characterization — GraphCast and AIFS have known biases for specific weather regimes (deep winter cyclones, tropical convection diurnal cycle, orographic precipitation) that MOS and postprocessing calibration can partially correct. This is a high-value technical skill in 2025-2026: national meteorological services and private weather companies both need meteorologists who can build and evaluate AI postprocessing systems rather than simply consuming AI output. Invest in Python-based statistical learning (scikit-learn, PyTorch) and NWP verification toolkits (METplus) — the meteorologist who can quantify AI model uncertainty and build calibration corrections for their region is increasingly the most technically valuable forecaster at any NWS office or private weather firm.

AI is sitting alongside you hereAnalyze satellite remote sensing data for operational weather monitoring and research — processing GOES-R series (GOES-16/18/U) Level 1b radiance data through AI-based retrieval algorithms for cloud-top height, fire detection (FDC product), derived motion winds, and total precipitable water

Analyze satellite remote sensing data for operational weather monitoring and research — processing GOES-R series (GOES-16/18/U) Level 1b radiance data through AI-based retrieval algorithms for cloud-top height, fire detection (FDC product), derived motion winds, and total precipitable water; interpreting polar-orbiting (JPSS VIIRS, SNPP) microwave and infrared sounder retrievals for atmospheric profiling; applying convolutional neural network classifiers for tropical cyclone intensity estimation (Dvorak technique AI enhancement) and severe convective mode classification from satellite imagery.[14],[1],[12]

Where your edge is

GOES-R AI retrieval algorithms have automated the production of dozens of real-time meteorological products that previously required manual analyst interpretation — fire detection, cloud-top temperature, derived winds, and lightning density all arrive as pre-processed AI products rather than raw radiance data. Your scientific judgment is essential for two tasks: (1) diagnosing retrieval failures in challenging scenes (high solar zenith angle, multi-layer cloud, smoke aerosol contamination of fire detection algorithms) that produce systematic errors in AI products you are using downstream; and (2) integrating satellite data with ground truth observations and NWP output to assess whether a satellite-detected signal represents a real meteorological or environmental event. The GOES-U generation (launched 2024) adds a new solar imagery channel — learning its AI product suite is a near-term skill investment.

Where this role is heading

Natural next steps for someone with your foundation: not exits, evolutions.

A direction you could grow

Natural Sciences Managers

Senior atmospheric scientists at NOAA/NWS program offices, national labs (NCAR, PNNL, ORNL), and private weather companies naturally transition into Natural Sciences Manager roles — directing research programs, managing forecasting service operations, and overseeing the transition from research to operational AI weather systems. The WMO Early Warnings for All initiative, the NOAA Weather-Ready Nation program, and the expanding private climate risk market all require managers who combine meteorological domain authority with program management and AI strategy capability. The transition to management offers a meaningful CRI uplift because managerial roles are structurally more insulated from automation than individual-contributor forecasting — the AI NWP disruption affects the production forecaster more directly than the program director who manages the forecasting operation's AI adoption roadmap. Transition difficulty is Medium because program management requires acquiring budget, contracting, personnel evaluation, and stakeholder communication skills not typically developed in operational forecasting or research careers.

What you'd add
  • · Federal program management: NOAA and NSF grant and cooperative agreement management; CPIC (Capital Planning and Investment Control) for IT and AI system procurements; federal acquisition basics for weather AI vendor contracts; GPRA program performance metrics for reporting to Congress
  • · AI strategy and technology transition: structuring research-to-operations (R2O) pathways for AI weather models from NSSL/ESRL research to NWS operational implementation; AI system evaluation frameworks (forecast verification standards, operational acceptance testing); vendor assessment for private AI weather model procurements
  • · Personnel and team development: NWS WFO workforce planning; mentoring forecasters on AI tool adoption; performance evaluation for hybrid AI+human forecast operations; managing interdisciplinary teams across meteorologists, software engineers, and ML researchers
  • · Stakeholder and policy communication: congressional testimony and legislative affairs for NOAA programs; media spokesperson training for high-impact weather events; inter-agency coordination (FEMA, FAA, DOT) for multi-hazard weather response
  • · Budget management: PPBE (Planning, Programming, Budgeting, and Execution) cycle at a federal agency or R&D lab; cost-benefit analysis for AI NWP infrastructure investments; competitive grant writing for NOAA Climate Program Office and NSF Atmospheric Sciences
What it takesSome new skills to pick up
Share this year
Drops anyone you send it to straight into 2026.
Preview card
Part of Science · see all 27roles →
Different role?

See the same long-arc view for your own profession.

Browse the directory by industry, or search by title or SOC code. New roles ship every few weeks. Every profile cites every claim.

Browse all roles

The data behind this timeline

On record since1870
Latest tracked employment9,400 (US, 2024)
Latest median pay$97,450 (2024)
Outlook+0.7% by 2034 (BLS National Employment Matrix 2024-34)
View all 26 cited data points
YearUS employmentMedian annual paySource
1940400$2,200ESTIMATE
19456,400n/aESTIMATE
1980n/a$28,000ESTIMATE
20008,000n/aESTIMATE
20036,490$65,670BLS-OEWS
20047,070$70,100BLS-OEWS
20057,050$73,940BLS-OEWS
20068,250$77,150BLS-OEWS
20078,750$78,390BLS-OEWS
20088,860$81,290BLS-OEWS
20098,320$84,710BLS-OEWS
20108,640$87,780BLS-OEWS
20119,640$89,790BLS-OEWS
201210,190$89,260BLS-OEWS
201310,800$87,030BLS-OEWS
201410,850$87,980BLS-OEWS
201510,370$89,820BLS-OEWS
20169,800$92,460BLS-OEWS
20178,940$92,070BLS-OEWS
20189,310$94,110BLS-OEWS
20199,290$95,380BLS-OEWS
202010,210$99,740BLS-OEWS
20218,520$94,570BLS-OEWS
20229,900$83,780BLS-OEWS
20239,310$92,860BLS-OEWS
20249,400$97,450BLS-OEWS
Embed this timeline on your site

Free for any site. Paste this where the timeline should appear; it stays interactive, every datapoint stays cited, and it sets no cookies on your page. How embedding works

<iframe src="https://futurehistory.earth/embed/19-2021"
  width="100%" height="430" style="border:0"
  title="Atmospheric and Space Scientists, a Future History timeline"
  loading="lazy"></iframe>

See all roles in Science