Atmospheric and Space Scientists
19-2021.00Median wage $99,070/yr10,000 employed (US)Rank #158 of 923 scored · top 17% by substitution
Investigate atmospheric phenomena and interpret meteorological data, gathered by surface and air stations, satellites, and radar to prepare reports and forecasts for public and other uses. Includes weather analysts and forecasters whose functions require the detailed knowledge of meteorology.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution41
Exposure38
Augmentation74
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
27 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
11%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%37
panel mean rating 2.5/5 → substitution pressure 37/100
Technical feasibility todayw 20%40
panel mean rating 2.6/5 → substitution pressure 40/100
Cost vs. human wagew 15%40
panel mean rating 2.6/5 → substitution pressure 40/100
Adoption barriersw 20%inverted — strong barriers lower the score49
panel mean rating 3.0/5 (barrier strength) → substitution pressure 49/100
Sector adoption velocityw 10%42
panel mean rating 2.7/5 → substitution pressure 42/100
Task breakdown (27 tasks)
Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.
Gather data from sources such as surface or upper air stations, satellites, weather bureaus, or radar for use in meteorological reports or forecasts.
74CI 57–91 · exposure 64 · augmentation 75 · importance 4.0/5 · click for rater detail
Gather data from sources such as surface or upper air stations, satellites, weather bureaus, or radar for use in meteorological reports or forecasts.
74| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | National weather services, commercial forecast firms, and research institutions have all deployed automated data ingestion pipelines; this is standard practice in operational meteorology, reflecting high digitization and established tooling. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Meteorological and atmospheric science agencies have long-standing, deep automation of data collection pipelines, representing mature and pervasive adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems significantly assist meteorologists by ingesting multi-source data, flagging anomalies, and pre-processing for analysis, allowing human forecasters to focus on interpretation and decision-making rather than manual data assembly. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially enhances scientists' ability to aggregate, filter, and pre-process diverse data streams, though human oversight remains for quality control and interpretation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While data ingestion from automated sources (satellites, stations, radar) can be systematized, this task involves selecting appropriate sources, handling variable data quality, and contextual judgment about which data streams are relevant for specific forecasting goals—activities that current AI cannot yet fully substitute at production scale with equivalent quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Data ingestion from surface stations, satellites, and radar is highly structured and already automated via APIs, feeds, and pipelines that AI/software can orchestrate and query with minimal human intervention. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Data gathering itself carries no licensing barrier—meteorologists use public and proprietary feeds—and no law requires a human to physically collect the data, though institutions typically retain human oversight of source integrity and anomaly detection. |
| Adoption barriers | claude-sonnet-5 | 1/5 | Data gathering is a technical/logistical task with no licensing or liability requirement tying it to a human; it's already largely automated infrastructure. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated data pipelines and cloud-based ingestion are now substantially cheaper than manual data collection; once configured, marginal cost per forecast cycle is minimal compared to a full-time meteorologist's labor. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated data feeds and API-based ingestion cost a tiny fraction of manual data gathering, offering order-of-magnitude savings over human-performed collection. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed systems routinely ingest and aggregate satellite, radar, and station data automatically; however, human meteorologists still review and validate source selection and data anomalies, so end-to-end autonomous execution without oversight remains partial. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Automated meteorological data collection systems (NOAA, ECMWF, commercial weather platforms) already aggregate multi-source observational data reliably in production at scale. |
Prepare weather reports or maps for analysis, distribution, or use in weather broadcasts, using computer graphics.
72CI 70–75 · exposure 75 · augmentation 100 · importance 3.9/5 · click for rater detail
Prepare weather reports or maps for analysis, distribution, or use in weather broadcasts, using computer graphics.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Major weather services, broadcasters, and digital platforms are actively deploying AI-assisted graphics and report generation. The information and media sectors show rapid adoption of these tools in production workflows. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | National weather services and broadcast media have already widely adopted automated graphics generation and are increasingly integrating AI-driven visualization tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments meteorologists' productivity by auto-generating drafts, applying graphics, and handling routine formatting, freeing expert time for interpretation, unusual weather events, and high-stakes forecasts. This is a textbook augmentation scenario. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-assisted graphics tools dramatically speed up map creation and formatting, letting scientists focus on interpretation and forecasting judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Modern AI systems can generate weather maps, apply computer graphics, and format reports with high efficiency. However, meteorological interpretation and selection of which data to emphasize still benefits from human judgment, limiting full end-to-end automation to ~70–80% time savings rather than the full 100%. |
| Task automatability | claude-sonnet-5 | 4/5 | Generating weather maps and reports from model output is largely a data visualization and templated summarization task that current AI and automated systems already handle well, though final quality checks still add human time. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Weather broadcasts and official forecasts often require a credentialed meteorologist's sign-off for liability and regulatory reasons (NWS standards, broadcast licensing). This oversight requirement and customer trust in human expertise create moderate friction against full substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement mandates a human create the graphic, though official forecasts may require meteorologist sign-off for liability and public safety reasons. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven weather graphics and report generation cost a fraction of employing meteorologists for routine visualization and formatting tasks. Cloud-based inference and templated systems make per-report costs orders of magnitude lower than human labor. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated map/report generation is far cheaper per instance than manual preparation once systems are set up, though initial integration and data pipeline costs are non-trivial. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed weather visualization tools and AI-assisted graphics systems are in production at major weather services and broadcast operations. Systems reliably generate maps and basic reports, though human meteorologists typically review and edit final products for accuracy and context. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated weather graphics and report generation systems (e.g., NWS automated products, broadcast graphics software) are already deployed at scale in meteorological agencies and media. |
Develop or use mathematical or computer models for weather forecasting.
71CI 46–95 · exposure 67 · augmentation 88 · importance 4.5/5 · click for rater detail
Develop or use mathematical or computer models for weather forecasting.
71| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Weather forecasting has been among the earliest and deepest adopters of automated AI/ML systems globally; operational NWP models, ensemble systems, and machine-learning post-processing are standard in nearly all national meteorological services and commercial forecast providers. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Major meteorological agencies and tech companies are actively piloting AI weather models, but widespread operational reliance is still emerging rather than fully mainstream. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI model outputs directly augment meteorologist productivity by providing rapid baseline forecasts, ensemble guidance, and pattern recognition that humans refine and interpret; meteorologists use AI outputs to accelerate scenario analysis, warning issuance, and specialized forecasts while maintaining decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI models significantly speed up simulation, pattern detection, and ensemble generation, substantially boosting scientist productivity while humans retain oversight and interpretation roles. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Weather forecasting models are already highly automated end-to-end. AI systems can ingest data, run numerical weather prediction models, post-process outputs, and generate forecasts with minimal human intervention, easily exceeding 50% time savings while maintaining or improving forecast quality compared to manual approaches. |
| Task automatability | claude-sonnet-5 | 2/5 | Model development and forecasting rely heavily on physical simulation, domain expertise, and validation against novel atmospheric conditions that current AI cannot fully replace end-to-end, though AI can accelerate certain sub-tasks like code generation or data preprocessing. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Meteorologists retain significant interpretive authority over forecast communication and specialized applications (aviation, severe weather), and some jurisdictions prefer human sign-off on official forecasts, creating modest friction; however, the underlying model development and data assimilation is routinely automated without legal barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI use, but institutional validation standards, safety-critical nature of forecasts, and reliance on established meteorological authorities create moderate organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | The infrastructure cost per forecast run is a tiny fraction of what human meteorologists would cost to manually construct equivalent predictions; cloud-based and on-premise model inference is orders of magnitude cheaper than professional labor for routine forecasting. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-based forecasting models can run inference far cheaper than traditional NWP once trained, but development, training compute, and integration with existing infrastructure keep overall costs comparable to human-supervised systems currently. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature production systems like those deployed by NOAA, meteorological agencies worldwide, and private weather services (e.g., Weather.com, commercial NWP models) demonstrably perform automated weather forecasting reliably at scale with well-established benchmarks and operational performance metrics. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI weather models (e.g., GraphCast, Pangu-Weather) are deployed experimentally and show strong skill on certain metrics, but operational centers still rely on traditional NWP with AI as a complement rather than full replacement. |
Interpret data, reports, maps, photographs, or charts to predict long- or short-range weather conditions, using computer models and knowledge of climate theory, physics, and mathematics.
69CI 46–92 · exposure 67 · augmentation 88 · importance 4.4/5 · click for rater detail
Interpret data, reports, maps, photographs, or charts to predict long- or short-range weather conditions, using computer models and knowledge of climate theory, physics, and mathematics.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Adoption is rapid and deep in the information and professional services sectors; major weather and climate organizations have begun operationalizing AI forecasting models, with national meteorological services and private weather firms actively deploying these systems in production. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Government agencies (NOAA, NWS) and companies are actively piloting AI-based weather models, but full production reliance remains partial as human-in-the-loop verification is still standard practice in this specialized science sector. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems significantly augment meteorologists by automating data interpretation, generating initial forecasts, and highlighting anomalies, allowing experts to focus on high-stakes decisions, scenario analysis, and exception-handling where domain expertise adds value beyond the model's output. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI and machine learning models dramatically speed up data interpretation, pattern recognition, and ensemble forecasting, significantly boosting scientists' productivity while they retain interpretive and decision-making roles. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Current AI systems can replicate much of this task end-to-end: machine learning models trained on historical weather data reliably generate forecasts, interpret charts and maps, and apply physics-based models with >50% time savings on data interpretation and initial prediction pipelines. Tools like neural weather models (e.g., GraphCast, Pangu-Weather) now match or exceed human meteorologist skill on many prediction horizons. |
| Task automatability | claude-sonnet-5 | 2/5 | Core weather prediction relies on established numerical weather prediction (NWP) models run on supercomputers, but interpreting outputs, integrating diverse data sources, and making judgment calls under uncertainty still requires substantial human expertise not fully replicable by off-the-shelf AI today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While meteorologists are employed by government and private agencies, there is no regulatory requirement that a licensed human must personally perform weather prediction—the task outputs (forecasts) are the regulatory concern, not the performer. Institutional preference for human oversight exists but is not a hard legal barrier. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement mandates a human forecaster for most contexts, though critical warnings (e.g., severe weather alerts) often still require human sign-off due to liability and public safety concerns. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Once trained, inference cost for AI weather models is negligible (pennies per forecast) compared to the loaded cost of a meteorologist (>$100k annually), delivering orders of magnitude savings even accounting for integration and expert oversight. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI models can generate forecast data cheaply once trained, but the specialized compute infrastructure, data pipelines, and required human interpretation keep total costs roughly comparable to skilled scientist labor for now. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Weather prediction systems are deployed at scale in production by national meteorological services (NOAA, ECMWF, Met Office) and private weather firms. AI-driven weather models have demonstrated reliable performance in real operations, with many institutions already replacing traditional ensemble models with machine learning-based alternatives. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI-enhanced forecasting products (e.g., Google's GraphCast, IBM/NASA models) exist and show promising skill in production-adjacent settings, but human forecasters still review and adjust outputs for operational and high-stakes forecasts, so full reliability without oversight isn't yet standard. |
Perform managerial duties, such as creating work schedules, creating or implementing staff training, matching staff expertise to situations, or analyzing performance of offices.
59CI 30–87 · exposure 58 · augmentation 75 · importance 3.3/5 · click for rater detail
Perform managerial duties, such as creating work schedules, creating or implementing staff training, matching staff expertise to situations, or analyzing performance of offices.
59| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | HR tech and workforce management solutions are already widely deployed in large research institutions, government agencies, and private firms. Adoption is rapid in information-sector organizations (which employ many atmospheric scientists). Public data shows strong uptake of HR automation and analytics tools across these sectors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric/space science offices (often government or academic) are not fast adopters of AI-driven management tools; broader managerial AI adoption is still nascent even in more digitized sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists managers in this domain by automatically generating schedule drafts, pulling training content, flagging expertise-skill gaps, and auto-generating performance summaries. Managers remain in control of final decisions while their productivity per task increases substantially through AI-powered templates, recommendations, and dashboards. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with drafting schedules, summarizing staff performance data, and suggesting training content, improving efficiency while the manager retains decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | This task is largely administratively repetitive—scheduling, training content generation, expertise matching, and performance analysis. Current AI systems (LLMs, workflow automation tools, and analytics platforms) can handle all these elements end-to-end with significant time savings: scheduling can be automated with constraint solvers, training materials generated or curated by LLMs, staff-expertise matching performed via embeddings/matching algorithms, and performance dashboards/reports built by BI tools. A 50% time saving at equal quality is readily achievable. |
| Task automatability | claude-sonnet-5 | 2/5 | Managerial tasks like scheduling and performance analysis have components AI can assist with (drafting schedules, summarizing metrics), but matching expertise to situations and staff training design require contextual judgment and interpersonal knowledge AI cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers prevent automating these administrative duties. The main friction is organizational (manager resistance, desire to maintain personal oversight of staff), and some preference for human judgment in sensitive staffing decisions. However, nothing legally requires a human manager to personally create schedules or dashboards. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement blocks AI use, but organizational trust, HR policy, and accountability for personnel decisions create moderate friction against full automation of managerial judgment calls. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Scheduling, training delivery, and performance reporting are labor-intensive manual tasks for managers. SaaS HR platforms and BI tools cost a fraction of even one FTE manager's salary, and LLM-based assistants add negligible incremental cost once deployed. AI systems are at least an order of magnitude cheaper than the fully loaded cost of manual managerial labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While scheduling tools are cheap, the judgment-heavy components (training design, performance evaluation, personnel matching) still require significant human oversight, keeping all-in cost comparable to or only modestly cheaper than a manager's time. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products exist for most components: Workday and SuccessFactors handle scheduling and training workflows at scale; analytics platforms (Tableau, Power BI) are standard for performance reporting; and AI-driven HR analytics tools are in production use. However, end-to-end integration and the requirement to 'match staff expertise to situations' with nuance remains partially manual in most organizations, preventing a full 5. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Generic scheduling and HR-analytics software exist and some AI features are embedded in workforce management tools, but no deployed product reliably handles the full scope of managerial judgment described (training design, expertise-matching) in a scientific office context. |
Analyze historical climate information, such as precipitation or temperature records, to help predict future weather or climate trends.
58CI 50–66 · exposure 55 · augmentation 88 · importance 3.7/5 · click for rater detail
Analyze historical climate information, such as precipitation or temperature records, to help predict future weather or climate trends.
58| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Climate research and meteorological organizations have rapidly integrated ML and AI-driven analysis into production workflows. National weather services, universities, and climate research centers actively deploy neural network forecasts and automated trend detection, with measurable adoption in the past 3–5 years. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Atmospheric science has adopted AI/ML tools for pattern recognition and forecasting at a moderate pace, with notable production deployments in weather forecasting but slower uptake in long-term climate trend research specifically. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI substantially augments atmospheric scientists by automating data curation, running multiple model scenarios in seconds, and highlighting patterns that would take humans weeks to discover manually. Scientists remain in the loop for validation, interpretation, and policy application, while productivity on exploratory analysis rises dramatically. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially augments this task by rapidly processing large historical datasets, identifying patterns, and running predictive models, significantly speeding up analysis while scientists retain interpretive and validation roles. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate data retrieval, preprocessing, and run standard statistical/ML models on climate records to identify patterns and generate trend forecasts. However, the task requires domain expertise to interpret results, select appropriate methodologies, and communicate uncertainty—elements that still require significant human judgment and setup for each new analysis. |
| Task automatability | claude-sonnet-5 | 3/5 | AI/statistical models can process historical climate data and generate trend analyses, but the task requires domain expertise to select appropriate models, validate physical plausibility, and interpret uncertainty, limiting full automation.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Regulatory frameworks (WMO standards, NOAA guidelines) and scientific rigor expectations create some friction around validation and sign-off, but no hard legal requirement mandates that a licensed human must personally perform the analysis. Organizational and journal-publication norms favor human interpretation, creating moderate adoption friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement mandates a human perform this specific analytical task, though scientific publication and institutional accountability create some oversight expectations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference on climate datasets is relatively inexpensive compared to the loaded cost of a skilled atmospheric scientist conducting the same analysis. Once models are trained, running predictions on new data costs a fraction of human labor, though initial setup and validation still carry overhead. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Running climate models and data analysis pipelines has substantial compute and data costs; while cheaper than extensive manual analysis, it's not dramatically cheaper than a scientist using existing tools since specialized infrastructure and expertise are still required. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (climate ML libraries, weather services using neural networks, ensemble forecast systems) reliably perform trend analysis and prediction on historical climate data in production. Some error rates and model limitations exist, but the capability is demonstrably deployed at scale by meteorological agencies and climate research institutions. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Machine learning and statistical climate models are deployed operationally (e.g., ECMWF's AI-based forecasting, NOAA tools), but human scientists still curate, validate, and interpret outputs rather than relying on fully autonomous systems. |
Estimate or predict the effects of global warming over time for specific geographic regions.
57CI 30–84 · exposure 58 · augmentation 88 · importance 3.4/5 · click for rater detail
Estimate or predict the effects of global warming over time for specific geographic regions.
57| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Climate research and weather services (high-digitization, information-heavy sectors) are rapidly adopting AI-assisted forecasting and impact modeling, with ML downscaling, emulators, and automated scenario generation now in routine use at major research centers and national meteorological agencies. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and climate research are adopting ML tools for specific subtasks (emulation, pattern detection) but broad production adoption replacing physical modeling pipelines remains slow and pilot-stage. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments scientist productivity by rapidly generating and visualizing regional scenarios, sensitivity analyses, and uncertainty bounds, allowing researchers to focus on interpretation, stakeholder communication, and policy-relevant insights rather than computational workflows. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI/ML significantly aids scientists in data analysis, pattern recognition, model emulation, and literature synthesis, meaningfully speeding up parts of the prediction workflow while humans retain interpretive and modeling control. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Climate modeling and projection is highly computational and data-driven; current AI systems can integrate observational data, run ensemble forecasts, and produce regional climate impact summaries end-to-end with significant time savings compared to manual analysis, meeting the ≥50% threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | Climate prediction requires running complex physical models, calibrating against observational data, and interpreting uncertainty in ways current AI cannot fully replace end-to-end; AI can accelerate parts (e.g., data processing, statistical downscaling) but not the core scientific modeling and validation.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No legal licensing requirement mandates human sign-off on climate projections; regulatory and policy contexts vary, but the core technical task itself faces minimal legal or organizational adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement, but scientific credibility, peer review, and institutional trust in climate projections used for policy create meaningful friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven climate projection pipelines can process vast datasets and generate regional forecasts at a fraction of the cost of traditional ensemble runs and manual interpretation, making computational cost several times cheaper than dedicated human scientist-years. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Running and validating climate models still requires substantial compute and expert oversight comparable to or exceeding human scientist costs; AI reduces some labor but doesn't yet cut the overall cost by an order of magnitude for rigorous regional projections. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature climate modeling platforms (including those leveraging ML for downscaling and impact assessment) are deployed in research institutions and government agencies; however, some uncertainty quantification and validation steps still require expert oversight, preventing a full 5. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI-based emulators and ML-assisted climate downscaling tools exist in research settings, but production-grade regional climate projections still rely on physics-based GCMs/RCMs with human scientist oversight, not deployed AI products doing this autonomously. |
Develop computer programs to collect meteorological data or to present meteorological information.
53CI 51–55 · exposure 50 · augmentation 75 · importance 4.0/5 · click for rater detail
Develop computer programs to collect meteorological data or to present meteorological information.
53| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Scientific and research institutions are steadily adopting AI coding assistants, but adoption is primarily for augmentation rather than replacement. Deployment is middling—pilots and early adoption are common in academic and government labs, but deep, systematic replacement of human programming in this domain is not yet widespread. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Software development broadly has seen fast AI coding-assistant adoption, but atmospheric science as a niche scientific/government sector adopts general-purpose coding AI at a moderate pace tied to institutional IT policies. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI coding tools substantially boost productivity for meteorological software development: they accelerate boilerplate generation, suggest data-handling patterns, and help debug. Scientists remain firmly in the loop for scientific design and validation, making this a strong augmentation scenario where AI transforms velocity without removing human judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI coding assistants substantially speed up writing, debugging, and documenting programs for data collection and visualization, with the scientist retaining control over domain-specific logic and validation. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | Portions of this task—particularly code generation for data collection pipelines and visualization—can be significantly assisted or partially automated by modern AI coding tools (Copilot, ChatGPT). However, the scientific logic of what meteorological data to collect and how to validate it for domain-specific requirements typically requires human expertise, preventing full end-to-end automation at equal quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI coding assistants can generate substantial portions of data collection scripts and visualization dashboards, but integrating with specialized meteorological data formats, APIs, and quality-control pipelines still requires domain expertise and iterative human refinement. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | There are no hard licensing or regulatory barriers preventing AI-assisted code development for meteorological systems. However, organizations often require human sign-off on code quality and scientific validity, and institutional preference for human-reviewed solutions creates some friction to full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement governs writing internal software tools, though organizational standards and data integrity requirements in scientific/government agencies create some friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While AI coding assistants are cheap per inference, the task still requires a skilled atmospheric scientist to architect, validate, and maintain the system. The human labor cost dominates; AI reduces iteration time but doesn't remove the need for the expert, making total cost only modestly lower than pure human development. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI coding assistance reduces development time meaningfully but a domain scientist or software engineer is still needed for review, testing, and integration, keeping costs roughly comparable rather than order-of-magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI-assisted code generation tools are deployed in production and reliably help with boilerplate and common patterns, but no system fully handles the domain-specific meteorological data validation, API integration decisions, and scientific correctness verification required here without human oversight. Products exist but with material gaps in scientific context. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like GitHub Copilot and ChatGPT are used in production to write and debug scientific code, but no deployed system autonomously builds full meteorological data pipelines without expert oversight and validation. |
Speak to the public to discuss weather topics or answer questions.
51CI 30–71 · exposure 45 · augmentation 75 · importance 3.3/5 · click for rater detail
Speak to the public to discuss weather topics or answer questions.
51| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Weather agencies and media organizations have piloted AI-driven public communication tools, but large-scale replacement of human meteorologists for public engagement is still limited; adoption is emerging but not yet deeply embedded across the sector. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Government and academic meteorology organizations adopt AI slowly for public communication roles, prioritizing human trust and accountability in public-facing science communication. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can draft responses, curate relevant meteorological data, and handle routine Q&A, freeing human scientists to focus on complex public dialogue, media interviews, and crisis communication—substantially raising their productivity and allowing focus on high-value interactions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can help draft talking points, summarize data, and generate visualizations or FAQ responses, meaningfully boosting a scientist's efficiency while they remain the face of public communication. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can generate coherent public-facing weather explanations and answer factual questions about weather with significant time savings; however, the task requires nuanced audience calibration, real-time responsiveness to unpredictable questions, and occasional judgment calls that still benefit from human oversight, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 2/5 | Public-facing speaking requires real-time interaction, credibility, and adaptive communication that current AI cannot fully replicate end-to-end, though scripted Q&A support is feasible. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | There are few legal or licensing barriers to automating public weather communication; organizational preference for human credibility and brand trust provides some friction, but this is manageable rather than a hard regulatory or authorization requirement. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement bars AI from speaking, but public trust, institutional branding, and preference for a credentialed expert create real friction against substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | An AI system handling public weather communication incurs minimal marginal inference costs and requires minimal oversight once deployed, making it orders of magnitude cheaper than funding a human atmospheric scientist's time to answer repeated public inquiries. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI-generated weather summaries are cheap, live public engagement (interviews, talks, press briefings) still requires human presence and judgment, keeping cost savings limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed chatbots and language models can handle routine weather Q&A and public education at scale, but they occasionally produce inaccurate meteorological details or fail on edge cases; production systems exist but with material error rates in specialized or context-dependent scenarios. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Chatbots and voice assistants can answer basic weather FAQs, but no deployed product substitutes for a scientist engaging the public live on nuanced or emerging weather topics. |
Prepare forecasts or briefings to meet the needs of industry, business, government, or other groups.
46CI 32–59 · exposure 42 · augmentation 75 · importance 4.1/5 · click for rater detail
Prepare forecasts or briefings to meet the needs of industry, business, government, or other groups.
46| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Government and academic meteorological operations are gradually integrating AI for data processing and model ensemble analysis, but the core task of preparing client-facing forecasts and briefings remains human-led. Adoption is moderate, with pilots and hybrid approaches common but not rapid replacement. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Meteorological agencies and private weather companies are adopting AI-assisted forecasting and text generation tools, but full-scale replacement of expert briefings remains limited and cautious given safety implications. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists by rapidly analyzing large datasets, generating scenario summaries, formatting briefing templates, and highlighting anomalies. Forecasters can leverage AI to explore multiple scenarios and prepare more thorough briefings faster while retaining full decision-making authority and scientific judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up drafting of routine forecast text and briefing materials, letting scientists focus on interpretation, anomalies, and client-specific nuances. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate draft forecasts from meteorological data and produce structured briefing documents, the task requires domain expertise, client-specific insight, and judgment about uncertainty that current systems cannot fully replicate end-to-end. Significant human review and revision are necessary, limiting time savings below the 50% threshold. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft forecast summaries and briefings from weather model data, but interpreting model discrepancies, unusual events, and tailoring for specific client needs still requires expert judgment, so only partial automation meets the equal-quality bar. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory agencies (NOAA, NWS) and government clients often require forecasts and briefings to be prepared or signed off by credentialed atmospheric scientists. Liability for forecast errors and the trust placed in official briefings creates both legal and organizational barriers to full automation without human authority. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing strictly requires a human forecaster for most commercial briefings, though liability concerns for high-impact events (aviation, marine, severe weather warnings) create some friction and institutional caution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Generating baseline forecasts with AI is cheap, but the domain expertise required to translate raw forecasts into actionable, client-specific briefings with appropriate confidence qualifications demands highly trained specialists. The all-in cost of AI + human oversight often approaches or exceeds the cost of a domain expert working directly. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating draft forecasts and briefing text via AI models is very cheap compared to analyst time, though some human review keeps the ratio from reaching the extreme low end. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI tools can assist with data processing, trend analysis, and document generation for weather and space forecasts, and some operational systems use AI components. However, production systems still rely on human meteorologists and space scientists for final validation, interpretation, and custom briefing tailoring, making fully autonomous deployment unreliable. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products exist that auto-generate weather briefings and forecast narratives (e.g., automated NWS text products, AI weather summarization tools), but human meteorologists still validate and customize outputs for high-stakes or specialized audiences. |
Create visualizations to illustrate historical or future changes in the Earth's climate, using paleoclimate or climate geographic information systems (GIS) databases.
41CI 30–52 · exposure 42 · augmentation 75 · importance 3.1/5 · click for rater detail
Create visualizations to illustrate historical or future changes in the Earth's climate, using paleoclimate or climate geographic information systems (GIS) databases.
41| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Atmospheric and space science remains a relatively small, research-oriented sector with slow digital transformation compared to finance or tech. While GIS automation is advancing in water resources and environmental management, climate visualization for publication or policy use remains heavily driven by human experts, with limited production-scale AI agent deployment. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and geoscience research is a moderately-digitized but specialized academic/government sector where AI adoption for domain-specific GIS/paleoclimate visualization remains in early pilot stages rather than widespread production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments this task: machine learning can rapidly process large paleoclimate datasets, suggest visualization parameters, generate multiple scenarios, and flag anomalies, while the scientist retains control over interpretation, methodology, and publication standards. This human-in-the-loop pattern strongly accelerates workflow without replacing the expert judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially assists by automating boilerplate code, suggesting visualization types, and speeding up figure generation and data formatting, while the scientist retains control over data selection, interpretation, and validation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate visualizations from climate data and GIS databases with modern tools, the task requires domain expertise to select appropriate datasets, define meaningful temporal scales, and ensure scientific accuracy. Current AI cannot reliably navigate paleoclimate databases or validate climate GIS methodology without expert oversight, making full automation well below the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 3/5 | AI tools can generate charts, maps, and code (e.g., Python/GIS scripting) for visualizing climate data, but selecting appropriate paleoclimate datasets, ensuring scientific accuracy, and interpreting geological proxies still require substantial human expertise and validation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: paleoclimate and climate visualization require peer-reviewed scientific standards, institutional liability for accuracy in published or policy-relevant outputs, regulatory standards for climate data representation, and organizational expectation that qualified atmospheric scientists validate interpretations. Automation faces strong friction from quality assurance and reputational risk in climate science. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this specific task, but scientific credibility, peer review standards, and institutional trust in specialized paleoclimate interpretation create moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Cloud-based GIS platforms and AI visualization tools are moderately priced, but integration with specialized paleoclimate databases, quality assurance, and the expertise required to oversee output make the all-in cost competitive with or exceeding the wage of a skilled scientist-technician, especially when accounting for necessary validation. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI can reduce time spent on coding and initial visualization drafts significantly, but the specialized data wrangling, quality control, and domain expertise needed still require costly human scientist time comparable to current costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Data visualization and GIS mapping tools with AI components exist and function in production (e.g., ArcGIS with AI-assisted analysis, Python libraries with machine learning), but they require significant human judgment for paleoclimate interpretation, appropriate spatial/temporal resolution selection, and scientific validity checks. Current products handle routine mapping but not the nuanced climate science decisions embedded in this task. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI coding assistants and data visualization tools (e.g., ChatGPT with code interpreter, GIS plugins) are used in production for generating draft visualizations, but full pipelines integrating specialized paleoclimate/GIS databases with scientific rigor are not yet fully automated in deployed products. |
Broadcast weather conditions, forecasts, or severe weather warnings to the public via television, radio, or the Internet or provide this information to the news media.
39CI 30–48 · exposure 38 · augmentation 75 · importance 4.1/5 · click for rater detail
Broadcast weather conditions, forecasts, or severe weather warnings to the public via television, radio, or the Internet or provide this information to the news media.
39| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While automated alert systems are common, human meteorologists remain the primary public face of weather communication. Broadcasters and news organizations have not moved toward replacing on-air meteorologists with AI presentations, keeping adoption in the slow-to-middling range despite technical capability in forecasting. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Broadcast and media sectors are adopting AI for scriptwriting and automation but human presenters still dominate live severe weather communication, so deployment is slow in this specific niche. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augmentation is substantial: AI forecasting models, visualization tools, and automated data synthesis dramatically improve a meteorologist's ability to identify threats and prepare communications faster and more comprehensively, while the human remains the final interpreter and communicator. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can quickly draft warning scripts, translate forecasts into multiple languages, and prepare visualizations, meaningfully boosting scientist/broadcaster productivity while the human retains final delivery. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate weather forecasts and identify conditions, the task of selecting what to broadcast and delivering it with appropriate tone, urgency calibration, and public communication requires human judgment. An AI system could potentially draft or auto-generate some elements, but end-to-end automation achieving equal quality would require replacing the editorial and presentation decisions that a meteorologist makes. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can generate forecast summaries and even synthetic voice/video presentations, but live broadcasting, ad-hoc interpretation of severe weather, and public trust dynamics still require human presence for most of this task. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Public trust, regulatory frameworks (NWS coordination), and audience expectations that a credible human communicates warnings create friction, though these are not hard legal barriers. Media organizations prefer human anchors and meteorologists; automation faces organizational and reputational resistance rather than licensure blocks. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement to be a broadcast meteorologist, but public trust, liability during severe warnings, and audience preference for human communicators create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-generated weather data and automated alerting have reduced some labor, but professional meteorologists command significant salaries, and the cost of maintaining AI forecasting infrastructure plus human oversight remains comparable to or potentially higher than full human production in many markets. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated text-to-speech and templated forecast generation is cheap, but production-quality broadcast integration, oversight, and correction for severe weather nuances add cost comparable to a human presenter in many contexts. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI weather models and severe-weather-detection systems exist in production (e.g., NOAA models, automated alerts), but the full task of broadcasting or writing for news media still requires human meteorologists to interpret data, select key information, and deliver it. Automated severe-weather alerting systems are deployed, but human-led broadcast is the standard. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some AI-generated weather broadcasts and automated voice alerts exist (e.g., synthetic TV meteorologist avatars), but they remain niche and not the standard in most newsrooms or emergency systems. |
Prepare scientific atmospheric or climate reports, articles, or texts.
37CI 25–50 · exposure 38 · augmentation 75 · importance 3.8/5 · click for rater detail
Prepare scientific atmospheric or climate reports, articles, or texts.
37| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption in atmospheric science remains slow and limited to minor assist roles (drafting intros, polishing language) rather than autonomous report generation. The sector values rigorous, original analysis and maintains strong professional gatekeeping, resulting in cautious and measured uptake. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Academic and research sectors show growing but uneven adoption of AI writing tools; usage is common for drafting/editing but full pipeline automation is still limited and cautious due to accuracy concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist atmospheric scientists by generating initial drafts, summarizing literature, formatting tables, and refining prose, thereby raising productivity on writing mechanics. However, the core scientific analysis, interpretation, and novel contributions remain firmly the scientist's domain. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI writing assistants substantially speed up drafting, editing, summarizing data, and improving clarity for atmospheric scientists producing reports and articles, while the scientist retains responsibility for content accuracy. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate drafts of atmospheric reports and synthesize existing research, current systems lack the domain expertise, novelty of analysis, and scientific rigor required for publishable outputs without extensive human revision. Atmospheric scientists must interpret complex data, develop original hypotheses, and ensure accuracy—tasks where AI achieves <50% time savings due to high error rates and need for substantial reworking. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft substantial portions of technical writing, summarize data, and generate text sections, but synthesizing novel scientific findings, ensuring accuracy, and framing original interpretations still requires significant expert input and revision. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Peer review, publication standards, and professional accountability create strong barriers: journals and funding agencies require human scientist authorship and accountability for accuracy. Additionally, liability for incorrect climate or atmospheric predictions and institutional norms strongly favor human sign-off on scientific publications. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement to write reports, but scientific credibility, peer review norms, and institutional accountability for accuracy create meaningful friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference and integration costs for specialized scientific writing are modest, but the oversight required from atmospheric scientists to validate accuracy, methodology, and novel contributions means total cost per usable report approaches or exceeds the cost of direct human authorship. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI drafting reduces time spent on writing, but oversight, fact-checking, and domain-expert review remain necessary, keeping all-in costs only moderately below human-only authorship. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI writing tools exist and can produce coherent scientific text, but they frequently introduce factual errors, misrepresent data, and fail to capture nuanced interpretations needed in peer-reviewed work. No mature product reliably performs full report generation without domain expert oversight; deployment remains limited to draft-assist roles rather than autonomous production. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | LLM-based writing assistants and tools like ChatGPT/Claude are commonly used by scientists today for drafting and editing, but reliable, error-free production of full scientific reports without expert oversight is not yet standard. |
Formulate predictions by interpreting environmental data, such as meteorological, atmospheric, oceanic, paleoclimate, climate, or related information.
34CI 28–41 · exposure 30 · augmentation 88 · importance 4.3/5 · click for rater detail
Formulate predictions by interpreting environmental data, such as meteorological, atmospheric, oceanic, paleoclimate, climate, or related information.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Meteorological and climate research institutions are adopting AI-assisted analysis and hybrid forecasting models, but adoption remains primarily in support (augmentation) rather than replacement; production systems still rely on human meteorologists for final interpretation and forecasting. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Meteorological and climate science agencies are actively piloting ML-based forecasting tools, but full operational deployment replacing human interpretation remains limited and cautious. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists atmospheric scientists by rapidly processing massive datasets, identifying patterns, generating scenario simulations, and highlighting anomalies, substantially raising productivity while the scientist remains in the loop for interpretation, validation, and decision-making. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI dramatically accelerates data processing, pattern detection, and ensemble analysis, significantly boosting scientists' productivity while they retain interpretive and validation roles. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI excels at pattern recognition in environmental datasets, formulating *predictions* requires interpreting domain-specific data in context, integrating multiple data streams, validating assumptions, and communicating uncertainty—tasks where current AI systems need significant human oversight and cannot achieve ≥50% time savings at equal quality end-to-end. |
| Task automatability | claude-sonnet-5 | 2/5 | Core scientific prediction requires integrating numerical models, domain expertise, and judgment about physical processes that AI can assist with but not independently perform to expert standard end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory and professional standards in climate and weather forecasting require human meteorologists to interpret, validate, and sign off on predictions, especially for public warnings and policy decisions; liability and error-cost asymmetry create strong organizational friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement for interpretation itself, but institutional reliance on validated models, liability for public forecasts, and scientific rigor standards create moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Inference costs for large climate models are non-trivial, and the integration, validation, and expert oversight required to produce trustworthy predictions add significant expense; the total cost per prediction often approaches or exceeds the loaded wage of an atmospheric scientist. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | ML weather models can generate forecasts far cheaper than traditional supercomputing runs, but human interpretation, calibration, and validation still require significant expert oversight, balancing the cost advantage. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI-powered tools for environmental data analysis and forecasting exist in production (e.g., ML-based weather models, climate simulators), but they perform narrow subtasks with material error rates and require expert interpretation; no single deployed system reliably formulates predictions across the full scope of meteorological, oceanic, and paleoclimate domains without expert validation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI-based weather/climate models (e.g., GraphCast, FourCastNet) exist and show promise but are still largely research-stage or supplementary to operational NWP systems, not replacing expert interpretation broadly. |
Conduct numerical simulations of climate conditions to understand and predict global or regional weather patterns.
32CI 28–37 · exposure 30 · augmentation 75 · importance 3.4/5 · click for rater detail
Conduct numerical simulations of climate conditions to understand and predict global or regional weather patterns.
32| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Major climate and weather forecasting institutions (NOAA, Met Office, ECMWF) are actively piloting ML-accelerated components and surrogate models, but end-to-end replacement of human-guided simulation workflows remains rare; adoption is advancing but still largely experimental. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Atmospheric science is increasingly incorporating ML-based emulators and hybrid models, but adoption is still largely pilot/research-stage rather than fully embedded in standard operational forecasting workflows. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems meaningfully augment atmospheric scientists by accelerating model emulation, automating parameter sweeps, and highlighting anomalies or patterns in large output datasets, while human experts remain essential for model design, validation, and interpretation of results. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids in speeding up simulations, downscaling, anomaly detection, and interpreting large datasets, significantly boosting scientist productivity while humans retain oversight of model design and validation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in parts of the simulation pipeline (data preprocessing, parameter tuning, emulation), the full end-to-end task of designing, validating, and interpreting complex climate models requires substantial human expert judgment and decision-making that current systems cannot autonomously perform at parity with human specialists. |
| Task automatability | claude-sonnet-5 | 2/5 | Running climate/weather numerical simulations requires domain expertise, specialized HPC modeling frameworks, and physical understanding that AI can assist but not independently execute end-to-end at equal quality today.assistance is mainly in coding, data prep, and interpretation rather than full model design and validation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory and scientific standards (e.g., IPCC assessment criteria, forecast verification protocols) effectively require human expert review and sign-off on climate predictions used for policy or public communication; liability and accuracy expectations create strong institutional friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement, but scientific rigor, peer review, and institutional validation standards create meaningful friction against fully automated adoption. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Conducting full climate simulations remains computationally expensive and human-expert-intensive; while ML acceleration of specific components can reduce wall-clock time, the overhead of model development, validation, and interpretation by domain scientists means total cost per simulation is not yet an order of magnitude cheaper than traditional approaches. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Running and validating large-scale climate simulations still requires substantial human expertise and HPC resources; AI reduces some labor but does not yet undercut costs by an order of magnitude for this specialized scientific task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed products (e.g., neural network emulators, ML surrogates for physics modules) exist and are used in research and operational contexts, but they typically handle narrow subproblems rather than the complete simulation workflow; integration into production forecasting pipelines remains limited and requires careful validation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools exist for weather nowcasting (e.g., GraphCast, some ML weather models) but full climate simulation workflows in production still rely on traditional numerical models run and interpreted by scientists, not autonomous AI systems. |
Develop and deliver training on weather topics.
30CI 30–30 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail
Develop and deliver training on weather topics.
30| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Academic and meteorological organizations have adopted e-learning platforms and AI-assisted content tools, but training delivery itself remains predominantly human-led. Adoption of AI for autonomous training is slow relative to information/tech sectors; most institutions augment rather than replace instructors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and public-sector meteorology are not leading adopters of AI-driven training tools; adoption is more common in generic corporate training L&D functions than in scientific/technical instruction. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists training development by generating customized lesson plans, visualizations, practice problems, and explanatory content, allowing instructors to focus on delivery, interaction, and refinement. AI tools measurably enhance the productivity of trainers while they remain in control of pedagogical decisions and live instruction. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist in drafting training slides, summarizing research, generating quizzes, and creating visual aids, significantly speeding up preparation while the scientist retains delivery and subject authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate training materials and draft lesson content, but delivering effective training requires real-time interaction, gauging audience understanding, adapting explanations, and answering unexpected questions—all of which demand human pedagogical judgment. Automated systems could create slides and scripts but cannot replace the adaptive, interactive delivery that characterizes training instruction. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft training content and materials, but designing, delivering, and adapting live training on weather topics still requires subject-matter expertise, presentation skills, and audience interaction that current AI cannot fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Training delivery often occurs in academic or government contexts where stakeholders prefer qualified human instructors and may require credentialed expertise. However, there are no hard legal barriers preventing AI from assisting with or partially automating training material creation, though organizational preference for human expertise provides moderate friction. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement for training delivery itself, but organizational expectations for expert credibility, accuracy, and public trust in weather information create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-generated training materials are cheap at scale, but human meteorologists command high wages; integrating AI content generation with human delivery oversight keeps total costs comparable to—or potentially higher than—direct human instruction for specialized weather science training. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate draft materials, the need for expert review, live delivery, and audience interaction means overall costs remain comparable to or only modestly below human-led training. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools can generate training content (chatbots, LLMs drafting curricula), no deployed system reliably delivers comprehensive weather training end-to-end with the domain expertise, contextual sensitivity, and instructional design that scientists expect. Products exist for content drafting but not for autonomous training delivery at the fidelity required. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products like chatbots and content generators can assist with slide creation or Q&A support, but no deployed product independently develops and delivers full weather training curricula in production settings. |
Apply meteorological knowledge to issues such as global warming, pollution control, or ozone depletion.
30CI 28–32 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail
Apply meteorological knowledge to issues such as global warming, pollution control, or ozone depletion.
30| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Climate and atmospheric research organizations have adopted AI-assisted modeling and data analysis at moderate pace, with many pilots and increasing computational support, but the pace remains measured because core policy recommendations and scientific assessments still require human meteorologists and are not undergoing rapid displacement. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Atmospheric science and environmental research increasingly use AI/ML for climate modeling and data analysis, but adoption for applied policy-relevant judgment tasks remains at the pilot stage rather than deep production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments meteorologist productivity by enabling rapid scenario modeling, pattern detection in vast datasets, and visualization of complex atmospheric phenomena; these tools allow scientists to explore more hypotheses and refine recommendations faster while remaining in the decision-making loop. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids literature synthesis, data pattern detection, and modeling scenarios, meaningfully boosting scientist productivity while the human retains interpretive and judgment responsibilities. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in analyzing large climate datasets and modeling scenarios, applying meteorological knowledge to complex environmental issues requires integrating diverse data sources, evaluating competing theories, and making judgment calls about policy implications—tasks that demand human expertise and cannot currently be fully automated to meet the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | This task requires original scientific reasoning, synthesis of complex atmospheric data, and judgment about policy implications that current AI cannot reliably perform end-to-end; AI can assist with literature review and data analysis but not replace the scientific application of expertise. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | High barriers exist: decisions on global warming, pollution control, and ozone policy often require regulatory sign-off from human experts, carry significant liability for incorrect recommendations, and demand institutional trust in human judgment; government and research organizations mandate credentialed scientists oversee such analyses. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensure is typically required, but institutional, peer-review, and credibility standards in scientific and policy contexts create meaningful friction against pure AI-generated conclusions being accepted. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools (supercomputer time, software licenses, maintenance) for atmospheric simulation are expensive, and their deployment still requires highly-paid atmospheric scientists to interpret results and formulate recommendations, making the all-in cost remain comparable to or higher than direct human analysis. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human expert scientists remain necessary for validation and interpretation, so while AI can cut some research time, the overall cost of a credible scientific assessment is still dominated by human labor and oversight. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed products exist for climate modeling and pollution simulation, but they serve as tools requiring expert interpretation rather than autonomous systems that can apply meteorological knowledge end-to-end; no production system replaces the meteorologist's role in evaluating trade-offs and making evidence-based recommendations. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (e.g., literature summarization, climate model output analysis) exist but no deployed product independently applies meteorological expertise to novel pollution or climate policy questions at professional scientific standards. |
Conduct meteorological research into the processes or determinants of atmospheric phenomena, weather, or climate.
29CI 25–32 · exposure 25 · augmentation 75 · importance 4.3/5 · click for rater detail
Conduct meteorological research into the processes or determinants of atmospheric phenomena, weather, or climate.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Meteorological research remains concentrated in academic and government institutions (NOAA, universities) with slower AI adoption cycles compared to commercial sectors. While AI is used for operational weather forecasting, research-level adoption of autonomous AI-driven investigation is nascent. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Atmospheric science and climate research increasingly use AI/ML tools (e.g., for climate modeling, forecasting), representing middling-to-growing adoption, but full automation of research remains rare. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments meteorological research by accelerating data processing, pattern detection, and simulation runs; researchers can explore hypotheses faster and visualize complex atmospheric interactions more intuitively. The human scientist remains central but operates with substantially amplified analytical capacity. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids in data analysis, pattern recognition, simulation, and literature synthesis, meaningfully boosting researcher productivity while humans retain interpretive and design control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can process large meteorological datasets and run simulations faster than humans, the task requires forming novel research hypotheses, designing experiments, interpreting anomalies, and contextualizing findings—all of which demand human scientific judgment. AI assists with data analysis but cannot autonomously conduct the full research cycle to 50% time-saving at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Original scientific research involving hypothesis generation, novel modeling, and experimental design cannot be fully automated end-to-end; AI can accelerate specific subtasks like data analysis and literature review but not the full research process at equal quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Research integrity, peer review, and publication standards require human authorship and accountability; funding agencies and institutions mandate human expertise and decision-making in grant-funded atmospheric research. Liability for incorrect climate or weather predictions also creates organizational and regulatory friction. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement for research itself, but peer review, scientific credibility, and institutional funding structures create moderate friction against full AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference on meteorological datasets is cheap, but the integrated cost of model development, domain validation, and expert human oversight to ensure research rigor approaches or exceeds the cost of a meteorologist performing the research themselves. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply handle data processing and pattern detection, but the overall research task still requires expensive expert oversight, domain expertise, and validation, keeping costs comparable to human-led research. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product autonomously conducts meteorological research end-to-end. AI tools exist for data analysis and model post-processing, but they operate within human-designed frameworks; production systems require meteorologists to frame questions, validate outputs, and interpret results. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (e.g., ML weather models, literature search assistants) are used in production for specific components, but no deployed product independently conducts full meteorological research programs reliably. |
Analyze climate data sets, using techniques such as geophysical fluid dynamics, data assimilation, or numerical modeling.
28CI 25–30 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail
Analyze climate data sets, using techniques such as geophysical fluid dynamics, data assimilation, or numerical modeling.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption is slow in academic and government climate science; these sectors prioritize methodological rigor and human accountability, with AI largely confined to computational acceleration rather than autonomous decision-making. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science is a specialized research-heavy field with slower AI tool adoption compared to fast-moving sectors like finance or general professional services, though ML weather/climate modeling is a growing niche. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments atmospheric scientists through automated data ingestion, computational acceleration of fluid dynamics simulations, and visualization of complex datasets, allowing scientists to focus on hypothesis generation and interpretation rather than routine computation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids in processing large datasets, running simulations faster, detecting patterns, and assisting with coding for numerical models, meaningfully boosting scientist productivity while they remain central to interpretation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in numerical modeling and data processing steps, the task requires domain expertise to select appropriate techniques, validate assumptions, and interpret results critically—human oversight remains essential for high-stakes climate analysis. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with data processing, pattern detection, and code generation for numerical models, but the core scientific analysis of complex climate datasets requires domain expertise, judgment, and validation that current AI cannot autonomously perform at equal quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: domain credibility requires human scientists to validate findings, peer review mandates human responsibility for methodology, and liability for climate projections feeding policy demands expert human sign-off on outputs. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement per se, but scientific rigor, peer review, and institutional trust in model outputs create moderate friction against pure AI substitution for interpretive analysis. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI tools reduce computational costs for some modeling components, but the specialized expertise required, high oversight demands, and need for custom integration make the all-in cost of AI-assisted analysis comparable to or higher than hiring domain experts. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Running large numerical models and geophysical fluid dynamics simulations requires substantial compute and specialized oversight, so AI-assisted analysis is not necessarily cheaper than skilled scientist labor when accounting for validation and interpretation costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed products exist for individual components (data processing, visualization, basic modeling), but no production system autonomously performs full end-to-end climate data analysis at the quality required for published research or policy decisions. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | There are research tools (e.g., ML-based climate emulators, GraphCast-style models) but these are not yet standard deployed products replacing scientists' analytical workflows in operational climate science. |
Consult with other offices, agencies, professionals, or researchers regarding the use and interpretation of climatological information for weather predictions and warnings.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.7/5 · click for rater detail
Consult with other offices, agencies, professionals, or researchers regarding the use and interpretation of climatological information for weather predictions and warnings.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Meteorological agencies and research centers have been slow to automate expert-to-expert consultation; while weather prediction algorithms are widely deployed, the interpersonal consultation and interpretation aspect remains human-centric. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and government meteorological agencies have historically slower AI adoption for interpretive/advisory functions compared to fully digitized sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by summarizing climatological datasets, highlighting anomalies, and drafting briefing materials for human consultation, but the collaborative judgment and interpretation remain with the experts. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help synthesize climatological data, generate summaries, and prepare briefing materials, enhancing the scientist's ability to consult effectively. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can retrieve and summarize climatological information, the core task requires interpersonal consultation, judgment synthesis across multiple expert perspectives, and contextual interpretation for specific operational decisions that AI systems cannot reliably perform end-to-end today. |
| Task automatability | claude-sonnet-5 | 2/5 | This task requires interactive, contextual professional dialogue and judgment-based interpretation of climate data with stakeholders, which current AI can support but not conduct autonomously end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: weather warnings carry legal/liability implications, regulatory frameworks require qualified meteorologists to sign off on forecasts, and organizational protocols mandate human expert review before public communication. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While not strictly licensed, inter-agency coordination on weather warnings often involves institutional trust, liability for public safety warnings, and established professional protocols. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI systems capable of meaningful climatological interpretation plus human oversight and validation would likely approach or exceed the cost of direct expert consultation, given the high stakes of weather warnings. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human expert consultation still dominates due to need for trust, accountability, and nuanced judgment; AI can lower prep costs but not replace the interaction itself. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product systematically handles multi-party expert consultation and interpretation of climatological data for operational weather decisions; LLMs can draft talking points but cannot replace the collaborative expert assessment this task demands. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product independently consults across agencies on climatological interpretation; existing tools assist analysts but don't replace the expert consultative role. |
Research the impact of industrial projects or pollution on climate, air quality, or weather phenomena.
28CI 25–30 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail
Research the impact of industrial projects or pollution on climate, air quality, or weather phenomena.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Academic and governmental research institutions are adopting AI for specific subtasks (climate modeling, data processing) but not for the strategic research design and interpretation work; adoption remains slow in this domain. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and environmental research sectors are adopting AI/ML tools in specific areas (climate modeling, remote sensing) but broad production-scale replacement of researchers remains limited and slow given the specialized, publicly funded nature of the field. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems today meaningfully assist by accelerating data analysis, literature review, climate model execution, and visualization, substantially raising researcher productivity while keeping humans in control of framing, validation, and conclusions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids literature reviews, data analysis, simulation acceleration, and pattern detection in large datasets, meaningfully boosting researcher productivity while humans retain interpretive and methodological control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with data analysis, modeling, and literature synthesis, this task requires formulating research questions, designing studies, interpreting results in context, and making judgments about causation and impact—activities that demand human scientific expertise and cannot be fully automated to meet the 50% time-saving bar today. |
| Task automatability | claude-sonnet-5 | 2/5 | This research requires original data collection, hypothesis generation, model design, and interpretation of complex physical systems, which current AI cannot fully replicate end-to-end; AI can accelerate literature review, coding, and data analysis but not the full research task. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | This research informs policy, regulation, and litigation; outputs must be defensible and often require credentialed scientists to author and sign off on findings, creating legal and institutional barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement to publish research, but regulatory and policy-relevant findings often require credentialed scientists, peer review, and institutional accountability, creating moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI systems require expensive infrastructure (compute for modeling, data access, expert human review of outputs), and this task demands domain expertise that cannot yet be cost-effectively replaced; human researchers remain far more efficient at the core research function. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Specialized climate/atmospheric modeling still requires substantial computational and human expert oversight; AI reduces some labor costs (literature synthesis, data wrangling) but doesn't yet approach order-of-magnitude savings for the full research task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably executes end-to-end atmospheric impact research independently. AI tools exist for data processing and literature review, but the synthesis, hypothesis generation, and validation of causal claims remain research-stage or require substantial human oversight. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (e.g., climate models, ML-based emissions analysis) exist but are used as components within human-led research pipelines, not as standalone products producing peer-reviewed environmental impact assessments. |
Conduct wind assessment, integration, or validation studies.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.5/5 · click for rater detail
Conduct wind assessment, integration, or validation studies.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Energy and atmospheric science sectors show moderate digitization but slow AI adoption for core technical tasks. Most organizations still rely on established methodologies, human expert teams, and domain-specific tools rather than AI-driven automation of wind studies. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science is a specialized research-oriented field with moderate digitization; AI adoption exists in specific modeling and forecasting tools but broad production-level automation of assessment/validation studies is still uncommon. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist with data processing, statistical analysis of wind patterns, and visualization of modeling results, improving efficiency in data-heavy phases. However, site selection, field interpretation, and final validation remain human-led, making augmentation useful but not transformative across the full task. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and machine learning tools significantly assist in data processing, pattern recognition, and preliminary modeling for wind studies, meaningfully boosting scientist productivity while humans retain interpretive and validation responsibilities. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Wind assessment involves complex field measurement, modeling, and interpretation of atmospheric data that requires domain expertise and judgment. While AI can assist with data processing and some modeling tasks, the full end-to-end work—site evaluation, sensor placement, integration with existing systems, and validation of results—relies heavily on human expertise and field knowledge that AI cannot autonomously perform at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Wind assessment involves complex data collection, sensor calibration, physical modeling, and interpretation of atmospheric dynamics that current AI cannot fully perform end-to-end without significant human expertise and judgment.atmospheric wind study data. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Wind assessment results inform major infrastructure and energy policy decisions with significant financial and safety implications. Engineering sign-off, liability, regulatory compliance (siting, environmental impact), and stakeholder trust in human expertise create strong adoption barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement mandates a human scientist per se, but organizational and scientific rigor standards, peer review, and use in regulatory/safety contexts (e.g., aviation, wind energy siting) create meaningful friction against pure automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The specialized nature of wind assessment, combined with the need for field work, domain expertise, and integration with existing tools, means AI tools do not yet achieve cost parity with human specialists when all overhead is included. Human wind engineers command high wages, but AI cannot yet replace their full output at lower cost. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can reduce computational costs for certain modeling components, the overall task still requires substantial human oversight, specialized instrumentation, and domain expertise, keeping costs comparable to or only modestly below human-only approaches. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Products exist for narrow components (wind resource modeling, data analysis) but no deployed system reliably performs the full assessment-to-validation pipeline end-to-end. Most wind assessment remains anchored to specialized domain software and human-led site surveys rather than AI-driven automation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some AI/ML tools exist for wind forecasting and data analysis in production (e.g., weather modeling systems), but full validation studies requiring domain expertise and integration of multiple physical models are not reliably automated by deployed products. |
Direct forecasting services at weather stations or at radio or television broadcasting facilities.
26CI 25–28 · exposure 25 · augmentation 75 · importance 4.1/5 · click for rater detail
Direct forecasting services at weather stations or at radio or television broadcasting facilities.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Broadcast and weather station sectors have adopted AI-assisted forecasting tools and automated model output, but adoption of AI-directed forecasting services remains limited; most stations still employ human meteorologists to curate and direct forecasts. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Meteorological and broadcasting sectors have adopted AI for data crunching and visualization but adoption of AI to direct or manage forecasting operations is minimal and mostly experimental. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augments forecasters substantially by providing real-time model ensembles, pattern detection, and automated data processing, which measurably increases productivity in selecting and communicating the most reliable forecasts while the meteorologist retains final authority and interpretation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially augments the underlying forecasting work (model outputs, data visualization, alerts) that a director oversees, improving decision speed and accuracy while the human retains directional control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Weather forecasting itself is increasingly automated via numerical weather prediction models, but directing forecasting *services* involves coordination, communication, and decision-making about which forecasts to use, how to present them, and managing station operations—tasks requiring human judgment that current AI cannot fully perform end-to-end with 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Directing forecasting services involves managerial oversight, coordination of staff, and real-time judgment calls that AI cannot autonomously perform end-to-end today, though AI can assist with underlying data analysis.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Directing forecasting services at broadcast facilities involves regulatory compliance (FCC rules, public safety communication standards), professional liability for incorrect forecasts, and contractual expectations that a qualified meteorologist oversee output, creating legal and organizational friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Directing forecasting services often carries institutional accountability (e.g., NWS certification, broadcast licensing context) and requires human judgment and liability for public safety communications, creating strong organizational and regulatory barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI weather model inference is cheap, but integrating AI into station operations, maintaining broadcast-quality output, and ensuring liability compliance still requires substantial human expertise and oversight, making the total cost per service comparable to or higher than a human forecast director. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Directing a team/service still requires a human manager; AI tools may reduce some data-processing costs but the managerial role itself remains costly to replace with equivalent oversight and accountability. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can generate forecasts and assist with data analysis, no deployed system reliably directs the full suite of forecasting services (choosing models, communicating to media outlets, managing broadcast schedules, quality control) without human oversight at production scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI products exist for weather modeling and forecast generation, but no deployed product directs or manages forecasting operations or broadcasting facility staff in production. |
Teach college-level courses on topics such as atmospheric and space science, meteorology, or global climate change.
25CI 25–25 · exposure 25 · augmentation 75 · importance 3.7/5 · click for rater detail
Teach college-level courses on topics such as atmospheric and space science, meteorology, or global climate change.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Higher education has been slow to adopt AI for core instruction despite digitization; most colleges still mandate human faculty and have resisted large-scale automation of teaching roles. While some institutions pilot AI tutoring supplements, genuine replacement of college-level course instruction remains rare and contested. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Higher education adopts AI tools gradually for content support, but full course delivery automation is rare and adoption is slow due to institutional and pedagogical norms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially augment a college instructor by generating lecture drafts, creating practice problems, grading quizzes, and offering personalized student support, significantly multiplying their productivity. The human instructor remains in control of learning objectives, assessment integrity, and student mentorship while offloading routine content production and initial feedback. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids lecture prep, quiz generation, explanations, and answering student questions, meaningfully boosting instructor productivity while the instructor remains central. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate lecture outlines, slides, and explanatory content, teaching college-level atmospheric science requires live interaction, real-time question-handling, assessment design, and adaptive pedagogy that current systems cannot reliably deliver end-to-end. AI might assist with content creation but cannot fully replace the instructor's classroom presence and dynamic engagement. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft lectures, slides, and materials but cannot autonomously deliver a college course including live instruction, mentoring, and adaptive engagement with students at equal quality.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Institutional accreditation, faculty governance, union contracts, and legal liability for educational outcomes create substantial barriers; colleges are generally required to employ credentialed instructors, and curriculum approval processes are rigid. Student expectation of human instruction and potential regulatory scrutiny of AI-delivered higher education add friction. |
| Adoption barriers | claude-sonnet-5 | 4/5 | University teaching typically requires credentialed faculty, accreditation standards, and institutional oversight, creating strong structural barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI content generation and tutoring systems are cheap in isolation, but providing a replacement for a full college course requires significant setup, moderation, assessment design, and human oversight, raising costs closer to parity with adjunct instructor wages. The infrastructure and validation burden currently offsets any cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate supporting materials, but replacing the full teaching function still requires human oversight, accreditation, and interaction, keeping costs comparable rather than dramatically lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably teaches college-level science courses independently; AI can draft materials and answer simple questions but lacks the pedagogical autonomy, student interaction capability, and institutional integration required for accredited instruction. Pilot systems and demos exist, but production deployment of AI as primary instructor is not yet realized at scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some AI-based tutoring and content-generation tools exist but no deployed product independently teaches a full college course in this specialized domain at scale. |
Design or develop new equipment or methods for meteorological data collection, remote sensing, or related applications.
25CI 20–30 · exposure 20 · augmentation 63 · importance 3.6/5 · click for rater detail
Design or develop new equipment or methods for meteorological data collection, remote sensing, or related applications.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Scientific instrument and method development remains a slow-moving, research-driven process with limited digitization and automation; adoption of AI tools in atmospheric science labs is nascent and primarily assistive rather than transformative. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science and instrumentation R&D remain a specialized, moderately slow-adopting sector where AI is used for data analysis more than for equipment design innovation.' |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist meteorologists by automating literature synthesis, optimizing simulation parameters, and suggesting design variations, thereby accelerating iteration, but the creative and validative core remains human-centric. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids literature review, simulation modeling, generative design exploration, and data analysis, meaningfully boosting productivity of scientists designing new methods or equipment.' |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in optimizing sensor design parameters and analyzing existing methods, designing fundamentally new meteorological equipment or data-collection approaches requires novel engineering integration, physical validation, and domain-specific innovation that current systems cannot execute end-to-end without substantial human direction and hands-on prototyping. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a creative engineering and scientific design task requiring novel hardware/method innovation, physical prototyping, and domain expertise; AI can assist with simulations and literature review but cannot autonomously design and validate new sensing equipment end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Design and development of new scientific equipment often requires domain credentials, regulatory compliance (e.g., frequency allocations for remote sensing), institutional approval, and peer validation before deployment; organizations typically require human accountability for novel methodologies. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement blocks AI assistance, but engineering design work is subject to organizational validation, safety testing, and institutional review before deployment, creating moderate friction.' |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI services (design assistance, simulation) are incremental cost additions to the human expert's work rather than replacements; the specialized expertise required means the loaded human wage for a meteorologist engineer remains the dominant cost. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can lower costs for simulation and literature synthesis, but the bulk of design, engineering, and physical testing still requires expensive skilled human labor, keeping overall cost comparable to human-driven R&D.' |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No production systems demonstrably design or develop novel meteorological equipment autonomously; AI tools exist for CAD assistance and literature review, but the core task—conceiving and validating new collection methods—remains human-driven with AI playing only a supporting role in real scientific workflows. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product independently designs meteorological instruments or remote sensing methods in production; this remains a research-stage aspiration at best.' |
Measure wind, temperature, and humidity in the upper atmosphere, using weather balloons.
9CI 5–13 · exposure 0 · augmentation 38 · importance 3.2/5 · click for rater detail
Measure wind, temperature, and humidity in the upper atmosphere, using weather balloons.
9| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Weather balloon operations remain a manual, human-intensive field practice with minimal digital automation adoption. Sectors performing this task (meteorology, atmospheric research) are conservative in automation deployment for core measurement activities. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Atmospheric science field operations remain a physically-dependent, moderately digitized sector where robotic or autonomous balloon launch systems are still niche and not widely deployed. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can assist post-collection analysis of balloon data (temperature, humidity pattern recognition) but offers limited real-time assistance during the physical launch, deployment, and retrieval phases that define the core task. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI assists significantly in analyzing the collected sensor data, flight path prediction, and integrating readings into forecasting models, even though it doesn't perform the physical launch itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Launching and retrieving weather balloons requires physical manipulation in outdoor conditions, site-specific setup, and real-time decision-making based on atmospheric conditions. Current AI systems cannot operate field equipment autonomously or replace the human judgment needed to execute this task in practice. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical field task requiring launching and tracking weather balloons with instrument payloads; AI cannot physically deploy or retrieve balloon-based sensors. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant regulatory oversight of upper-atmosphere measurements (FAA airspace, meteorological standards), requirement for licensed trained personnel to operate specialized equipment, and liability concerns for data quality all create meaningful adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While not licensed in the legal sense, operational protocols, safety requirements near airspace, and standardized meteorological procedures create moderate institutional friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of deploying autonomous robotics to replicate balloon launches and recoveries vastly exceeds the human labor cost of a trained atmospheric scientist performing the task directly. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot replace the physical equipment, helium, launch site operations, and human labor needed to release and track balloons, so there is no AI cost substitute for the physical task itself. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product autonomously launches, monitors, and retrieves weather balloons. While AI can *analyze* data collected by balloons, the physical act of measurement itself requires human operation and remains research-stage only. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product performs the physical launching, tracking, or handling of radiosonde balloons; this remains a manual/mechanical operation with automated data logging only. |
Collect air samples from planes or ships over land or sea to study atmospheric composition.
7CI 5–10 · exposure 0 · augmentation 25 · importance 3.8/5 · click for rater detail
Collect air samples from planes or ships over land or sea to study atmospheric composition.
7| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | This task occurs in physically-grounded, highly regulated sectors (aviation, maritime, atmospheric research) with limited digitization and no evidence of AI-driven automation of sample collection activities. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Atmospheric field research is a low-digitization, physically-intensive niche with minimal AI adoption for the collection step itself, though data analysis downstream may see more uptake. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with planning sampling routes, predicting optimal collection times based on weather or atmospheric models, and post-collection data analysis, but offers minimal assistance to the core physical act of sample collection itself. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI can help plan sampling routes, optimize timing/locations, and analyze collected data, but offers little assistance for the physical act of sample collection itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | The task requires physical collection of air samples from moving aircraft or ships, which demands on-site human presence, equipment handling, and real-time decision-making about sampling conditions. Current AI systems have no capability to perform physical sample collection at scale. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical sample-collection task requiring deployment of instruments on aircraft or ships in real-world environments; no AI system can perform this physical fieldwork. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant regulatory and safety barriers exist around operation of sampling equipment on commercial aircraft and vessels, chain-of-custody requirements for scientific samples, and airworthiness/maritime compliance that would require licensed personnel oversight regardless of automation level. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically for sample collection, but operational barriers exist: aircraft/ship access, safety protocols, and specialized equipment logistics limit substitution regardless of AI capability. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Automating this task would require developing robotics and sampling systems whose cost would far exceed the loaded wage of a technician, and such systems do not exist in production today. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the physical collection process, so there is no viable AI cost basis to compare against human/vehicle operation costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system can independently collect physical air samples from planes or ships; this is a fundamentally manual, embodied task that requires human technicians and specialized equipment in operational settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product physically collects atmospheric samples; this remains a manual/instrumentation-based field science task performed by humans and specialized hardware. |
Related occupations — Life, Physical & Social Science
How to read this
A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.
What would change this score
New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.