NOAA Data Portal

It is NOAA's data homepage and the front door to an observing ecosystem that ingests tens of terabytes daily into an archive approaching one exabyte. Collection platforms named on the page run from geostationary and polar-orbiting satellites through Hurricane Hunter aircraft, radars, buoys and autonomous gliders, standardized to WMO, IHO and ISO formats. Content organizes into nine themes: Weather, Climate, Ocean, Satellite, Fisheries, Coastal, Great Lakes, Geospatial Data and NOAA Publications.

Structurally it is a map, not a single download: the hub routes technical users toward a unified catalog search, cloud-hosted holdings, a weather and forecast service and the NCEI archive service, and leaves the joining to you. Within Datadory's shelf this record owns that discovery layer - one query that surfaces the specialist hazard feeds underneath, including two of its own siblings: the billion-dollar disasters table and its time-series companion.

One typed row per cataloged holding, identical in shape whether the record points at a station network or a satellite constellation - identifier, title, theme, format, grain, window. Three real entries lead the block above: the Storm Events Database reaching back to 1950, the U.S. Billion-Dollar Weather and Climate Disasters table covering 1980-2024, and GOES satellite imagery resolving to full scenes.

Beneath the index rows sits the shape of a delivered hazard extract: one row per event x geography x period, carrying the event type, the geography your scope names, the observation window and the measured value. Those cells populate once you name perils and regions - they were not transcribed during the research pass, and Datadory does not pad samples with invented observations.

Read the coverage columns as provenance: they are what lets a diff between pulls surface newly published holdings immediately.

Ten fields make up the confirmed spine of every delivered row: five identify the holding (record_id, title, theme, observing_system, format) and five describe what you can plan around (granularity, temporal_coverage, spatial_coverage, update_cadence, citation).

An honesty note the hub itself buries: field-definition confidence here is inferred, not parsed. The August 2026 research pass verified the hub's structure and scope but did not enumerate the thousands of subordinate dataset schemas underneath it, so the dictionary describes the normalized layer Datadory delivers rather than any single upstream file. Anything beyond the spine folds under an explicit request note instead of guessed columns - the full dictionary follows in tabular form below.

Geography - split by design: global for satellite and ocean products, US-focused for weather, climate and coastal observations. One pull therefore spans a Pacific typhoon scene and a Gulf Coast rain gauge without a second project.

Temporal - the widest span on the shelf. Live observation feeds sit at the leading edge while some climate station archives run past a century, so a single slice serves both today's watch boxes and a 1930s drought study.

Granularity - point observations and station records through gridded model output to full satellite scenes, stacked in one normalized record.

Set against the wider catalog - average quality score 7.81 across all 1,744 datasets - this record scores 8/10, carried by breadth and freshness and held back by exactly the heterogeneity normalization exists to fix.

API, files, or your warehouse. Daily, weekly, or hourly.

Your cadence is yours alone. Take a one-time snapshot of the slice you name, or keep a warehouse table current so each new pull diffs cleanly against yesterday's rows. Deliveries arrive normalized to the field dictionary above with record identifiers intact, so hazard extracts join against your own exposure book without hand-parsing text.

Two postures work well in practice. Take the discovery layer alone as a monitored map of what the ecosystem holds. Or take it joined to the deeper extracts - storm events, hurricane tracks, satellite scenes - aligned on the record key where the index row and the data share it. Most production users end up wanting both, in that order.

• Catastrophe-model feature engineering - station, gridded and satellite products become exogenous weather features in demand and damage models.
• Storm-footprint reconstruction - tracks, storm events and precipitation grids rebuild a named event county by county.
• Exposure screening against physical hazard - coastal and weather observations flag recurring-peril geographies before a property book prices them.
• Weather-anomaly attribution - recorded storms and temperature anomalies explain demand swings in competitor results.
• Climate context for market research - official observation archives anchor regional analyses in measured conditions.
• Catalog monitoring - diffing coverage stamps between pulls surfaces newly published holdings immediately.

Developers and data-product builders lead fit: a documented catalog that arrives as one fixed schema regardless of theme is precisely what monitoring features want; see developers builders use cases. Data scientists and ML engineers (equally strong) fuse the same material into a weather feature store keyed on geography and time; see data scientists use cases. Journalists, academics and students verify hurricane tracks and climate records against primary observations before publication; see journalists academics use cases. Competitive intelligence teams attribute demand swings to recorded anomalies; see competitive intel use cases. Investors and quants cross-check weather-driven narratives against official archives; see investors & quants use cases. Market researchers and consultants ground location studies in measured climate; see market researchers use cases.

This record owns the raw observations territory. Its nearest shelf neighbours own territories beside it: the billion-dollar disasters table holds all 403 US disasters over $1B since 1980 with CPI-adjusted loss figures, and its time-series companion renders Year x peril x geography aggregates exportable as CSV, JSON or XML - the faster routes when you already want loss aggregates rather than the weather itself. FEMA NFIP Redacted Claims v2 holds 2.72 million paid flood claims; NAIC Research and Insurance Data covers regulation; First Street Foundation Climate Risk Data models forward-looking risk per building.

Only this slice answers what did the atmosphere actually do - because only an observing agency measures it continuously. The price-versus-peril contrast is worked through in FRED Insurance-Related Series vs NOAA Data Portal.

Because the artifact is a homepage over an exabyte-scale estate - nine themes, four service families, six serialization formats and thousands of subordinate schemas - and most questions want one table. Datadory normalizes the slice to the ten-field spine, joins the deeper extracts on keys your warehouse already speaks, and schedules deliveries on your cadence so consecutive pulls diff cleanly.

Browse the rest of the industry on the property casualty insurance data hub, the best property-casualty-insurance datasets ranking, or the full catalog; the archive side of the family is profiled at the NCEI source page.