Data Sources

Every major weather data source — aggregated, normalized, and delivered as clean JSON through a single API.

Currently Integrated

SourceTypeCoverageResolutionDetails
NOAA GFSGlobal forecastGlobal28 kmView details ↓
ECMWF IFSGlobal forecastGlobal28 kmView details ↓
NOAA HRRRHigh-res forecastCONUS (US)3 kmView details ↓
NWS APIForecasts + alertsUS + territories2.5 kmView details ↓
ERA5Historical reanalysisGlobal31 kmView details ↓

Coming Soon

SourceTypeStatus
ECMWF AIFSAI global forecastIntegration in progress
NOAA AIGFSAI forecast (GraphCast-based)Integration in progress
NVIDIA Earth-2AI model suite (Apache 2.0)Planned Q2 2026
Open-MeteoMulti-model aggregationPlanned Q2 2026
GOES SatelliteImagery, cloud, lightningPlanned Q3 2026
NEXRAD RadarReflectivity, velocityPlanned Q3 2026

How Model Blending Works

When you request models=best (the default), Qanto selects the optimal data source for each field based on:

  1. HRRR for US locations within 0–48 hours (highest resolution, radar-assimilated)
  2. ECMWF IFS for global 0–10 day forecasts (most accurate global model)
  3. GFS for 10–16 day extended forecasts and as a fallback
  4. NWS for current US observations and severe weather alerts
  5. ERA5 for historical data requests

Every data point in the response includes a source field so you always know which model produced it.

Data Freshness

SourceUpdate FrequencyTypical Latency
NOAA GFSEvery 6 hours3.5-4 hours after init time
ECMWF IFSEvery 6 hours6-8 hours after init time
NOAA HRRREvery hour45-60 minutes after init time
NWS APIEvery 1-2 hoursNear real-time
ERA5Monthly (5-day lag)5 days behind present

Licensing Note

All source data is free or open-licensed. NOAA data is US public domain. ECMWF opened its data fully under CC-BY-4.0 in October 2025. ERA5 data is available under the Copernicus license (free for commercial use with attribution). The moat is the intelligence layer, not the data.


NOAA GFS

The Global Forecast System — NOAA's workhorse global weather model.

ProviderNOAA (National Oceanic and Atmospheric Administration)
TypeGlobal numerical weather prediction (NWP)
Resolution28 km (0.25°) horizontal, 127 vertical levels
Forecast Range384 hours (16 days)
Update FrequencyEvery 6 hours (00z, 06z, 12z, 18z)
CoverageGlobal
LicenseUS Public Domain (no restrictions)

What GFS Provides

GFS produces global weather forecasts for temperature, wind, humidity, precipitation, pressure, cloud cover, and dozens of other atmospheric variables. It runs on NOAA's supercomputers four times daily and is the backbone of most free weather services worldwide.

Strengths and Weaknesses

Strengths

  • Free and unrestricted (public domain)
  • 16-day extended forecasts
  • Global coverage with consistent quality
  • Massive user community and documentation

Weaknesses

  • Lower resolution than regional models (28 km vs 3 km HRRR)
  • Less accurate than ECMWF IFS for medium-range (3-10 days)
  • 6-hour update cycle means stale data between runs
  • Raw data in GRIB2 format requires specialized parsing

How We Use It

GFS serves as the primary fallback for global forecasts and is the exclusive source for 10–16 day extended forecasts. For US locations within 48 hours, HRRR takes priority. For 0–10 days globally, ECMWF IFS is preferred when available.

Raw Access

Direct NOAA GFS access (GRIB2)
# Download GFS 0.25° forecast for hour 6
curl -O "https://nomads.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/gfs.20260315/00/atmos/gfs.t00z.pgrb2.0p25.f006"

# Note: Raw GRIB2 files are 300-500 MB each.
# Qanto handles the download, parsing, and normalization for you.

ECMWF IFS

The Integrated Forecasting System — widely considered the world's most accurate global forecast model.

ProviderECMWF (European Centre for Medium-Range Weather Forecasts)
TypeGlobal numerical weather prediction (NWP)
Resolution28 km (0.25°) horizontal, 137 vertical levels
Forecast Range360 hours (15 days)
Update FrequencyEvery 6 hours (00z, 06z, 12z, 18z)
CoverageGlobal
LicenseCC-BY-4.0 (open since October 2025)

Why ECMWF Is Special

ECMWF consistently outperforms every other operational global forecast model in independent verification. It has the most sophisticated data assimilation system in the world, ingesting over 800 million observations daily from satellites, weather stations, radiosondes, aircraft, and ocean buoys. The model is funded by 35 member and cooperating states.

For medium-range forecasting (3–10 days), ECMWF is the gold standard. It is the model that energy traders, airlines, and governments rely on for critical decisions.

ECMWF AIFS

ECMWF has also deployed AIFS (Artificial Intelligence Forecasting System), the first AI weather model running operationally at a major weather center. AIFS is trained on ERA5 reanalysis data and produces forecasts competitive with the physics-based IFS at a fraction of the compute cost. Qanto is integrating AIFS output alongside IFS.

How We Use It

ECMWF IFS is the preferred model for all global forecasts in the 0–10 day range. When you request models=best, ECMWF IFS data takes priority over GFS for temperature, wind, precipitation, and pressure forecasts.

Open Data vs. Operational

Since October 2025, ECMWF provides its full operational output under CC-BY-4.0, which allows unrestricted commercial use with attribution. Previously, only a subset of data was available through the ECMWF Open Data initiative. The full open release includes HRES deterministic forecasts, ensemble forecasts (ENS), and AIFS output.


NOAA HRRR

High-Resolution Rapid Refresh — 3 km resolution, hourly updates, radar-assimilated.

ProviderNOAA / NCEP (National Centers for Environmental Prediction)
TypeConvection-allowing regional forecast
Resolution3 km horizontal, 50 vertical levels
Forecast Range48 hours
Update FrequencyEvery hour
CoverageContinental US (CONUS)
LicenseUS Public Domain

Why HRRR Matters

HRRR is the highest-resolution operational weather model in the US. At 3 km, it can resolve individual thunderstorms, sea breezes, mountain winds, and urban heat islands that global models at 28 km completely miss. It updates every hour by assimilating the latest radar observations, making it the most responsive model available.

Best Use Cases

Use CaseWhy HRRR Excels
Drone operations3 km wind data for pre-flight planning and in-flight monitoring
Construction planningHourly precipitation and wind forecasts for site management
Outdoor eventsPrecise thunderstorm timing and location
Energy tradingHourly wind and solar irradiance for renewable generation forecasts
Severe weatherBest at predicting individual storm cells and their tracks
AgricultureField-level frost warnings and precipitation timing

Limitations

  • US-only coverage (CONUS) — not available for international locations
  • Maximum 48-hour forecast horizon — use GFS/ECMWF for longer range
  • Large data volume — each hourly run produces ~2 GB of GRIB2 data
  • Occasional initialization issues during severe weather outbreaks

How We Use It

HRRR is the top-priority model for all US locations within the 0–48 hour forecast window. When you request weather for a US location with models=best, HRRR data is used for hourly forecasts and current conditions. For forecast hours beyond 48, the system seamlessly transitions to ECMWF IFS or GFS.

Raw Access

Direct NOAA HRRR access (GRIB2)
# Download HRRR forecast for hour 6 from the 12z run
curl -O "https://nomads.ncep.noaa.gov/pub/data/nccf/com/hrrr/prod/hrrr.20260315/conus/hrrr.t12z.wrfprsf06.grib2"

# Or via Google Cloud (faster, more reliable):
gsutil cp gs://high-resolution-rapid-refresh/hrrr.20260315/conus/hrrr.t12z.wrfprsf06.grib2 .

NWS API

The National Weather Service — official US government forecasts, observations, and severe weather alerts.

ProviderNWS (National Weather Service, NOAA)
TypeGridded forecasts, point observations, alerts
Resolution2.5 km gridded forecasts
Forecast Range7 days (detailed), 7 days (extended)
Update FrequencyEvery 1-2 hours
CoverageUS states, territories, and marine zones
LicenseUS Public Domain

What NWS Provides

The NWS API provides three key data types:

  • Gridded forecasts — High-resolution (2.5 km) forecasts produced by local NWS offices with human meteorologist input
  • Observations — Real-time data from ASOS/AWOS stations at airports and other automated stations
  • Alerts — Severe weather watches, warnings, and advisories with polygon geometries and urgency levels

Why We Use NWS

NWS forecasts incorporate human meteorologist expertise on top of model output. Local forecasters adjust for terrain effects, microclimates, and event-specific conditions that pure model output misses. For current conditions, NWS observation stations provide ground-truth measurements rather than model estimates.

NWS alert data is the authoritative source for severe weather warnings in the US. There is no commercial alternative for this data — it comes directly from the government.

Known Limitations and How We Handle Them

  • Reliability — The NWS API has occasional outages and slow responses. We implement aggressive caching and fallback to model data when the API is unavailable.
  • Rate limits — NWS has undocumented rate limits that can throttle heavy users. We distribute requests and cache aggressively.
  • US-only — NWS only covers US locations. For international locations, we rely on GFS and ECMWF for observations and forecasts.
  • Inconsistent format — NWS returns data in GeoJSON with inconsistent field naming. We normalize everything into our standard JSON schema.

Raw Access

NWS API — Point forecast
# Get forecast for a specific latitude/longitude
curl "https://api.weather.gov/points/40.7128,-74.0060" \
  -H "User-Agent: (qanto.com, contact@qanto.com)"

# The response contains links to:
#   - /forecast (7-day detailed)
#   - /forecastHourly (hourly forecast)
#   - /forecastGridData (raw gridded data)
NWS API — Active alerts
# Get active alerts for New York state
curl "https://api.weather.gov/alerts/active?area=NY" \
  -H "User-Agent: (qanto.com, contact@qanto.com)"

ERA5 Reanalysis

80+ years of global weather data — the most complete historical weather dataset ever created.

ProviderECMWF / Copernicus Climate Change Service (C3S)
TypeGlobal atmospheric reanalysis
Resolution31 km (0.25°) horizontal, 137 vertical levels
Time Range1940 to present (5-day lag)
Temporal ResolutionHourly
CoverageGlobal
LicenseCopernicus License (free, commercial use OK with attribution)

What Is Reanalysis?

Reanalysis combines historical observations (weather stations, satellites, radiosondes, ships, buoys) with a modern weather model to produce a complete, consistent picture of past weather. Think of it as “running today's best forecast model backwards through history.”

ERA5 is the fifth generation of ECMWF reanalysis. It provides hourly estimates of dozens of atmospheric, land, and ocean variables from 1940 to the present day, at 31 km resolution globally. It is the most widely used climate dataset in research and industry.

Use Cases

Use CaseDescription
Historical comparisonCompare current conditions to 80+ years of history at any location
Anomaly detectionIs today's temperature unusual? ERA5 provides the climatological baseline
Energy modelingWind and solar resource assessment using decades of hourly data
Insurance pricingHistorical risk assessment for parametric insurance products
Agriculture planningGrowing degree days, frost dates, and precipitation patterns
AI model trainingTraining data for AI weather models (GenCast, GraphCast, AIFS all trained on ERA5)

API Availability

ERA5 data is available through the Qanto API via the /v1/weather/:location/history endpoint. Specify a date range to retrieve historical hourly data for any location worldwide. Available on Pro plan and above.

Data Lag

ERA5 data has approximately a 5-day lag from the present. This is because the reanalysis requires collecting and quality-controlling observations before running the assimilation. For the most recent 5 days, Qanto uses forecast model data instead.

Raw Access

Python — CDS API for ERA5
import cdsapi

client = cdsapi.Client()

client.retrieve(
    "reanalysis-era5-single-levels",
    {
        "product_type": "reanalysis",
        "format": "netcdf",
        "variable": [
            "2m_temperature",
            "10m_u_component_of_wind",
            "10m_v_component_of_wind",
            "total_precipitation",
        ],
        "year": "2025",
        "month": "06",
        "day": ["01", "02", "03"],
        "time": [f"{h:02d}:00" for h in range(24)],
        "area": [42, -75, 40, -73],  # North, West, South, East
    },
    "era5_nyc_june2025.nc",
)

# Note: Raw ERA5 downloads can be slow (hours for large requests).
# The Qanto API pre-indexes ERA5 data for instant retrieval.