yr.no Weather Forecast

yr.no is a free weather service operated by the Norwegian Meteorological Institute (MET Norway). The Locationforecast 2.0 API provides global point-based weather forecasts in GeoJSON format. No API key is required -- only a User-Agent header identifying the application.

Key characteristics

Property	Value
Provider	MET Norway (Norwegian Meteorological Institute)
Spatial resolution	Point-based (single lat/lon coordinate)
Temporal resolution	Hourly (0-48h), 6-hourly (2-10 days)
Coverage	Global
Forecast window	Up to ~9 days from current time
Historical data	None -- forecast only
Format	GeoJSON (RFC 7946)
CRS	EPSG:4326 (WGS 84)

API endpoints

The Locationforecast 2.0 API offers three endpoints:

Endpoint	Description
/compact	Subset of variables sufficient for most use cases
/complete	Full set of variables including percentiles and cloud layers
/status	API health check

The pipeline uses the complete endpoint to access all available variables:

https://api.met.no/weatherapi/locationforecast/2.0/complete?lat={lat}&lon={lon}

API requirements

User-Agent header

A custom User-Agent header is mandatory. It must contain the application name and contact information (GitHub URL, email, or website). Missing or generic User-Agent strings return 403 Forbidden. Falsifying this information risks permanent blacklisting.

The pipeline sets:

User-Agent: prefect-climate/1.0 github.com/mortenoh/prefect-examples

Coordinate precision

Coordinates must use no more than 4 decimal places. Using 5 or more decimals returns 403 Forbidden. This restriction enables effective server-side caching. Four decimal places give ~11 m precision, which is more than sufficient for weather data.

# Coordinates are truncated to 4 decimals before the API call
lat = round(lat, 4)  # e.g. 8.4842
lon = round(lon, 4)  # e.g. -13.2344

HTTPS

Always use HTTPS. Unencrypted HTTP traffic may be throttled or blocked.

Caching and conditional requests

MET Norway strongly encourages caching. Every response includes:

• Expires -- when the forecast data will next be updated. Do not re-request before this time.
• Last-Modified -- when the data was last changed.

Use If-Modified-Since with the cached Last-Modified value on subsequent requests. A 304 Not Modified response means the cached data is still current.

Rate limiting

No hard published limit, but approximately 20 requests/second is the practical ceiling. Monitor for 429 Too Many Requests responses. The pipeline processes org units sequentially, which naturally stays well within limits.

Status codes

Code	Meaning
200	Success, new data returned
304	Not Modified -- use cached data
403	Forbidden -- bad User-Agent or coordinate precision
429	Throttled -- reduce request rate

Response format

The API returns a GeoJSON FeatureCollection with a timeseries array. Each entry has an ISO 8601 timestamp and nested data:

{
  "type": "Feature",
  "geometry": { "type": "Point", "coordinates": [-13.2344, 8.4842] },
  "properties": {
    "timeseries": [
      {
        "time": "2026-03-03T12:00:00Z",
        "data": {
          "instant": {
            "details": {
              "air_temperature": 28.5,
              "air_pressure_at_sea_level": 1010.2,
              "relative_humidity": 68.0,
              "wind_speed": 2.3,
              "cloud_area_fraction": 45.0,
              "dew_point_temperature": 22.1,
              "wind_from_direction": 210.5
            }
          },
          "next_1_hours": {
            "summary": { "symbol_code": "partlycloudy_day" },
            "details": { "precipitation_amount": 0.0 }
          },
          "next_6_hours": {
            "summary": { "symbol_code": "partlycloudy_day" },
            "details": {
              "precipitation_amount": 0.2,
              "air_temperature_max": 30.1,
              "air_temperature_min": 27.8
            }
          }
        }
      }
    ]
  }
}

• instant.details -- atmospheric variables at the exact timestamp.
• next_1_hours.details -- accumulated values for the next 1 hour (available for first ~48 hours).
• next_6_hours.details -- accumulated values for the next 6 hours (available for the full forecast window).

Available variables

Instant variables

Variables marked with compact are included in both /compact and /complete; others are /complete only.

Variable	Unit	Compact	Description
air_temperature	C	yes	Air temperature at 2 m height
air_temperature_percentile_10	C	no	10th percentile air temperature
air_temperature_percentile_90	C	no	90th percentile air temperature
air_pressure_at_sea_level	hPa	yes	Air pressure reduced to sea level
cloud_area_fraction	%	yes	Total cloud cover (all heights)
cloud_area_fraction_high	%	no	Cloud cover above 5000 m
cloud_area_fraction_medium	%	no	Cloud cover 2000-5000 m
cloud_area_fraction_low	%	no	Cloud cover below 2000 m
dew_point_temperature	C	no	Dewpoint temperature at 2 m
fog_area_fraction	%	no	Area covered in fog (visibility < 1 km)
relative_humidity	%	yes	Relative humidity at 2 m
ultraviolet_index_clear_sky	index	no	UV index for cloud-free conditions (0-11+)
wind_from_direction	degrees	yes	Direction wind is coming from (0 = N, 90 = E)
wind_speed	m/s	yes	Wind speed at 10 m (10-minute average)
wind_speed_percentile_10	m/s	no	10th percentile wind speed
wind_speed_percentile_90	m/s	no	90th percentile wind speed
wind_speed_of_gust	m/s	no	Maximum gust at 10 m

Period variables

Variable	Unit	Period	Description
precipitation_amount	mm	1h, 6h	Expected precipitation for period
precipitation_amount_max	mm	1h, 6h	Maximum likely precipitation
precipitation_amount_min	mm	1h, 6h	Minimum likely precipitation
probability_of_precipitation	%	1h, 6h, 12h	Chance of precipitation
probability_of_thunder	%	1h, 6h	Chance of thunder
air_temperature_max	C	6h	Maximum air temperature over period
air_temperature_min	C	6h	Minimum air temperature over period
symbol_code	string	1h, 6h, 12h	Weather icon code

Variable names follow the international CF Standard Name vocabulary (required by the EU INSPIRE directive).

Variables used in the pipeline

Six variables are extracted from the /complete response and aggregated from hourly/6-hourly to daily values:

Variable	API field	Source	Daily aggregation	Unit
Temperature	air_temperature	instant	Mean of all hourly values	C
Precipitation	precipitation_amount	next_1_hours / next_6_hours	Sum of all periods	mm
Relative humidity	relative_humidity	instant	Mean of all hourly values	%
Wind speed	wind_speed	instant	Mean of all hourly values	m/s
Cloud cover	cloud_area_fraction	instant	Mean of all hourly values	%
Air pressure	air_pressure_at_sea_level	instant	Mean of all hourly values	hPa

Daily aggregation logic

For each calendar day (UTC), the pipeline:

1. Groups all timeseries entries by date.
2. For instant variables (temperature, humidity, wind speed, cloud cover, pressure): computes the arithmetic mean of all hourly values within the day.
3. For precipitation: sums the precipitation_amount values. Prefers next_1_hours (higher resolution) when available; falls back to next_6_hours for the later part of the forecast window.

Days with no data for a given variable produce None (skipped during DHIS2 import).

Approach: point-based forecasts via centroids

Unlike CHIRPS and ERA5 (which provide raster data processed with zonal statistics), yr.no returns point forecasts for specific coordinates.

The flow computes the centroid of each org unit polygon (mean of all coordinate vertices) and queries yr.no for that single point. This is a simple approximation that works well for org units where the weather is relatively uniform across the area.

from prefect_climate import centroid

lat, lon = centroid(org_unit.geometry)
# e.g. (8.4842, -13.2344) for Bo District, Sierra Leone

For large or irregularly shaped org units, the centroid may not be representative of conditions across the entire area. Consider this limitation when interpreting data for mountainous regions or coastal boundaries.

Pipeline overview

1. Fetch org unit geometries from DHIS2 (Polygon/MultiPolygon only)
2. Ensure 6 data elements and 1 daily data set exist in DHIS2
3. For each org unit: a. Compute polygon centroid (mean lat/lon) b. Truncate coordinates to 4 decimal places c. Fetch forecast from yr.no /complete endpoint d. Aggregate hourly timeseries to daily values
4. Build DHIS2 data values (period format: YYYYMMDD)
5. POST data values to DHIS2 /api/dataValueSets
6. Create markdown artifact with import summary

@flow(name="dhis2_yr_weather_import", log_prints=True)
def dhis2_yr_weather_import_flow(query: ClimateQuery | None = None) -> ImportResult:
    client = get_dhis2_credentials().get_client()
    org_units = ensure_dhis2_metadata(client, query.org_unit_level)
    forecast_results = []
    for ou in org_units:
        result = fetch_org_unit_forecast(ou)  # centroid + yr.no API
        forecast_results.append(result)
    data_values = build_data_values(forecast_results)
    return import_to_dhis2(client, dhis2_url, org_units, data_values)

DHIS2 data model

Data Set: WF: yr.no: Weather Forecast (WfYrFrcSet1)  [Daily, openFuturePeriods=10]
  |- Data Element: WF: yr.no: Temperature       (WfYrTmpEst1)  -- mean C
  |- Data Element: WF: yr.no: Precipitation      (WfYrPrcEst1)  -- sum mm
  |- Data Element: WF: yr.no: Relative Humidity   (WfYrHumEst1)  -- mean %
  |- Data Element: WF: yr.no: Wind Speed          (WfYrWndEst1)  -- mean m/s
  |- Data Element: WF: yr.no: Cloud Cover         (WfYrCldEst1)  -- mean %
  |- Data Element: WF: yr.no: Air Pressure        (WfYrPrsEst1)  -- mean hPa

UID	Name	Aggregation	Unit
WfYrTmpEst1	WF: yr.no: Temperature	Daily mean	degrees Celsius (C)
WfYrPrcEst1	WF: yr.no: Precipitation	Daily sum	millimetres (mm)
WfYrHumEst1	WF: yr.no: Relative Humidity	Daily mean	percent (%)
WfYrWndEst1	WF: yr.no: Wind Speed	Daily mean	metres per second (m/s)
WfYrCldEst1	WF: yr.no: Cloud Cover	Daily mean	percent (%)
WfYrPrsEst1	WF: yr.no: Air Pressure	Daily mean	hectopascals (hPa)

The data set uses openFuturePeriods=10 since forecasts extend ~9 days into the future.

Building a time series

The yr.no API provides only future forecasts -- there is no historical archive. Each run of the flow imports the current ~9-day forecast window. Running the flow on a regular schedule (e.g. daily via cron) builds up a historical record of forecast data in DHIS2 over time. Overlapping forecast days from consecutive runs are updated in place.

Health relevance

Variable	Health relevance
Temperature	Heat-related illness, malaria transmission windows (optimal 20-30 C)
Precipitation	Waterborne disease risk, flooding, mosquito breeding habitat
Relative humidity	Pathogen survival and airborne transmission, respiratory illness
Wind speed	Disease vector dispersal, air quality, extreme weather preparedness
Cloud cover	UV exposure estimation, solar energy availability
Air pressure	Migraine triggers, respiratory conditions, storm prediction

Comparison with ERA5 and CHIRPS

Property	yr.no	ERA5-Land	CHIRPS
Data type	Forecast (future)	Reanalysis (historical)	Observation (historical)
Spatial	Point-based (centroid)	~9 km raster (zonal stats)	~5 km raster (zonal stats)
Temporal	Daily (from hourly)	Monthly	Monthly (from daily)
Variables	6 weather indicators	7 climate indicators	Precipitation only
API key	None (User-Agent only)	CDS API key required	None (direct download)
Latency	Real-time forecast	~5 days behind present	~3 weeks behind present

yr.no complements ERA5 and CHIRPS by providing near-term forecast data rather than historical observations. This enables early warning and preparedness workflows in DHIS2.

Running the flow

The flow is intended for manual or scheduled runs. Each run fetches the current forecast window (~9 days) and imports it into DHIS2.

# Direct run
uv run python flows/dhis2/dhis2_yr_weather_import.py

# Via Docker deployment
docker compose build dhis2-yr-weather-docker
docker compose up -d dhis2-yr-weather-docker

Direct API usage

from prefect_climate.yr import fetch_daily_forecasts

# Fetch forecast for Freetown, Sierra Leone
forecasts = fetch_daily_forecasts(lat=8.48, lon=-13.23)
for day in forecasts:
    print(f"{day.date}: {day.temperature}C, {day.precipitation}mm rain")

from prefect_climate.yr import fetch_forecast, aggregate_forecast_daily

# Two-step: fetch raw response, then aggregate
raw = fetch_forecast(lat=8.48, lon=-13.23)
daily = aggregate_forecast_daily(raw)

References

• yr.no Locationforecast 2.0 documentation
• Locationforecast data model
• Locationforecast HOWTO
• MET Norway terms of service
• CF Standard Names
• DHIS2 Climate Tools