meteoblue generally uses 2 model runs per day. The runs are based on the assimilations for 00:00 UTC and 12:00 UTC. Initialisation generally takes place 2 hours after assimilation. Actualisations are made between 6 and 8 hours after assimilation. Depending on the update timing by the customer, the last change may be reflected (or not). In unstable weather, these changes may be significant.
Some data packages contain hourly updates (post-processing) from weather station, some not. If a forecast for the same location is made using different data packages, these may result in different values for the variables which are modified by post-processing. As a result, you may occasionally get substantial differences. Further differences originate from different providers using different forecast models.
There are 3 types of post-processing methods:
Using past 1 year of measurement data to correct the forecast for a particular variable (temperature, relative humidity, windspeed, radiation). This is station dependent.
Using past 1 week of measurement data to correct the forecast for particular variables (temperature, relative humidity, windspeed, radiation). This is station dependent.
Using past 1 hour of measurement data to correct the forecast for a particular variables (temperature, relative humidity, wind speed, radiation, precipitation in some areas). This is area and station dependent, and can also happen independent of a customer station, by using other nearby data sources.
Weather data are constantly actualised as the weather evolves.
Weather data based on measurements are actualised when measurement data become available.
To actualise weather data, meteoblue uses the following prioritisation scheme:
- Global weather model: this is the most available source of information and supplies most weather variables. They are updated twice a day, though the are also frequently the least precise of all available sources.
- Regional weather model (also called "LAM" = local area model): these cover most regions of interest (with a "domain") and supply all necessary weather variables. They are updated 2(-4) times a day, and are more precise than global models.
- Observations: these are mostly based on remote sensing (such as satellites, radar), and are available for certain regions. Update intervals vary between 6 hours and 15 minutes, depending on the source and region. Variables detected by sensors (such as cloud cover, precipitation) will be used to adjust the simulations (e.g. cloud cover and radiation, rainfall) in a process, in which observation and simulation are matched using plausibility checks.
- Measurements: these are mostly obtained from weather stations. Update intervals vary between 12 hours and 15 minutes, depending on the station and variable. Station data is used to update information in the surroundings for the measured variables, if they are available, and for any locations in the surroundings that have similar conditions to the station, such as the altitude and the proximity. If there are no updates from the station, or if the chosen location has different altitudes (although it may be close to the station), then the measurements will not be used for updating.
Most actualisations are not archived. Historic data are available for each of these data sources, and possible corrections can be made using comparisons between the different data sources.
A schematic example of actualisations is shown for 4 locations (with and without actual measurements) and 3 actualisation timings:
(Legend for data source: n=Model; m=MOS, o=Observation: x = not available. AX= actualisation time)
|Location with historical and current measurements|
|Location with historical (and without current) measurements|
|Location with (no historical) and current measurements|
|Location without measurements|
With 2 model runs per day, starting at 00 and 12 UTC, the resulting forecast range of 168 hours will vary by timezone. The resulting forecast hours available (in local time) are displayed in the following graph: