Weather simulation data
This help page gives an overview of:
- Weather simulation data - what's that?
- Why meteoblue simulation data is comparable to measurement data
- Quality comparison: simulation data vs. measurement data
- Profiles of grid cells with simulation data and measurement data
- Calculation of simulation data
- Availability of meteoblue simulation data vs. measurement data
Weather simulation data – what’s that?
Basically, meteoblue forecasts are calculated with own models, not from individual weather stations. These weather models are based on the NMM (Nonhydrostatic Meso-Scale Modelling) or NEMS (NOAA Environment Monitoring System) technology, which enables the inclusion of detailed topography, ground cover and surface cover. Each forecast is archived by meteoblue at least once daily. From these data, we create a complete hourly history of the weather situation, which we can deliver for every place in the world from 1984 until today.
Official and reliable weather stations are scarce, and most of the weather station measurement data is useful only for a radius of 3 to 12km surrounding the station. This means that less than 1% of the worlds surface is covered by measured data. For the atmosphere, much less measurement data exist. Since weather stations are located unevenly on lands surface, you will find just a few places with weather stations in the vicinity. In most areas, stations are widely spaced and there are many places (especially in South America /Africa/ Asia/ unpopulated areas) where you will not find any weather station in the vicinity.
meteoblue weather simulations, however, cover the whole Earth surface, also the seas and mountainous areas. Therefore, meteoblue can offer forecast and archive data for any place in the world, including the atmosphere above us.
Why meteoblue simulation data is comparable to measurement data
There are a lot of reasons, why the meteoblue simulation data is comparable to measurement data. The most important criteria is stated in the table below:
|meteoblue simulation data||weather station measurement data|
< 1 km
|Number of weather variables||
|Number of years||
2-30 (with gaps)
hourly, 3 hourly, daily
(hourly), 3 hourly, daily
Quality comparison: simulation data vs. measurement data
This picture shows, where weather station measurements and meteoblue simulation data is available. meteoblue simulation data is available worldwide (grid), whereas rmeasurement data is only valid for an area of 1 to 3 km around a weather station. The color shows the temperature deviation from measurements and simulation data to real temperature.
This picture shows the quality of simulation data with MOS. The temperature deviation can be reduced by weather measurement data (data must be available at least for one complete year). meteoblue uses gridded MOS for most of the Earth's surface, so you can expect the same precision in most areas, based on our very thorough evaluation of the precision of temperature simulations. The results are validated daily at the spatial scales at which measurements exist (which varies largely). However the verification does not have any influence on the data itself as it is not a reanalysis were the forecast is merged with observations. We will, in addition, include the ERA-5 reanalysis from ECWMF once it becomes available. We currently keep all models untouched as many of our customers want the raw historical data to run training.
Profiles of grid cells with simulation data and measurement data
The resolution of the grid cell depend on the model domain. meteoblue offers weather data for 3-30km resolution.
Calculation of grid cell data in flat terrain
Forecast data: meteoblue can calculate weather data for every point with a specific altitude in the grid cell.
Archive data: meteoblue archives each forecast data set at least once daily. The archive data is calculated for the average altitude of the grid cell (see “average temperature of the grid cell). The temperature deviance for flat terrain is between 0 and 2K.
Calculation of grid cell data in mountainous terrain
Forecast data: meteoblue can calculate weather data for every point with a specific altitude in the grid cell (see “point temperature at specific location”).
Archive data: meteoblue archives each forecast data set at least once daily. The archive data is calculated for the average altitude of the grid cell (see “average temperature of the grid cell). The temperature deviance for mountainous terrain is between 0 and 5K.
|Flat terrain||Mountainous terrain|
|Forecast data||Forecast data calculated for every point in grid cell|
|Archive data||Forecast data calculated for average of grid cell|
Calculation of simulation data
Forecasts can be very wrong after just 10 days, so how can we compute 30 years? In fact we only compute a 1 day forecast, starting from real weather observations. This short term forecast has been validated against 10’000s of weather stations over many years, and is very close to reality (it corresponds to what is called a "reanalysis" which is equivalent to the best possible forecast available for the day-ahead). We repeat this procedure for every day of the year for 30 years. That makes 8760 one-day forecasts per year, which are then joined together into a continuous >30-year time series since - the meteoblue weather history.
Availability of meteoblue simulation data vs. measurement data
meteoblue simulation data is available without any gaps since 1984 with our global model (NEMSGLOBAL). This global model has a resolution of 30 km and we offer more than 30 years of hourly data.
Applications for simulation vs. measurement data
meteoblue simulations can replace measurements in most areas. A simulation will be more precise for a given place than measurements taken 20 or more km away (in mountain areas, the simulation is already more accurate than a measurement at a distance of 3-10 km away). In places with very high need for reliability, such as airports, city centres, building management, irrigation systems, renewable energy production sites, high traffic roads and some others, the additional precision offered by measurements justify the (substantially) higher cost.