The Meteomatics weather API downscales on the fly from the models' native resolutions to 90m horizontal resolution by using NASA's SRTM topography measured by the space shuttles. Since the weather models typically have a much coarser resolution (1km to 25km or even coarser) topographically challenging environments as e.g. alpine regions, are not very well modelled and the difference between the actual elevation and the modelled elevation can be rather large. That of course has an influence on the quality of the forecasted parameters.
An example is mount Säntis in Switzerland. It is roughly 2600m above mean sea level and Schwägalp, the lower terminus, which is located at a horizontal distance of just about 1km to mount Säntis, is about 1000m lower. Due to their horizontal proximity, both locations end up being in the same simulation cell and the model forecasts roughly the same temperature for both locations. In reality, however, we know that mount Säntis, with its high altitude, typically is much colder. In dry adiabatic conditions, the temperature is about 10 degrees lower on Säntis than at Schwägalp. That means that the temperature on Säntis is often forecasted too high and the temperature at Schwägalp too low.
By applying our downscaling technique, we can significantly reduce these systematic errors. The whole computation is done on the fly, so whenever you query the temperature at a certain location, we don't just deliver that one single precomputed value, but we look at the whole atmosphere and deliver the best possible forecast for you!
WMS via Cesium showing the difference between our downscaled temperature forecasts (on the left) and raw model output (on the right) in the Alps.