The weather becomes unpredictable on large timescales when the forecast error is magnified. When we create a forecast for the next couple of hours based on the latest observations, the behaviour of weather becomes more predictable. However, the information rapidly becomes out of date so we need to create predictions instantaneously to deliver the information before it gets old - this requires a great deal of computational power.
If we process all available measurements quickly enough to create a forecast for the next hour or two, we are able to deliver short term minute-by-minute forecasts with unbelievable accuracy.
Minute-by-minute precipitation data
All raw radar data comes from national services around the world which operate a network of multiple radar stations. All those data are then merged. Because weather radar data are extremely noisy, we need to eradicate all artifacts that can be confused for precipitation. This process is relatively complicated as it is not always easy to distinguish between noise and rain in radar images.
When the initial data preparation is finished, we obtain a time series of past precipitation. This can be used to feed our model which predicts any changes for the next few hours. As the output is highly localised, it is important to check the exact location to receive an accurate minute-by-minute forecast.
Forecast for each minute
Thanks to our technological and meteorological experience, we are able to run our model for most of the inhabited places around the world. You can see the results of our minute-by-minute forecast on the homepage for our weather API or our mobile app. The minute-by-minute forecast is also available using our standard and flexi subscription plans.
Our minute-by-minute forecast includes precipitation type and intensity, as well as start and end times for precipitation for the next hour or two, specific to a GPS location. To achieve the best results, it is essential to provide the exact location of the user.