AI tool opens the door to even better precipitation forecasts
The production of SMHI’s precipitation forecasts is now being enhanced with new technology. The introduction of an AI model marks a step forward in the process of creating these forecasts.
The AI model, CHIMP – developed at Chalmers University of Technology – has been trained at SMHI to improve the analysis of current precipitation conditions. The model has been trained with four years of satellite data, while weather radar data have been used as ground truth.
“The AI model can provide an analysis of the current precipitation situation over a much larger area than our radar network covers, and with a quality that represents a significant improvement across the entire region. A better analysis opens the door to more accurate forecasts,” says Bengt Rydberg, researcher in meteorological satellite applications at SMHI.
Training the AI model shows clear results
In addition to the fact that SMHI's forecast model works in a larger area than that covered by the Swedish radar network, ground-based radar technology itself has limitations. For example, in mountainous regions the topography can obstruct the radar’s line of sight. The researchers therefore excluded the Swedish mountain range from the training data. Nevertheless, once trained, the AI model produced better-quality precipitation analyses – both in the mountainous regions and across the broader forecast area.
Visible improvements in the precipitation forecast
The AI model is now part of the input used for SMHI’s weather forecasts. Users of the SMHI weather app can already see improvements in the “Precipitation” map feature, where the latest updates show fewer disturbances in the visualization of recent precipitation and a smoother transition to the forecast for upcoming precipitation.
The AI model CHIMP is now part of the input to the Precipitation map feature in the SMHI weather app. Following the latest update of the weather app, users can see the current precipitation analysis across the entire forecast area, along with a smoother transition from the current analysis to the precipitation forecast.
Screen recording from the Precipitation map feature in SMHI’s weather app, 4 November 2025.
The research and development of AI-based precipitation analysis from satellite data is part of a larger SMHI initiative to improve precipitation forecasts, partly funded by the Swedish Civil Contingencies Agency.
