Temporal Context Conditioning for Seasonality-Aware Precipitation Nowcasting of High-Intensity Rainfall

The paper introduces the Time-Aware Small-Attention U-Net (TA-SmaAt-UNet), which enhances the SmaAt-UNet architecture by incorporating lightweight temporal conditioning layers that utilize cyclical encodings of time-of-day and time-of-year. Experiments conducted on KNMI radar precipitation data demonstrate that this approach significantly improves the model's performance for high-intensity rainfall events and enhances the representation of seasonal variability in precipitation predictions. This advancement is relevant for practitioners as it offers a method to improve the accuracy and reliability of precipitation nowcasting, particularly in scenarios involving rare weather events.