Search papers, labs, and topics across Lattice.
The paper introduces AIFL, a deterministic LSTM model for global daily streamflow forecasting, addressing the performance gap between reanalysis and operational forecast data. AIFL is trained on 18,588 basins from the CARAVAN dataset using a two-stage approach: pre-training on 40 years of ERA5-Land reanalysis data followed by fine-tuning on IFS control forecasts. Results on an independent test set (2021-2024) demonstrate high predictive skill (median KGE' of 0.66 and median NSE of 0.53), comparable to state-of-the-art systems, with improved reliability in extreme event detection.
AIFL closes the gap between reanalysis and operational weather forecasts for streamflow prediction, achieving state-of-the-art accuracy with a transparent and reproducible pipeline.
Reliable global streamflow forecasting is essential for flood preparedness and water resource management, yet data-driven models often suffer from a performance gap when transitioning from historical reanalysis to operational forecast products. This paper introduces AIFL (Artificial Intelligence for Floods), a deterministic LSTM-based model designed for global daily streamflow forecasting. Trained on 18,588 basins curated from the CARAVAN dataset, AIFL utilises a novel two-stage training strategy to bridge the reanalysis-to-forecast domain shift. The model is first pre-trained on 40 years of ERA5-Land reanalysis (1980-2019) to capture robust hydrological processes, then fine-tuned on operational Integrated Forecasting System (IFS) control forecasts (2016-2019) to adapt to the specific error structures and biases of operational numerical weather prediction. To our knowledge, this is the first global model trained end-to-end within the CARAVAN ecosystem. On an independent temporal test set (2021-2024), AIFL achieves high predictive skill with a median modified Kling-Gupta Efficiency (KGE') of 0.66 and a median Nash-Sutcliffe Efficiency (NSE) of 0.53. Benchmarking results show that AIFL is highly competitive with current state-of-the-art global systems, achieving comparable accuracy while maintaining a transparent and reproducible forcing pipeline. The model demonstrates exceptional reliability in extreme-event detection, providing a streamlined and operationally robust baseline for the global hydrological community.
