Toward robust fine-scale decadal precipitation forecasts through dynamically consistent subsampling

Abstract. Reliable decadal predictions of regional precipitation are critical for managing water-resources and developing climate services, yet they remain a major challenge. To address this gap, we present a 5-step framework that integrates recent advances in decadal predictions of large-scale sea-level pressure (SLP) modes to enhance prediction skill of precipitation at a fine scale resolution. We first identify key atmospheric indices controlling precipitation variability over France, including the winter and summer North Atlantic Oscillation (NAO), the winter West Atlantic Pressure Anomaly, and the summer Mediterranean-Scandinavia index. These indices are predicted through an improved post-processing method applied on the multi-model Decadal Climate Prediction ensemble. The resulting decadal forecasts of the indices are used to select dynamically consistent members from a large uninitialized climate model ensemble, thereby avoiding initial drift from decadal climate predictions. The selected forecasts are then statistically bias-corrected and downscaled to an 8-km grid, providing relevant predictions for local scale and impact studies. The last step of the framework is the skill evaluation: over France, winter precipitation forecast based on the NAO achieves significant Anomaly Correlation Coefficient across 78 % of grid cells. Summer skill, though weaker, improves notably when combining NAO with the Atlantic Multidecadal Variability (significant over 48 % of grid cells). This approach offers a transferable pathway toward actionable, fine scale hydroclimate information at the decadal scale, potentially useful for climate services. The methodology is adaptable to other regions and variables, offering promising opportunities for improving decadal-scale hydroclimate predictions.