Improving Week-Ahead Precipitation Ensemble Forecasts through a Refined Moist Potential Vorticity Perturbation Strategy

      This study evaluates and optimizes the Moist Potential Vorticity (MPV) perturbation method to improve week-ahead (120-hour) ensemble precipitation forecasts over East China. Five diverse precipitation periods from June–July 2019 were selected to compare the performance of the proposed MPV perturbations against ensembles using perturbations from the NCEP Global Ensemble Forecast System (GEFS). A key focus was mitigating pervasive false alarms in extended-range forecasts, largely driven by environmental errors. Building on a baseline MPV strategy (perturbing both convective and baroclinic instabilities), a modified approach was tested that selectively perturbs only the convective instability component (MPV1) via anti-phase adjustments to potential temperature and water vapor. Verification methods, including the Fraction Skill Score (FSS), Bias, Symmetric Extremal Dependency Index (SEDI), ensemble spread, Continuous Ranked Probability Score (CRPS), and reliability diagrams, were used for comprehensive evaluation.

      Results show that the MPV1-only perturbation strategy effectively suppressed false alarms, leading to superior deterministic skill (e.g., higher FSS) and more reliable probabilistic forecasts beyond 48 hours compared to the baseline strategy, which was less skillful than the GEFS-only ensemble in week-ahead predictions. Furthermore, combining the optimized MPV perturbations with GEFS-based lateral boundary condition (LBC) perturbations synergistically enhanced ensemble diversity and improved probabilistic reliability. Thermodynamic analysis indicates that this combination is effective because environmental errors stem from both local prediction and LBCs. Specifically, underestimating stability in the middle layer of the lateral boundary is a key source of environmental error in week-ahead forecasts. Overall, the study confirms the necessity of accounting for environmental errors in ensemble forecasts.