Translating Spatial Precipitation Forecast Errors into Operational Criteria for Dynamic Reservoir Operation

The practical value of precipitation forecasts depends not only on their meteorological skill but also on their ability to support operational decisions. "Dynamic" reservoir operation, here defined as reservoir operation in which storage is proactively managed by using precipitation forecasts to create temporary flood storage before an event, is one of the most demanding applications of precipitation forecasting. While forecast verification metrics are widely used to quantify model performance, their direct interpretation in terms of operational feasibility remains largely unexplored.

This work proposes a forecast-oriented framework that translates spatial precipitation forecast errors into indicators of operational suitability for dynamic flood management. Forecasts are first evaluated over the Lombardy Region (Northern Italy), using two complementary spatial verification metrics: the Precipitation Attribution Distance (PAD), which quantifies the average displacement of precipitation features, and the Precipitation Smoothing Distance (PSD), which characterizes their spatial-scale mismatch. To capture event-specific displacement patterns at the catchment scale, a two-dimensional spatial shift approach is subsequently applied. The resulting displacement vectors are projected onto the principal axes of each reservoir hydrological catchment and normalized by basin dimensions, producing dimensionless indicators that directly relate forecast positional uncertainty to catchment geometry.

The proposed methodology translates spatial forecast errors into an indicator of whether precipitation forecasts are spatially accurate enough to support dynamic reservoir operation. Reservoirs with normalized displacement errors smaller than their characteristic basin dimensions can be distinguished from those where forecast uncertainty remains too large for reliable dynamic operation. The framework therefore discriminates reservoirs according to the effective usability of meteorological forecasts rather than forecast skill alone.

This approach establishes a quantitative bridge between spatial forecast verification and decision-making in flood-risk management. By relating meteorological forecast errors to hydrological catchment extension, the proposed framework provides an objective methodology for assessing whether precipitation forecasts are sufficiently accurate to support dynamic reservoir operation.