Complimentary Assessment of Forecast Performance with Climatologically Based Approaches

Abstract

Precipitation is a critical weather parameter that is highly variable in space and time, exhibits sharp gradients while is often dominated by zero values. It is also dependent on the underlying climate of a region, thus aggregating statistics regarding forecast performance over large climatologically-heterogeneous regions must be done with care. On the other hand, NWP (Numerical Weather Prediction) models provide grid-box average precipitation, while rain gauges represent a point measurement, leading to potentially significant representativeness errors. The WMO (World Meteorological Organization) recommends procedures to follow for the verification of precipitation, including accounting for the climate differences between stations. These can be addressed by scaling the quantity to be verified by the local climate of the station. The newly developed (by ECMWF (European Center of Medium Range Forecasts) SEEPS (Stable Equitable Error in Probability Space) verification score differentiates the forecast performance into precipitation intensity categories based on local station climatology and is thus simultaneously equitable and more stable. The complimentary use of SEEPS with metrics that focus on extreme events, such as the Symmetric Extremal Dependence Index (SEDI) that is adjusted to the climatological distribution of precipitation at each location, enables assessment of locally important aspects of the forecast while providing a reliable performance measure. This approach is applied for a period of one year over Greece for forecasts from two modelling systems, a 7 km regional model (COSMO) and the ECMWF global model.