Global assessments of the NCEP Ensemble Forecast System using altimeter data

Forecasts of 10-m wind (U10) and significant wave height (Hs) from the National Centers for Environmental Prediction (NCEP) Ensemble Forecast System are evaluated using altimeter data. Four altimeter missions are selected for the assessment in 2017 that provide a total of 33,229,297 data points matching model state to altimeter measurement. This large quantity of data allows the investigation of the error as a function of forecast ranges, quantiles, and location. Special attention is given to the comparison between the arithmetic mean of the ensemble forecast and the deterministic forecast control run. Error metrics are selected to quantify and separate the systematic and scatter components of the error. Results indicate a large reduction of the scatter errors (SCrmse) in the ensemble mean compared to the control run; more evident for U10, where large SCrmse of 5 m/s associated with strong winds at mid-latitudes beyond forecast day 7 drops to 3 m/s for the ensemble mean. This benefit is transferred to Hs and the largest SCrmse of 1.8 m at the control run is reduced to 1.3 m for the ensemble mean. Although the overall forecast skill of the ensemble forecast is improved, the extreme quantiles of Hs and U10 beyond forecast day 5 tend to underestimate the observations. This implies a need for bias correction algorithms applied during post-processing of the NCEP ensemble products. We conclude that for reliable wind and wave forecasts beyond 7 days at mid- and high latitudes, it is essential to use ensemble forecast products, especially when associated with extratropical areas in the Southern Hemisphere.