A New Technique for Subdaily Week-3 Forecast Updates Using the ECMWF Monthly Model

Sub-Seasonal (Weeks 3-5) forecasts of temperature and precipitation are becoming more frequently demanded by clients in the B2B (Business-to-Business) space. These demands are often in the form of a deterministic forecast of weekly aggregated departures from normal for a respective field (i.e., Temperature or Precipitation). The most commonly used Numerical Weather Prediction (NWP) system for these forecasts are the ECMWF Monthly Model, which updates twice a week (Monday and Thursday). There are no updates for the ECMWF Week 3 forecast between Tuesday and Thursday (AM), which can sometimes result in the Monday-initialized run becoming stale. This is especially true during times when the ECMWF Ensemble Prediction System (EPS) shows increased volatility in it’s 11-15 day period forecast (between Tuesday-Thursday). In order to better serve the needs of subseasonal forecast users during times when the ECMWF Monthly model is not updating, a statistical technique has been developed to produce twice-a-day updates (between 0Z Sunday and 12Z Thursday). These twice-a-day updates are derived from: i) Clustering data from the original Monday ECMWF Week 3 forecast and ii) 0Z and 12Z forecast updates from the ECMWF EPS. This talk will discuss the methodology behind the technique, verification statistics, and an unexpected use case of the technique.