TJ10.4 A seamless system for medium range, subseasonal and seasonal probabalistic forecasts of energy demand

Monday, 7 January 2013: 4:45 PM
Room 6A (Austin Convention Center)
Judith A. Curry, Georgia Institute of Technology, Atlanta, GA; and J. Belanger, M. Jelinek, V. Toma, and P. Webster

An overview is presented of the operational energy demand forecasting system that we have developed, called OmniCast, which is based upon ECMWF's hierarchical prediction system: a high resolution Deterministic Atmospheric Model (0.125o) with forecasts for 1-10 days, an Ensemble Prediction System (0.25o – 0.5o) with forecasts of 1-15 days, a Monthly product (0.5o) with forecasts 1-32 days, and Seasonal forecasts (1.5o) for 1-7 months. The ensemble is corrected to adjust for bias and distributional errors using model hindcast (reforecasts) and real time verification data from reanalyses and in situ observations, with the calibration solution varying with forecast lead time. The ensemble is interpreted in the context of pdfs and thresholds for extreme events. A new “Bayesian” clustering approach is used for monthly and seasonal forecasts that selects a high-predictability cluster based upon initial verification of each ensemble member by subsequent observations or subsequently initialized forecasts. An objective confidence assessment is presented for monthly and seasonal forecasts, based on historical predictability analysis (based on lead time, month, teleconnection regimes, and region), phase and amplitude of the MJO and ENSO, ensemble intercorrelation and spread, and recent prediction verification statistics. Recent verification statistics are presented, along with examples of graphical representations of the forecast information.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner