809 Developments in Tools for Monitoring Observation Sensitivity in the Global Forecast System

Tuesday, 14 January 2020
Hall B (Boston Convention and Exhibition Center)
Andrew Eichmann, NOAA, College Park, MD; and J. C. Alpert and K. Kumar

Operational forecasts from global weather models episodically produce abrupt drops in forecast skill (“busts” or “dropouts”) despite the long-term improvement of overall forecast skill. Experiments at National Center for Environmental Prediction (NCEP) and elsewhere have ameliorated dropouts in the Global Forecast System (GFS) (both fv3gfs and legacy spectral versions) hindcasts by using initial conditions (ICs) derived from those of the European Centre for Medium-Range Weather Forecasts model (EC). These results have prompted examination and monitoring of data assimilation (DA) and quality control under the Global Forecast Dropout Prediction Tool (GFDPT) project. We have developed tools using satellite radiance observation diagnostics to identify regions where observations have a high impact on forecast skill. A current project investigates Forecast Sensitivity to Observation Impact (FSOI) techniques applied to the current operational system FV3-GFS to provide additional data monitoring diagnostics. FSOI has been used successfully to select detrimental observations to withhold in Observation System Experiments (OSEs), resulting in improved forecast skill. FSOI metrics are also used to examine the role of detrimental observations in forecast busts as well.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner