6.3 Global Forecast Dropout Prediction Tool in Support of the NCEP Model Evaluation Group (MEG)—A Collaborative Project between JCSDA/NESDIS & NWS

Wednesday, 13 January 2016: 2:15 PM
Room 345 ( New Orleans Ernest N. Morial Convention Center)
V. Krishna Kumar, NOAA/NESDIS/Center for Satellite Applications and Research/JCSDA, COLLEGE PARK, MD; and E. Maddy, J. C. Alpert, and S. A. Boukabara

Coordination efforts are currently underway between JCSDA/NESDIS Directed Research Team (DRT) and NCEP/EMC Model Evaluation Group (MEG) to contribute to the MEG project focusing attention on forecast system and product quality on a daily basis with feedback into the model development cycle. The MEG evaluates the daily performance of NCEP forecast and analysis, identifies model biases and conducts post-mortem studies of high-impact, poorly-forecast events. NCEP and Regional Centers, the NWS offices, and private customers interact through the MEG to alert them of model biases and issues and provide a forum for users to report problems they have seen in NCEP global and regional model analyses and forecasts.

The goal of this project is to contribute directly to the MEG synoptic perspective and verification with emphasis on the veracity of the initial conditions (IC) and analysis from a standpoint of the quality of all satellite and conventional observations over the globe. NCEP Global Forecast System (GFS) Forecast Skill Team (called the “Dropout Team” which was the forerunner of the MEG) modernize software to depict and isolate assimilation errors and reveal observation errors and identify quality control issues in terms of their origins focusing on poor model skill score events.

A global forecast dropout prediction tool (GFDPT) developed by the Dropout team, the JCSDA/NESDIS Community Observation Assessment Tool (COAT) and Independent Assessment Tools (IAT) will be transitioned from Research to Operation (R2O) to benefit MEG and prototype use in an operational environment. GFDPT will have three components: (i) prediction of a dropout with 5 day lead time, (ii) detection of actionable volumes where extreme differences between GFS and ECMWF are found, and (iii) diagnosis of classes of identified conventional and non-conventional observations over a volume and demonstration of the feasibility of calculating radiance from model output to compare with that from co-located instrument observations. Such diagnostics will enhance the MEG collaboration between operational forecasters at NCEP Centers/WFOs and model developers and to alert the occasional forecast busts, track the location of large analysis and forecast errors and identify the suspected conventional and satellite data sources that resulted in large analysis differences.

We will provide functional examples of the prediction and diagnosis components of the GFPDT applied to a few recent GFS forecast dropouts.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner