Global Systems Division contributions to Warn-on-Forecast. Steve Koch Director, ESRL Global Systems Division. February 18, 2010. Topics and Tasks. Best approaches to radar data assimilation Storm-scale ensemble predictability studies MADIS Metadata and QC improvements
Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
Director, ESRL Global Systems Division
February 18, 2010
Function of DFI: Remove high-frequency oscillations (particularly, gravity waves) from the initial state for the forecast
3-km High-Resolution Rapid Refresh (HRRR)
Four IOPs: IOP1, 4, 10, 12
Models: 2 WRF-ARW (Thompson and Ferrier), MM5 (Schultz), RAMS
over the American River Basin
Reliability curves and the Brier skill score improved.
Internal frequency histograms changed.
Error bars: 90% confidence intervals
NOAA Must Make HPC a Top Priority Investment !
Innovate or become obsolete …
NOAA’s ability to meet its mission via HPC is falling further behind by any measure. The science will go where there is computing capability to advance it.
GSD is researching Graphical Processor Units (GPU)
Meteorological Assimilation and Data Ingest System (MADIS) supporting WoF
Surface Data Density Before MADIS
Surface Data Density After MADIS
MADIS Computing Environment supporting WoF
and Quality ControlIT Architecture
Port the existing GSD MADIS software to an integrated NWS TOC and NCO distributed environment, with a supporting backup and research-to-operation test environment at GSD.
Current Mesonet Stations with 5-minute Data - 690 supporting WoF
Current Mesonet Stations with 15-minute Data - supporting WoF13,810
Current Stations + UrbaNet + ASOS/AWOS + APRSWXNET + AWS with 5-minute Data by 2011 – 14,574
Blue – current Red – UrbaNet Brown – ASOS/AWOS Black – APRSWXNET and AWS
Despite the tremendous number of sites being ingested, QC’d, and distributed through MADIS, the data are still largely distributed like “oases and deserts”. Adaptive multi-scale analysis techniques that utilize the temporal information (GSD STMAS multi-grid 3Dvar) are required.
Challenge: Non-uniform data distribution