Quick and Dirty Validation of GEFS Reforecast Calibrated Precipitation Forecasts

Validation of GEFS reforecast calibrated precipitation forecasts against NARR analyses, focusing on reliability, upper and lower tercile outcomes, and Brier skill scores for different lead times. Challenges with NARR analyses in capturing snowfall events in the northern Great Plains are discussed, suggesting the use of alternative datasets like station data, Higgins/CPC data, or CCPA data set to improve reliability. Questions about the reliability discrepancies and potential factors affecting skill degradation are raised, with a willingness to share code for further analysis.

• GEFS
• Reforecast
• Validation
• Precipitation
• NARR

chaelaza Follow

Uploaded on Sep 12, 2024 | 0 Views

Download Presentation

Please find below an Image/Link to download the presentation.

The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. Download presentation by click this link. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.

Presentation Transcript

1. (Quick and dirty) validation of GEFS/reforecast calibrated precipitation forecasts Tom Hamill tom.hamill@noaa.gov (303) 497-3060

2. Methods Rank analog method described briefly in the submitted BAMS article (here) and the 2006 MWR article (here). Validation against 32-km NARR analyses. I d note that NARR analyses seem to be particularly bad in picking up snowfall events in the northern Great Plains, which degrades winter skill. Evaluated for 1985-2010 forecasts over CONUS.

3. Reliability, upper and lower tercile, days 5-10 (120-240 h lead) Comparatively, in 2004 article, for version-1 reforecasts validated against station data, 6-10 day RPS skill was ~ 0.06.

4. Reliability, upper and lower tercile, days 7-14 (168-336 h lead) Comparatively, in 2004 article, for version-1 reforecasts validated against station data, 8-14 day RPS skill was ~ 0.03.

5. What about upper decile? Any skill?

6. Brier skill scores, days 5-10

7. Brier skill scores, days 5-10

8. Sample case

9. Thoughts / questions I think skill is in no small measure degraded by NARR analyses. Ought to use other data Station data Higgins/CPC data Recent CCPA data set of Yan Luo I m curious why I don t get better reliability. I have been getting excellent reliability for shorter-lead forecasts using the same method (e.g., here, slide 17). I suspect that it s still related to training sample size. Happy to share code if it helps, though I m off soon for 2 weeks foreign travel.