1 / 28

Assessment of the CFSv2 real-time seasonal forecasts for 2014

Assessment of the CFSv2 real-time seasonal forecasts for 2014. Wanqiu Wang, Mingyue Chen, and Arun Kumar CPC/NCEP/NOAA. Relevance. Diagnostics/monitoring of CFS real-time forecasts. Real-time skill assessment Improve forecast through post-processing Impact of initial condition

donnak
Télécharger la présentation

Assessment of the CFSv2 real-time seasonal forecasts for 2014

An Image/Link below is provided (as is) to download presentation 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. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Assessment of the CFSv2 real-time seasonal forecasts for 2014 Wanqiu Wang, Mingyue Chen, and Arun Kumar CPC/NCEP/NOAA

  2. Relevance Diagnostics/monitoring of CFS real-time forecasts • Real-time skill assessment • Improve forecast through post-processing • Impact of initial condition • Systematic errors

  3. Outline • SST indices • Spatial maps • Anomaly correlation skill • Prediction of sea ice extent minimum

  4. 1. SST indicies

  5. SST indices Nino34 Nino34 • Stronger amplitude of both positive and negative phases • Delayed transition of ENSO phases at longer lead-time DMI • Failed to reproduce positive DMI in 2012 • Good forecast for 2010 negative DMI. MDR • Too warm for Jan-Sep 2014 for 3 and 6 month lea. Too cold after Oct 2014. • Underestimate the amplitude of warm anomalies during Jan-Jul 2011 DMI MDR

  6. CFSv2 Nino34 SST raw anomalies • Nino3.4 SST is near normal during 2013 and has been positive since April 2014. Forecast warm anomalies in 2014 too strong.

  7. CFSv2 Nino34 SST with PDF correction • The correction reduced amplitude of forecast warm anomalies in 2014.

  8. 2. Spatial maps Anomaly = Total – Clim1999-2010 CFSv2 forecast is at a lead of 20 days or so. For example, forecast for Jun-Jul-Aug is from initial conditions of May 1-10th. Impacts of atmospheric initial conditions should be largely removed.

  9. Forecast for MAM 2014 • Slightly weaker SST anomalies in the Tropical Pacific. • Reasonable rainfall anomaly patter.

  10. Forecast for MAM 2014 • Good forecast of negative T2m anomalies North America. Most of forecasted warm anomalies over Eurasia also verified.. • Failed to forecast negative anomalies around north pole. Forecasted tropical and mid-latitude positive Z200 anomalies are consistent with the observed.

  11. Forecast for JJA 2014 • Reasonable forecast SST anomaly pattern, but with erroneous anomalies over some areas (e.g., and .to the east of Argentina) • Tropical rainfall anomalies in the Pacific and Atlantic is good, but forecast anomaly pattern in Indian Ocean is wrong.

  12. Forecast for JJA 2014 • Observed T2m anomalies are quite spotty, which are largely not captured by CFSv2. • CFSv2 reproduced the observed tropical positive Z200 anomalies, but failed to capture observed anomalies in the mid-high latitudes.

  13. Forecast for SON 2014 • CFSv2 captured the overall SST warmth and spatial anomaly pattern.. • Both observed and forecast precipitation anomalies are weak. Forecast wet positive anomalies in tropical western Pacific are not well defined in the observation.

  14. Forecast for SON 2014 • CFSv2 failed to forecast the cold anomalies in the northern hemisphere. • Consistently, CFSv2 did not reproduce observed Z200 in the northern mid-high latitudes.

  15. Forecast for DJF 2014/2015 • CFSv2 reproduced slightly weaker SST anomalies in the tropical Pacific, but stronger negative anomalies in the Atlantic. • CFSv2 predicted the observed wet anomalies in the western Pacific, but failed to reproduce the dry anomalies in the equatorial Indian Ocean, and wet anomalies in the northern tropical Atlantic.

  16. Forecast for DJF 2014/2015 • The observed warm anomalies in the Eurasian mid-high latitudes and western North Amrica, and cold anomalies in the eastern North America are not forecasted by CFSv2. • CFSv2 failed to capture the observed Z200 anomalies pattern in the mid-high latitudes.

  17. 3. Anomaly correlation skill

  18. Pattern correlation over tropical Pacific 20S-20N • Tropical Pacific rainfall correlation is low, especially after 2014 summer. • The observed amplitude is small and thus difficult to predict. • The SST forcing is also not very strong, providing limited predictability of rainfall anomalies.

  19. Pattern correlation over tropical Indian Ocean 20S-20N • Tropical Indian Ocean rainfall correlation is low, possibly due to the weak observed amplitude and weak SST forcing.

  20. Pattern correlation over tropical Atlantic 20S-20N • SST correlation is generally less than 0.4. • Rainfall correlation is near 0.6 in for AMJ and JJA 2014, but becomes very low ASO 2014. • MDR SST index anomaly erroneously negative for DJF 2014/2015, where observed values is about 0.3. (slide 16)

  21. Pattern correlation over NH 20N-80N • T2m skill was reasonable for FMA and MAM 2014, but became relatively low thereafter. • Precipitation skill was changeable OND being the lowest in 2014. • Moderate PNA skill for ASO 2014 to NDJ 2014/2015, becoming near zero for DJF 2014/2015 consistent with spatial map (slide 16).

  22. 4. Artic September sea ice extent

  23. CFSv2 Obs=5.28

  24. CFSv2 Obs=5.28

  25. CFSv2 Obs=5.28

  26. CFSv2 predicted sea ice extent for September 2014 (106 km2) Obs=5.28 • CFSv2 raw data contains large errors • Bias correction based on 1997-2010 hindcasts helps but corrected sea ice extent is still too large for March to August forecasts • Bias correction based on more recent years further reduced the error but forecasts from May to Jul remain too large.

  27. Differences in sea ice extent between CFSR and NASA Team analysis (106 km2) • Significant jumps in 1997 and 2008 • The resulting time-dependent systematic bias in forecast is difficult to remove • The differences were changeable after 2008. Systematic forecast errors still exist even with a shorter period for bias correction

  28. Differences in sea ice volume between CFSR and PIOMAS analysis (103 km3) • Changeable differences depending on year and month • The larger SIV during 2014 summer may be another reason (in addition of SIE, slide 27) for the predicted SIE (slide 26). Courtesy of Thomas Collow

More Related