NOTE: The nws-changes mailing list is no longer active. The list archives are made available for historical reasons.
==================================================== Notice of INTENT to Change: New Sea-Surface Temperatures in Meso Eta Brief description of change(s): This change will replace the use of the 1 degree by 1 degree Reynolds optimum interpolation (OI) analysis of sea-surface temperature (SST) in the Meso Eta Model runs with a new SST analysis. EMC's Ocean Modeling Branch has designated this the Real Time Global SST (RTG_SST) which is performed on a 50 km grid and had been referred to in earlier messages as the 2DVAR. The following changes are also being made at this time but they do not effect the model evolution in any way. We are a) refining our restart capability which will be used for the 84 hour extension in February, b) adding output to the lateral boundary condition files that are generated each hour of the forecast for the nested Meso runs "threats" and allow the sea-level pressure reduction in the nest to match that in the parent, and c) correcting the procedure for diagnosing the 10 m winds. Reason for Change(s): EMC has been working towards this SST change since October, but it has been elevated to crisis status following the meetings of a Tiger Team recently assembled to examine the poor Meso Eta guidance for the DC and Baltimore areas for the winter storm of 30 December. Following discussions between NESDIS and EMC on 12 January, it was decided that the new high-resolution daily RTG_SST analysis with inhomogeneous (variable) background-error covariance function provided the best possible solution. This new SST analysis is expected to be a major improvement over the current operational SST analysis products (NCEP's Reynolds analysis and NESDIS' MCSST) and a slight improvement over the one studied during the Tiger Team evaluation which used a homogeneous (fixed) covariance length of 222 km. The new analysis includes in-situ observations and high-resolution (4 km) satellite SST retrievals, generating a daily analysis with variable-scale (100 to 450 km) covariance functions. These data are also used in the Reynolds OI, however it uses 7-days of data, with 1 degree super-obing resulting in a very smooth analysis. The NESDIS analysis product is at 50 km but uses the satellite retrievals only. The RTG_SST analysis uses only the most recent 24 hours of data, is performed on a higher-resolution (50 km) grid than the Reynolds and produces a sharper depiction of the warm core of the Gulf Stream and associated gradients. It also is the only analysis to properly depict the colder shelf water - a feature that was noted by the Tiger Team to be critical in getting an accurate model prediction for the recent east coast winter storm. The Meso Eta Model is capable of providing the proper mesoscale response to ocean forcing but only if it has the proper mesoscale structures depicted in the SST. EMC's parallel run (Etax) at the time was running with the fixed correlation length version of the RTG_SST and now uses the variable length version. The change to the 10 m wind diagnosis is in response to an inadvertent change that was implemented when operations were being established on the IBM SP following the fire in the Cray C-90. We had been testing ways to correct a low bias in 10 m wind speeds in the mountainous west and a code change that resulted in stronger winds everywhere was mistakenly implemented on the IBM. This change had the desired effect over elevated terrain of increasing the diagnosed 10 m wind speeds but the change had not been localized to elevated terrain and had a negative effect over areas of lower elevation. This can be seen by comparing the statistics for the same month before and after November 1999 at our web page of near surface verifications: http://www.emc.ncep.noaa.gov/mmb/research/nearsfc/nearsfc.verf.html . A localized variant of the correction to 10 m winds has been in the 22 km parallel since November. Verifications have been added to the above web page. Again, the computation of the 10 m wind is purely diagnostic so this change does not effect the model prediction. Schedule for change: Final testing: real-time parallel testing plus re-runs of two recent cases: 3 and 20 December 2000 . CAFTI Approval: None - this has been designated a crisis change Expected implementation: tentatively 30 January 2001 - subject to satisfactory reruns and review of results on 25 January. Description of testing: The new RTG_SST analysis algorithm has been evaluated by EMC's Ocean Modeling Branch (OMB) by running year-long sequences of comparative analyses and verifying against observations from moored buoys in both the Atlantic and Pacific Oceans. Further, since January 6, 2001, it has been used as the lower boundary condition over the ocean for the Eta Model in the parallel tests. These real-time 22 km parallel runs using the RTG_SST analysis are being produced twice daily with forecasts made to 60 hours. On 20 December this run began by using a RTG_SST analysis with a fixed 222 km covariance length. This is what was in place in the parallel for the winter storm of 30 December and produced less than half the precipitation in DC and Baltimore that the operational Meso run produced. Since 6 January, the RTG_SST analysis has used a variable covariance length that is sharper where there are tight climatological temperature gradients and is broader where these gradients are weaker. Eric Rogers has established 3 sites: 1) where the SST fields can be compared http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapll.sst/ and 2) where the parallel runs can be compared to AVN and NGM http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapll/ and 3) where the Eta vs Etax runs can be compared with more fields and more detailed focus on both east and west coasts http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapllsup/ . Geoff Manikin has prepared many plots for the Tiger Team which compare the operational Eta vs AVN vs NGM runs and the Eta vs Etax runs where the Etax is the run using RTG_SST SST. These can be found at http://www.emc.ncep.noaa.gov/mmb/research/tiger/tiger.html In addition, two re-runs are planned: a) of the 3 December storm for which model guidance predicted heavy snow for Raleigh, NC which failed to verify and b) the run from 12z 20 December which predicted too much coastal development and too much precipitation. These cases will be rerun with the inhomogeneous RTG_SST. In both of these cases, the operational Meso Eta used the NESDIS MCSST. Statistical summaries: Because of the crisis nature of this implementation, EMC's OMB will prepare a TPB and will brief CAFTI on the new RTG_SST analysis once the implementation is completed. Information on the 10 m wind change is available at the bottom of http://www.emc.ncep.noaa.gov/mmb/research/nearsfc/nearsfc.verf.html Specific examples of note: See 30 December storm. Anticipated impact on forecasts: NWS COMET scientists Bill Bua and Stephen Jascourt have prepared a "training vignette" on the Meso Eta Model, SST and the 30 December storm. It is under final review and will be appearing at http://www.emc.ncep.noaa.gov/cases/20001230_snow/index.htm Overall, the new SST will provide a more accurate lower boundary over water for the model and this will improve the marine boundary layer evolution etc. The need to get this in as soon as possible stems from the fact that the winter season is when there are the greatest temperature differences between air and water and that is when the fluxes are greatest. Differences between Reynolds and RTG_SST are greatest within the Gulf Stream and along the East coast and Gulf coast where there is cooler shelf water depicted in the RTG_SST. Differences are slight over the Pacific. Because the changes are towards cooler coastal shelf temperatures but a warmer Gulf Stream, there is no systematic response in east coast storm depth. We have seen coastal storms at various times be weaker when influenced by shelf waters and deeper when influenced by the Gulf Stream. Generally, coastal storms are farther east (further offshore) in the parallel Etax than in current operations. NOTE: There were other aspects of the initial conditions that contributed to the poor guidance from the Meso Eta Model of the 30 December storm that are not addressed by the SST replacement. We hope to have them resolved in time for the Spring 2001 change package. Points of contact for further information on the changes: Meso Eta: Eric Rogers (301-763-8000 ext7227, eric.rogers@xxxxxxxx) RTG_SST: Jean Thiebaux (301-763-8000 ext7216,Jean.Thiebaux@xxxxxxxx) Field Evaluation: This is ongoing at NCEPs HPC & MPC and the Etax parallel will likely be mentioned in their discussions between now and implementation. Peter Manousos (HPC SOO) will lead evaluation efforts overall and specifically for the retrospective cases. Review of results is targeted for 25 January. For the 30 December storm at least, the most intense scrutiny was performed by the Tiger Team whose membership included SOO's Steve Zubrick of the Sterling, VA office and Rich Grumm of State College, PA office as well as co-leads Jeff McQueen of NWS' Office of Science and Technology and Ralph Petersen Deputy Director of EMC, plus Russ Schneider of NCEP's Storm Prediction Center, Keith Brill of NCEP's Hydrometeorological Prediction Center, Geoff Manikin of EMC's Mesoscale Modeling Branch, Paul Kocin of The Weather Channel, Brian Wimer of Accuweather, Mike Kaplan of North Carolina State University and Lance Bosart of SUNY at Albany. Their final report is due in early February. Approvals: NCEP Director: Date: EMC Director: Date: Future changes: Requests for Meso Eta guidance to cover Day 3 have been received and will require an extension of 24 hours from 60 to 84. This is planned as an internal implementation within NCEP in February. These are being run off our parallel now and can be viewed at http://www.emc.ncep.noaa.gov/mmb/mmbpll/etapll_ext/ We expect that the 84 hour extension grids and BUFR soundings will be available via ftp. Another notice will be forthcoming. The tuning of the 3DVAR, assimilation of observed precipitation and upgrades to land-surface physics are planned to be implemented in Spring 2001.
nws-changes archives: