Global Parallel Experiment Log

Experiment Name

GSI hybrid retrospective 31 July 2006-5 November 2006

Parallel Slot

PRY

Control Slot

GFS

Start date of parallel experiment

2006/07/31

End date of parallel experiment

2006/11/05

Implementation date if any

May 1, 2007

Environmental Modeling Center scientists

whole GlobMob

Abstract (including Motivation, Hypothesis and Method)

Analysis changes
--Replace Spectral Statistical Interpolation (SSI) analysis system with Gridpoint Statistical Interpolation (GSI) analysis system.
--operational AIRS data (only center spot AIRS) used until 2006083012,new every field of view form of AIRS data used after that date
Model changes
--Modify sigma coordinate system to a hybrid sigma-pressure coordinate system
--Updated GFS radiation module
Questions? Contact Glenn.White@noaa.gov or Mark.Iredell@noaa.gov

Background links

Evaluation of parallel results

Summary of final statistics by Glenn White

Forecast statistics by Russ Treadon

Precipitation scores over continental United States by Russ Treadon

Verification against sondes and aircraft by Russ Treadon and Suru Saha

Effect of GSI hybrid on hurricane forecasts by Morris Bender

Forecast tracks of hurricanes (AVN0 opr. GFS; PARY GSI hybrid GFS; G2006 GFDL 2006 with AVNO; GHYB GFDL 2006 with PARY) by Morris Bender

Forecast tracks in East Pacific of hurricanes AVN0 opr. GFS; PARY GSI hybrid GFS; G2006 GFDL 2006 with AVNO; GHYB GFDL 2006 with PARY by Morris Bender

Verification of 2m-T, 2m-RH, 10m-wind and precip over United States by Helin Wei

Summary of statistics by Glenn White

Systematic errors and time-mean differences by Glenn White

Conclusion

GSI hybrid improved hurricane tracks for both GFS and GFDL models in 3-5 day forecasts in the Atlantic. (Morris Bender)
GSI hybrid significantly reduced day 1 rms error against analysis in height, temperature and tropical winds.
Skill of extratropical height forecasts very similar in operational and GSI hybrid
Skill of precipitation forecasts very similar
Tropical wind forecasts improved in GSI hybrid
Bias of 200 hPa wind over East Asia against radiosondes substantially reduced
Forecast heights fit radiosondes significantly better in GS hybrid

Specific comments

GSI hybrid affects vertical structure in tropics.
GSI hybrid much less oceanic stratus clouds in first 6 hours.
GSI hybrid pbl height less over ocean, more over land.
GSI hybrid improvements in stratosphere
GSI hybrid more transient eddy kinetic energy than operational GFS.
GSI hybrid more downward surface net heat flux in 6 hour forecasts. Operational appears to have too much downward already.
GSI hybrid more stable surface layer (T2m-Tsfc) (Glenn White)
* 2-m temp is warmer compared with OPS. This is
consistent with Suru's fits over North America.
* The warmer temp reduces the nighttime cool
bias over the east CONUS found in OPS. However, it increases
the morning session warm bias over the west CONUS.
* 2-m RH is lower.
* Lower RH reduces the daytime high RH bias but
introduces the nighttime low RH bias over the east CONUS. It
increases the low RH bias over the west CONUS
* Little change is found in 10-m wind speed for both cases.
* The improvement in the precip is very impressive
* The threat score is higher for all threats.
The bias score is a lot closer to 1 (very impressive). Much
more improvement is found in the east CONUS than in the west.
Overall I like the performance of PRY especially the improvement of
precipitation. (Helin Wei)

Global Parallel Home Page

Disclaimer