GSI Hybrid EnKF-3DVAR Upgrade

GSI Hybrid EnKF-3DVAR Upgrade• Components

– GPS RO bending angle rather than refractivity– Inclusion of compressibility factors for atmosphere– Retune SBUV ob errors – fix bug at top– Update radiance usage flags– Prepare for monitoring NPP and Metop-B (Use of data?)– Add GOES-13 data– Add Severi CSBT radiance product– Satellite monitoring stats code included in Ops.– New Sat wind data and QC– EnKF hybrid system – Update to current version of trunk– New version of Forecast model

• Restructured to include options for Semi-Lagrangian & NSST ModelUpdated postprocessor

• CAPE, CIN, & Lifted Index calculated from virtual temperature• Ability to output GRIB2 directly• Ability to read GFS spectral coefficient file which will be exercised during this

implementation to save post run time• 4 new variables for Fire Wx, 8 for Wind Energy, 3 for severe weather• Switch to use CRTM 2.0.2 lib and coefficient files

2

Motivation from Var

• Current background error covariance (applied operationally) in VAR – Isotropic recursive filters– Poor handle on cross-variable covariance– Minimal flow-dependence added

• Implicit flow-dependence through linearization in normal mode constraint (Kleist et al. 2009)

• Flow-dependent variances (only for wind, temperature, and pressure) based on background tendencies

– Tuned NMC-based estimate (lagged forecast pairs)

3

EnKF/3DVar hybrid

c''

o1T''

o'1

VarT''

Var 21

21 JJ HxyRHxyxBxx

J : Penalty (Fit to background + Fit to observations + Constraints)

x’ : Analysis increment (xa – xb) ; where xb is a backgroundBVar : Background error covarianceH : Observations (forward) operatorR : Observation error covariance (Instrument +

representativeness)yo’ : Observation innovationsJc : Constraints (physical quantities, balance/noise, etc.)

B is typically static and estimated a-priori/offline

6

Hybrid Variational-Ensemble

• Incorporate ensemble perturbations directly into variational cost function through extended control variable– Lorenc (2003), Buehner (2005), Wang et. al. (2007), etc.

111

ef

't

'o

1T't

'o

1Te

'f

1T'ff

'f 2

121

21 HxyRHxyLxBxx ,J

K

kkk

1

e'f

't xxx

f & e: weighting coefficients for fixed and ensemble covariance respectivelyxt: (total increment) sum of increment from fixed/static B (xf) and ensemble B k: extended control variable; :ensemble perturbationL: correlation matrix [localization on ensemble perturbations]

ekx

EnKFmember update

member 2 analysis

high resforecast

GSIHybrid Ens/Var

high resanalysis

member 1 analysis

member 2 forecast

member 1 forecast

recenter analysis ensemble

Dual-Res Coupled Hybrid

member 3 forecast

member 3 analysis

Previous Cycle Current Update Cycle

Status

• Coding/parallel scripting completed– Except possible changes due to NPP and GOES-15

data• Optimized version of post being used – so no

delay in delivery of post fields expected.• Some data handling issues being worked.• Parallel without relocation and ozone hybrid from

June 1, 2011- Oct. 7, 2011– Problems with extrapolation of ozone into data void

regions (South Pole)– Initial hurricane location degraded (forecast improved)

11

Status

• Modification to system– Reinstate hurricane relocation– Uncouple ozone analysis from hybrid

• New parallels started– Rerun beginning June 1, 2011 (will start when backup

machine returns)– Real time catch up starting Nov. 1, 2011 (up to Nov.

15, 2011)• Results that follow are from previous parallel

– Significant changes to meteorological fields not expected (except hurricane relocation)

12

500mb Anomaly Correlation

13

1000mb Anomaly Correlation

14

200mb Tropical winds

15

850mb Tropical winds

16

12-36 hour precipitation scores

17

36-60 hour precipitation scores

18

60-84 hour precipitation scores

19

20

21

22

Hurricane track

23

More complete diagnostics

24

http://www.emc.ncep.noaa.gov/gmb/wd20rt/experiments/prd12q3h/vsdb/

http://www.emc.ncep.noaa.gov/gmb/wd20rt/experiments/prd12q3r/vsdb/

Retrospective

Real Time catch up

25

Cost

• Analysis/GSI side– Minimal additional cost (~2 nodes)

• Reading in ensemble (3-9 hour forecast from previous cycle)• Some additional computations in J calculations• 1 additional field• Coding complete/in place

• Additional “GDAS” Ensemble (T254L64 GDAS)– EnKF-based perturbation update

• Cost comparable to current analysis [ 8-10 nodes, <40 minutes]– Includes ensemble of GSI runs to get O-F and actual ensemble update

step• Work ongoing to optimize coding and scripting

– 9hr forecasts, needs to be done only in time for next (not current) cycle

Concerns

• Rain/no-rain for 12-36 hour forecast (spin-up?)

• Changes in initial clouds– Part of spin-up?

• Tasked to attempt to incorporate NPP data.• Biggest concern is computer resources to catch

up to real time.

26

NPP

• ATMS data available – Needs some development because of different size FOVs

for different channels (code change or preprocessor)– Quality control check out– Bias correction spin up

• CrIS instrument to be turned on in Dec.– Major calibration of CrIS – Jan. 15– Should be similar to AIRS and IASI – significant

development (code changes) unlikely– Quality control check out– Bias correction spin up

27

NPP

• Tentative Schedule– RFC to NCO 1/3/12– Complete development for ATMS ~1/15/12– Major calibration update for CrIS 1/15/12– Check out quality control/bias correction 2/6/12– Short off-line test ends 2/16/12– NCO parallel testing begins 2/16/12– Implementation 4/1/12– Any unexpected finding that delays the schedule will stop

the implementation of NPP.• No chance for completion of NPP impact studies prior

to NCO parallel. (JCSDA)

28