Multi-center, multi-model ensemble and its application

Yuejian Zhu Environmental Modeling Center

NCEP/NWS/NOAA

Acknowledgements: Ensemble team and Dr. B. Lapenta

Present for US THORPEX workshop September 20 2012

THORPEX related projects • International projects

– THORPEX-HEPEX – THORPEX-IPY – THORPEX-TPARC – THORPEX-YOTC – THORPEX-TIGGE/GIFS – THORPEX-NAEFS

• National projects – THORPEX-HYDRO – THORPEX (NCEP and ESRL – NOAA THORPEX)

• Ensemble data assimilation • Model error – physical uncertainty • Systematic error – bias correction – NAEFS products • Target observation – WSR

• THORPEX related workshop, regional committee meeting • Best reference:

http://www.emc.ncep.noaa.gov/gmb/ens/THORPEX.html

3

Research and Operational Applications

In Multi-Center Ensemble Forecasting

http://wwwt.emc.ncep.noaa.gov/gmb/ens/index.html

Yuejian Zhu

Acknowledgements: Zoltan Toth (ESRL), Glenn Rutledge (NCDC), Andre Methot (MSC),

Dan Collins, Bo Cui, Richard Wobus (NCEP)

North American Ensemble Forecast System

International project to produce operational multi-center ensemble products

• Combines global ensemble forecasts from Canada & USA

– 40 members per cycle, 2 cycles per day from MSC & NWS

• 6-hourly output frequency

• 1*1 degree resolution (~100km)

• NDGD resolution (~5km, CONUS only)

• Forecasts out to 16 days

• Generates products for

– Weather forecasters

• E.g., NCEP Service Centers (US NWS)

– Specialized users

• E.g., hydrologic applications in all three countries

– End users

• E.g., forecasts for public distribution in Canada (MSC) and Mexico (NMSM)

• Operational outlet for THORPEX research using TIGGE archive

– Prototype ensemble component of THORPEX Global Interactive Forecast System (GIFS)

5

NCEP CMC NAEFS

Model GFS GEM NCEP+CMC

Initial uncertainty ETR EnKF ETR + EnKF

Model

uncertainty/Stochastic

Yes (Stochastic Pert) Yes (multi-physics) Yes

Tropical storm Relocation None

Daily frequency 00,06,12 and 18UTC 00 and 12UTC 00 and 12UTC

Resolution T254L42 (d0-d8)~55km

T190L42 (d8-16)~70km

(d0-d16) ~ 66km 1*1 degree

Control Yes Yes Yes (2)

Ensemble members 20 for each cycle 20 for each cycle 40 for each cycle

Forecast length 16 days (384 hours) 16 days (384 hours) 16 days

Post-process Bias correction

(same bias for all

members)

Bias correction

for each member

Yes

Last implementation February 14th 2012 August 17th 2011

NAEFS Current Configuration

Updated: February 14th 2012

NH Anomaly Correlation for 500hPa Height Period: January 1st – December 31st 2010

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

An

om

aly

Co

rre

lati

on

Forecast (days)

GFS GEFS NAEFS

NAEFS – 10.2d

GEFS – 9.7d

0.6 skill line

GFS – 8.0d

Benefit for forecast: 1. Ensemble mean

will extend 1.7 days forecast ability

2. NAEFS will add additional 0.5 day forecast skill

3. Post process will add another additional

7

Purpose • Improve reliability while maintaining resolution in NWP forecasts

Reduce systematic errors (improve reliability) while

Not increasing random errors (maintaining resolution)

• Retain all useful information in NWP forecast

Methodology • Use bias-free estimators of systematic error

• Need methods with fast convergence using small sample

• Easy implementation for frequency upgraded forecast system

Approaches – Computational efficiency • Bias Correction : remove lead-time dependent bias on model grid

Working on coarser model grid allows use of more complex methods

Feedback on systematic errors to model development

• Downscaling: downscale bias-corrected forecast to finer grid

Further refinement/complexity added

• No dependence on lead time

NAEFS Statistical Post-Process (SPP)

Bias corrected NCEP/CMC GEFS and NCEP/GFS forecast (up to 180 hrs), same bias

correction algorithm

• Combine bias corrected NCEP/GFS and NCEP/GEFS ensemble forecasts

• Dual resolution ensemble approach for short lead time

• NCEP/GFS has higher weights at short lead time

NAEFS products

• Combine NCEP/GEFS (20m) and CMC/GEFS (20m), FNMOC ens. will be in soon

• Produce Ensemble mean, spread, mode, 10% 50%(median) and 90% probability

forecast at 1*1 degree resolution

• Climate anomaly (percentile) forecasts also generated for ens. mean

Statistical downscaling

• Use RTMA as reference - NDGD resolution (5km/6km), CONUS and Alaska

• Generate mean, mode, 10%, 50%(median) and 90% probability forecasts

Current NAEFS SPP System

8

9

Variables Domains Resolutions Total 10/8

Surface Pressure CONUS/Alaska 5km/6km 1/1

2-m temperature CONUS/Alaska 5km/6km 1/1

10-m U component CONUS/Alaska 5km/6km 1/1

10-m V component CONUS/Alaska 5km/6km 1/1

2-m maximum T CONUS/Alaska 5km/6km 1/1

2-m minimum T CONUS/Alaska 5km/6km 1/1

10-m wind speed CONUS/Alaska 5km/6km 1/1

10-m wind direction CONUS/Alaska 5km/6km 1/1

2-m dew-point T CONUS 5km 1/0

2-m relative humidity CONUS 5km 1/0

Total cloud cover?

NAEFS downscaling parameters and products Plan: Q4FY2012 (NDGD resolutions)

All downscaled products based on 1*1 (lat/lon) degree globally

Products include ensemble mean, spread, 10%, 50%, 90% and mode 9

12hr 2m Temperature

Forecast Mean Absolute

Error w.r.t RTMA for CONUS

Average for September, 2007

GEFS raw forecast

NAEFS forecast

GEFS bias-corr. & down scaling fcst.

11

NCEP/GEFS raw forecast

NAEFS final products

4+ days gain from NAEFS

From Bias correction (NCEP, CMC)

Dual-resolution (NCEP only)

Combination of NCEP and CMC

Down-scaling (NCEP, CMC)

11

12

From Bias correction (NCEP, CMC)

Dual-resolution (NCEP only)

Combination of NCEP and CMC

Down-scaling (NCEP, CMC)

NAEFS final products

NCEP/GEFS raw forecast

8+ days gain

12

Tmax Tmin

Temperature

Latest evaluation for CONUS

temperature forecast by apply :

1. Bias correction at 1*1 degree for

NCEP GFS/GEFS, CMC/GEFS

2. Hybrid bias corrected NCEP GFS

and GEFS

3. Apply statistical downscaling for all

bias corrected forecast

4. Combined all forecasts at 5*5 km

(NDGD) grid with adjustment -

NAEFS

Mean abs error Mean abs error

Mean abs error

3.5 days gain

13

Tmax Tmin

Temperature

Latest evaluation for CONUS

temperature forecast by apply :

1. Bias correction at 1*1 degree for

NCEP GFS/GEFS, CMC/GEFS

2. Hybrid bias corrected NCEP GFS

and GEFS

3. Apply statistical downscaling for all

bias corrected forecast

4. Combined all forecasts at 5*5 km

(NDGD) grid with adjustment -

NAEFS

CRPS CRPS

CRPS

7 days gain

14

Dew point T RH

RMS & Spread RMS & Spread

Dew Point T RH

CRPS

CRPS

15

NCEP CMC FNMOC

Model GFS GEM Global Spectrum

Initial uncertainty ETR EnKF (9) Banded ET

Model uncertainty

Stochastic

Yes (STTP) Yes (multi-physics) None

Tropical storm Relocation None None

Daily frequency 00,06,12 and 18UTC 00 and 12UTC 00 and 12UTC

Resolution T254L42 (d0-d8)~55km

T190L42 (d8-16)~70km

L40 ~ 66km T159L42 ~ 80km

Control Yes Yes No

Ensemble

members

20 for each cycle 20 for each cycle 20 for each cycle

Forecast length 16 days (384 hours) 16 days (384 hours) 16 days (384 hours)

Post-process Bias correction for

ensemble mean

Bias correction for

each member

Bias correction for

member mean

Last

implementation

February 14th 2012 August 17th 2011 September 14 2011

NAEFS/NUOPC Configuration

Updated: February 14 2012

17

PQPF forecasts from various global ensembles and combination

18

Raw NCEP

NAEFS + FNMOC Stat. corr.

NAEFS

Combined NCEP – CMC (NAEFS) show further increase in skill (6.2d)

Addition of FNMOC to NAEFS leads to modest improvement (6.7d)

Raw NCEP ensemble has modest skill (3.4d)

Statistically corrected NCEP ensemble has improved skill (4.8d)

0.5 CRPS skill

Value-added by including FNMOC ensemble into NAEFS T2m: Against analysis (NCEP’s evaluation)

Winter evaluations – first try

19

Verification against observation (NH) Raw ensemble forecast, Period: 11/17-12/21/2010

Courtesy of Normand Gagnon (CMC/MSC)

2-meter temperature

10-meter wind speed

T2m analysis difference accumulation (out to 20101026 ~ 10 days)

Challenge: large surface temperature analysis variation from center to center

Good - after adjustment Bad – after adjustment

Still not good – after adjustment

Perfect – don’t need adjustment

NCEP NCEP

NCEP

NCEP

FNMOC

FNMOC

FNMOC FNMOC

Example for combine current NAEFS and downscaled SREF

Future Plan for Improving NAEFS

Probabilistic Products

• Improving global probabilistic products – Adding FNMOC global ensemble

• NAEFS-LAM – Combine NCEP/SREF and CMC/REFS

• Extend additional variables and other regions – Pending on RTMA’s development

– Currently applied to CONUS and Alaska only

– Future: Hawaii, Guan and Puerto Rico

• Adding post processed regional ensemble for downscaling probabilistic products – SREF for Northern American products

• Improving exist bias correction method – Adjust decaying coefficients – varied by lead time, domain and

season

• Adjust 2nd moment – Possible to get help from BMA