42
DYNAMO Webinar Series Dynamics of the Madden-Julian Oscillation Field Campaign Climate Variability & Predictability Application of DYNAMO/AMIE observations to validate and improve the representation of MJO initiation and propagation in the NCEP CFSv2 Joshua Xiouhua Fu (University of Hawaii) & Wangqiu Wang (NCEP) Wednesday, July 23 @ 2pm University of Hawaii at Manoa

DYNAMO Webinar Series Dynamics of the Madden-Julian Oscillation Field Campaign

  • Upload
    wilson

  • View
    35

  • Download
    0

Embed Size (px)

DESCRIPTION

DYNAMO Webinar Series Dynamics of the Madden-Julian Oscillation Field Campaign. Application of DYNAMO/AMIE observations to validate and improve the representation of MJO initiation and propagation in the NCEP CFSv2 Joshua Xiouhua Fu (University of Hawaii) & Wangqiu Wang (NCEP) - PowerPoint PPT Presentation

Citation preview

Page 1: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

DYNAMO Webinar SeriesDynamics of the Madden-Julian Oscillation Field Campaign

Climate Variability & Predictability

Application of DYNAMO/AMIE observations to validate and improve the representation of MJO initiation and propagation in the NCEP CFSv2

Joshua Xiouhua Fu (University of Hawaii) & Wangqiu Wang (NCEP)

Wednesday, July 23 @ 2pm

University of Hawaii at Manoa

Page 2: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

3

THE ULTIMATE GOAL of this project: To improve the prediction skill of Madden-Julian Oscillation (MJO) in the national climate forecast model (NOAA/NCEP CFSv2)

· ANALYZE the MJOs observed during DYNAMO period

· REVIEW operational models’ forecasting of the DYNAMO MJOs

· ASSESS the capability of CFSv2, GFS, and UH models

in MJO forecasts

QUANTIFY the impacts of air-sea coupling on MJO forecasting

· Experiment for cumulus parameterizations and SST uncertainty

Categorize MJO types: coupled and uncoupled

Outline

Page 3: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Observed MJO events during DYNAMO period

Page 4: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

SST and MJO-filtered OLR Anomalies in DYNAMO Period

Oct-MJO

Nov-MJO

SST (shading); OLR (contours)

IOP

Five MJO events

Thanks-giving TC duringNov. MJO

Only two MJO events(Nov. & Mar.) with robustcoherent positive SST anomalies leading the convection

Air-sea ‘coupling strength’ varies with individual MJO events

4

Page 5: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Prediction of the observed MJO by operational models

Page 6: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Good forecasts of two successive MJO events

IC: Oct_17 IC: Nov_07

Courtesy of NCEP MJO Discussion Summary led by Jon Gottschalck et al.

Bad forecasts of Sep. primary MJO event

5

IC: Sep_20 IC: Sep_12

Page 7: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

IC: Oct_03 IC: Oct_10

Maritime Continent Barrier

Weak Intensity

IC: Oct_24 IC: Nov_27

IC: Mar_05 IC: Mar_129

SlowPropagation

Page 8: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Prediction by GFS, CFSv2 and UH models

Page 9: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Nov. MJO initiation in CFSv2&UH models

IC: Nov_04

6

• Both CFSv2 and UH models capture the development of November MJO.

• Propagation in UH model quite realistic.

• Propagation in CFSv2 too slow.

Shadings: ObservationContours: Forecast

Red arrows: Observed minimum values. Green arrows: Forecast minimum values.

OLR anomalies

Page 10: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Extended-range forecasts of Nov. MJO initiation

7

Page 11: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Forecasts Initialized on Nov. 18, 2011

OBS

CFSv2

UH

13

• Propagation in UH model quite realistic.

• Propagation in CFSv2 too slow.

Shadings: ObservationContours: Forecast

OLR anomalies

Page 12: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Acknowledgment: Observational surface flux data from Revelle during DYNAMO period are provided by Chris Fairall, Simon de Szoeke, Jim Edson, and Ludovic Bariteau

Page 13: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

MJO Skills of GFS, CFSv2, and UH during DYNAMO

GFS: 13 days CFSv2&UH: 25/28 days

CFSv2&UH MME: 36 days

Fu et al. (2013)

(Wheeler-Hendon Index, Lin et al. 2008)

14

Page 14: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

MJO prediction in CFSv2 hindcasts (1999-2010)

Page 15: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Composites forecast for each initial phase

Observation CFSv2

Initial Phase 1 2 3 4 5 6 7 8

Obs(CFSv2-obs)

6.9(-1.7)

6.7(-1.2)

7.4(-1.2)

7.6(-0.5)

6.7(-1.3)

7.2(-2.0)

7.2(-1.2)

6.4(-1.3)

Phase speed (Degree/day)

Initial phases: 1, 3, 5, 7 Initial phases: 2, 4, 6, 8

11Wang et al. 2013. Climte Dyn.

Page 16: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Composite from initial phase 3

OLR (shading); U850 (contours)

Forecast

Observation

12

Page 17: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Bivariate correlation of Wheeler-Hendon index

as a function of target phase (MJO Days)

lead

tim

e (d

ays)

IOAfrica AtlMC WP

10

(Based on CFSv2 1999-2010 hindcasts)

Wang et al. 2013. Climte Dyn.

Page 18: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Courtesy of Owen Shieh

12 UTC Nov 28

November MJO & Thanks-giving TC

(TC05A)

15

Page 19: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Forecasts of GFS, CFSv2 and UH with IC on Nov. 11

Observed and forecasted U850 and OLR averaged for days-13-15

U850 (contours)OLR (shading)

16

Page 20: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Forecasts of GFS, CFSv2 and UH with IC on Nov. 18

U850 (contours)OLR (shading)

Observed and forecasted U850 and OLR averaged for days-13-15

18

Page 21: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

What caused the dramatic differences in MJO prediction between GFS and CFSv2/UH?

• Air-sea coupling• Model physics

What are needed for an improved MJO prediction in GFS and CFS?

Page 22: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Impacts of air-sea coupling on the prediction

Page 23: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Names of Experiments SST Settings

CPL Atmosphere-ocean coupled forecasts.

Fcst_SST (or fsst) Atmosphere-only forecasts driven by daily SST

derived from the ‘cpl’ forecasts.

Pers_SST (or psst) Atmosphere-only forecasts driven by persistent

SST.

TMI_SST (or osst) Atmosphere-only forecasts driven by observed

daily TMI SST.

UH Forecast Experiments with Different SSTs

19

Page 24: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

SST-Feedback Significantly Extends MJO Prediction Skill

Persistent SST CPL

Forecasted Daily SST

Observed Daily SST

Potential

20

Page 25: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Dependence on convection parameterization and SST

uncertainty

Page 26: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign
Page 27: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

NCEP GFS Forecast Experiments

1. Model• Atmosphere-only GFS (May 2011 version)• T126/L64

2. SSTs• Clim • NCDC OI analysis• TMI (TRMM Microwave Imager)

3. Convection parameterizations• SAS (Simplified Arakawa Schubert (Pan&Wu 1995)): Operational CFSv2• SAS2 (Revised Simplified A-S (Han&Pan 2011)): Operational GFS• RAS (Relaxed A-S (Moorthi and Suarez (1999))

4. Forecast runs• Initial conditions: CFSR• Initial dates: 1 Oct 2011 to 15 Jan 2022 (4 runs from 00, 06, 12, 18Z

each day)• 31 target days

21

Page 28: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

22

(Wang et al. 2014)

Page 29: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

23(Wang et al. 2014)

Page 30: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

OLR RMM indexAnomaly Correlation

(Wang et al. 2014)

Page 31: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Differences in forecast q (RAS – SAS2) with TMI SST from 7 Nov 2011

• The lower troposphere above PBL with SAS2 is consistently drier than that with RAS, even before Nov 12 when rainfall rate is small.

• The drier lower troposphere with SAS2 is related to the larger rainfall rate during the first few days, indicating that the SAS2 convection scheme tend to drive the atmosphere to a drier state to maintain the balance between convection and large-scale dynamics

• Establishment of such a drier lower troposphere with SAS2 results in a less strong convection response to the underlying SST anomalies.

24

Why does RAS scheme produce better MJO?

Page 32: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

CFSUse an alternative convection scheme, e.g., replacing SAS2 with RAS Improve SST accuracy with better intra-seasonal and diurnal variability

What can we do to improve MJO prediction in CFS and GFS?

25

GFSUse an alternative convection scheme, e.g., replacing SAS2 with RAS Specify SSTs from another coupled forecast system (e.g., CFS), or couple

GFS to a mixed-layer ocean model.

Page 33: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Categorization of MJO types:

Coupled and uncoupled

Page 34: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Different Roles of Air-sea Coupling on the Oct. and Nov. MJO Events (UH)

Fu et al. (2014) 26

Page 35: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Different Roles of Air-sea Coupling on the Oct. and Nov. MJO Events (GFS)

Oct-MJO

Nov-MJO

Dec-MJO

Need Daily SST Forcing

28

Page 36: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

29

Summary

· Only two of five observed MJOs during DYNAMO have robust coherent positive SST anomalies leading MJO convections.

· The initiation of successive MJO is more predictable than primary MJO. Major MJO forecasting problems include: slow eastward propagation, the Maritime Continent barrier and weak intensity.

· During DYNAMO period, the MJO forecasting skills for the GFS, CFSv2, and UH models are 13, 25, and 28 days. The equal-weighted MME of the CFSv2 and UH reaches 36 days.

· Air-sea coupling is important for MJO forecasting and still has plenty rooms to be improved.

Page 37: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

31

Summary

· The interactions between the Nov-MJO and Thanksgiving-TC have been much better represented in the UH and CFSv2 coupled models than that in the atmosphere-only GFS.

· CFSv2 MJO forecasting may be improved with an alternative cumulus parameterization (e.g., RAS) and more accurate SST prediction.

· GFS MJO forecasting with an alternative cumulus parameterization (e.g., RAS) and SSTs from CFS, or couple GFS to an mixed-layer ocean model.

· Two-type MJOs exist: strongly coupled to underlying ocean or largely determined by atmospheric internal dynamics.

Page 38: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

31

Publications 

Fu, X., J.-Y. Lee, P.-C. Hsu, H. Taniguichi, B. Wang, W. Q. Wang, and

S. Weaver, 2013: Multi-model MJO forecasting during

DYNAMO/CINDY period. Clim. Dyn., 41, 1067-1081.

Wang, W. Q., M.-P. Hung, S. Weaver, A. Kumar, and X. Fu, 2013:

MJO prediction in the climate forecast system version 2 (CFSv2).

Clim. Dyn.

Fu, X., W. Q. Wang, J.-Y. Lee and et al.: Distinctive roles of air-sea

coupling on different MJO events: A new perspective revealed from

the DYNAMO/CINDY field campaign. submitted.

Wang, W. Q., A. Kumar, and X. Fu: Dependence of MJO prediction on

sea surface temperatures and convection schemes. to be submitted.

Page 39: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Group Meeting, Honolulu, Mar 02, 2012

Page 40: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

Group Meeting, Honolulu, Mar 02, 2012

MJO Initiation

MJO-I

MJO-II

MJO-III

One Primary MJO Event

Three Successive MJO Events

Page 41: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign

10S-10N average OLR anomalies (Wm-2)

Observation SAS2 SASNCDC SST Day 12 forecast

RAS

Page 42: DYNAMO  Webinar  Series Dynamics of the Madden-Julian Oscillation Field Campaign