Towards Advancing the MJO and 1-30-day Weather Forecasting in the Fully Coupled NGGPS

- Status Update

PI: Joshua Xiouhua Fu (U. Hawaii)

Collaborators: Yuejian Zhu, Cecelia Deluca, Wei Li, and Xiaqiong (Kate) Zhou

August 29 2018

MJO Sensitivity study in GEFS retrospective forecasts to support SubX project: • Improved Stochastic Physics, • 2-tired SST approach • and Scale-aware Cumulus Parameterization

References: • Zhou, X., Y. Zhu, D. Hou, Y. Luo, J. Peng and D. Wobus, 2017: The NCEP Global Ensemble Forecast System with the

EnKF Initialization, Wea. Forecasting, Vol. 32, 1989-200 • Zhu, Y., X. Zhou, M. Pena, W. Li, C. Melhauser and D. Hou, 2017: Impact of Sea Surface Temperature Forcing on Weeks

3 & 4 Forecast Skill in the NCEP Global Ensemble Forecasting System, Wea. Forecasting, Vol. 32, 2159-2173 • Zhu, Y., X. Zhou, W. Li, D. Hou, C. Melhauser, E. Sinsky, M. Pena, B. Fu, H. Guan, W. Kolczynski, R. Wobus and V.

Tallapragada, 2018:Towards the Improvement of Sub-Seasonal Prediction in the NCEP Global Ensemble Forecast System (GEFS)”, JGR, p6732-6745

• Li, W., Y. Zhu, X. Zhou, D. Hou, E. Sinsky, C. Melhauser, M. Pena, H. Guan and R. Wobus, 2018: Evaluating the MJO Forecast Skill from Different Configurations of NCEP GEFS Extended Forecast, Climate Dynamics (in review process)

Configuration of Four Sensitivity Experiments

Label Stochastic Physics Scheme

SST Cumulus Scheme

STTP STTP Relax to Climatology

SAS

SPs (SPSA)

SPPT+SHUM+SKEB Relax to Climatology

SAS

SPs+CFSBC (SPSB)

SPPT+SHUM+SKEB Initial analysis+ bias corrected CFS forecast

SAS

SPs+CFSBC+CNV (SPSC)

SPPT+SHUM+SKEB Initial analysis+ bias corrected CFS forecast

Updated SAS

• Mode system – NCEP GEFSv11 • Resolution – 34km (0-8 days); 55km(8-35days) • Ensemble members – 21 members (20 perturbed forecast + control forecast) • Period – May 1st 2014 – May 26 2016 • Frequency – every 5-day at 00UTC

Part I: Overall Skill Assessment

ACC RMSE Intensity

(Standard Deviation)

PresenterPresentation NotesTo assess the impacts of different model configurations on MJO hindcasts in the GEFS, the ACC, RMSE, and intensity of MJO in terms of OLR, U850, U200, VP850, VP200 during two-year hindcasts are estimated with the method used in Fu et al. (2017, JGR-O), which is an alternative MJO skill measure complementary to WH index. With this kind of skill measure, it is found that the MJO skills steadily increase from the STTP, SPSA, SPSB, to SPSC. The dynamical variables (U850, U200, VP850, VP200) have systematically higher skills than the OLR. However, the model MJO intensity decreases steadily as lead time increases in all 4 configurations.

ACC RMSE Intensity

(Standard Deviation)

PresenterPresentation NotesContinued.

Part II: Mean States

Means

OLR Means, Biases, and Mean-differences among Four Different Experiments (STTP, SPSA, SPSB, SPSC)

Biases

1st 10-d 2nd 10-d 3rd 10-d

SPSC-SPSB

SPSB-SPSA

SPSA-STTP

PresenterPresentation NotesThe two-year mean OLR in the SPSC (and other three experiments) indicates that the simulations are systematically drier than the observations, which may explain why the MJO intensity decreases steadily with increased lead times in all hindcasts.

Part III: Spatial-Temporal Distributions

of Skill Changes in the Four Sensitivity Experiments

OLR ACC Differences among Four Experiments as Functions of Initial Dates and Lead Times

SPSA-STTP SPSB-SPSA

SPSC-SPSB SPSC-STTP

PresenterPresentation NotesTo explore the specific impacts of individual processes, the skill changes as a function of initial dates and lead times are assessed:The skill increases induced by the new stochastic physics and SST forcing appear, respectively, before day-18 and after day-15. The skill increases with new cumulus scheme look randomly. Overall, the cumulative effects of new stochastic physics, new cumulus scheme and SST forcing are near-uniform skill increases before day-15, and concentrated in three separated periods after day-15. The underlying reasons deserve further study.

OLR TCC and Differences among Four Experiments at Different Lead Times

STTP

1-10day 11-20day 21-30day

SPSA-STTP

SPSB-SPSA

SPSC-SPSB

SPSC-STTP

PresenterPresentation NotesThis plot illustrates the spatial distributions of skill changes at three lead periods (1-10day, 11-20day, and 21-30day ) by different physical processes. Apparent skill gains of the SPSC relative to the STTP are robust over tropical Indian and Pacific Oceans, in particular after day 11.

VP850 TCC and Differences among Four Experiments at Different Lead Times

STTP

1-10day 11-20day 21-30day

SPSA-STTP

SPSB-SPSA

SPSC-SPSB

SPSC-STTP

PresenterPresentation NotesThis plot illustrates the spatial distributions of skill changes at three lead periods (1-10day, 11-20day, and 21-30day ) by different physical processes. As for the OLR, the SPSC shows robust skill gains relative to the STTP over tropical Indian and Pacific Oceans, particularly after day-11. The contributions from the stochastic physics and SST are apparently larger than the scale-aware cumulus parameterization.

VP200 TCC and Differences among Four Experiments at Different Lead Times

STTP

1-10day 11-20day 21-30day

SPSA-STTP

SPSB-SPSA

SPSC-SPSB

SPSC-STTP

PresenterPresentation NotesThis plot illustrates the spatial distributions of skill changes at three lead periods (1-10day, 11-20day, and 21-30day ) by different physical processes. As for the OLR, the SPSC shows robust skill gains relative to the STTP over tropical Indian and Pacific Oceans, particularly after day-11. The contributions from the stochastic physics and SST are apparently larger than the scale-aware cumulus parameterization.

Preliminary summary • MJO related sensitivity study has been carried out from NCEP GEFS 35-d

retrospective forecasts for four difference configuration. • The main MJO skills are similar to EMC’s evaluation – SPSC has best skill score for

all lead times. • Individual variables - OLR, U850, U200, VP850 and VP200 have also indicated

better ACC and less RMS error for SPSC configuration, however, the intensity of these variables shows weaker than “control” with increasing forecast lead-time

• Through OLR analysis, all three configurations have systematic drier with increasing forecast lead-time, this could explain why the intensity is weaker.

• TCC has also indicated excellent improvement from all enhanced sciences (SPs, 2-tired SST and SA-convection parameterization.

• In order to dig out specific processes responsible for specific improvements, in-depth analyses are still going on (e.g., the regression analyses and degree of organization etc.).

• Ongoing – 2-d and 3-d regression for all these tropical key variables.

Towards Advancing the MJO and 1-30-day Weather Forecasting in the Fully Coupled NGGPS��- Status Update Slide Number 2Slide Number 3Slide Number 4Slide Number 5Slide Number 6Slide Number 7Slide Number 8Slide Number 9Slide Number 10Slide Number 11Slide Number 12Slide Number 13Preliminary summary