GFSv16: Further Advancements to the UFS Medium Range Weather Application in 2021

Fanglin YangNOAA/NWS/NCEP/EMC

Vijay Tallapragada, Daryl Kleist, Arun Chalwa, Jason Levit, Russ Treadon, Jeffery Whitaker, Jack Kain, Lucas Harris, Jian-Wen Bao, Jun Wang, George Gayno, Jongil Han, Weizhong Zheng, Helin Wei, Shrinivas Moorthi, Valery Yudin, Ruiyu Sun, Linjiong Zhou, Anning Chen, Rongqian Yang, Xingren Wu, Xiaqiong Zhou, Henrique Alves, Jessica Meixner, Ali Abdolali, Catherine Thomas, Cory Martin, Wansu Wu, Kristen Bathmann, Phillp Peigion, Andrew Collard, Xu Li, Haixia Liu, Yanqiu Zhu, Xiujuan Su, Shelley Melchior, Wen Meng, Hui-ya Chuang, Boi Vuong, Mallory Row, Guang-Ping Lou, Yali Ma, Jiayi Peng, Deanna Spindler, Roberto Padilla, Geoffery Manikin, Alicia Bentley, Logan Dawson, Kate Friedman, Mark Iredell, Hang Lei, Eric Rogers, George Vandenberghe, James Abeles, Gerhard Theurich, Edward Hartnett, Rusty Benson, Farida Adimi

UNIFIED FORECAST SYSTEM (UFS) USERS’ WORKSHOP

July 27-29, 2020.1

GFS.v15 Implementation -- June 12, 2019

First Finite-Volume Cubed-Sphere Dynamical Core (FV3) based GFS& GFDL Microphysics Scheme with Multiple Prognostic Cloud Hydrometers

Configuration:

● High-res: C768 (~13km)

● Data Assimilation: C384 (~25km, 80 member ensemble)

● 64 layer, top at 0.2 hPa

● Dycore: FV3, non-hydrostatic, single precision

● Physics: GFS Physics + GFDL Cloud Microphysics

Outstanding Issues

• PBL inversion

• Cold bias in the lower troposphere in winter

• Surface temperature biases

• Biases in the upper atmosphere

• Error of track forecast for strong storms (initial wind > 65 kts) in the Atlantic basin

GFS.v15 laid the foundation for further improvement of NOAA FV3-based forecast systems

2

3

Change History of GFS Configurations Mon/Year Lev Truncations Z-cor/dyncore Major components upgrade

Aug 1980 12 R30 (375km) Sigma Eulerian first global spectral model, rhomboidal

Oct 1983 12 R40 (300km) Sigma Eulerian

Apr 1985 18 R40 (300km) Sigma Eulerian GFDL Physics

Aug 1987 18 T80 (150km) Sigma Eulerian First triangular truncation; diurnal cycle

Mar 1991 18 T126 (105km) Sigma Eulerian

Aug 1993 28 T126 (105km) Sigma Eulerian Arakawa-Schubert convection

Jun 1998 42 T170 (80km) Sigma Eulerian Prognostic ozone; SW from GFDL to NASA

Oct 1998 28 T170 (80km) Sigma Eulerian the restoration

Jan 2000 42 T170 (80km) Sigma Eulerian first on IBM

Oct 2002 64 T254 (55km) Sigma Eulerian RRTM LW;

May 2005 64 T382 (35km) Sigma Eulerian 2L OSU to 4L NOAH LSM; high-res to 180hr

May 2007 64 T382 (35km) Hybrid Eulerian SSI to GSI

Jul 2010 64 T574 (23km) Hybrid Eulerian RRTM SW; New shallow cnvtion; TVD tracer

Jan 2015 64 T1534 (13km) Hybrid Semi-Lag SLG; Hybrid EDMF; McICA etc

May2016 64 T1534 (13km) Hybrid Semi-Lag 4-D Hybrid En-Var DA

Jun2017 64 T1534 (13km) Hybrid Semi-Lag NEMS GSM, advanced physics

Jun 2019 64 FV3 (13km) Finite-Volume NGGPS FV3 dycore, GFDL MP

17 years No change in vertical resolution

4

GFS V16: Major Upgrades to Forecast Model Model resolution:

Increased vertical resolution from 64 to 127 layers and raise model top from 54 km to 80 km

Physics updates PBL/turbulence: Replaced K-EDMF with sa-TKE-EDMF

Revised background diffusivity as a stability dependent function GWD: Added a parameterization for subgrid scale nonstationary gravity-wave dragRadiation: Updated calculation of solar radiation absorption by water clouds; Updated cloud overlap

assumptions.Microphysics: Updated GFDL microphysics scheme for computing ice cloud effective radiusNoah LSM: Revised ground heat flux calculation over snow covered surface; Introduced vegetation

impact on surface energy budget over urban area

Coupling to WaveOne-way coupling of atmospheric model with Global Wave Model (WaveWatch III)

Coupling to GLDASSpin up land states using CPC Gauge precipitation in the GDAS 00Z cycle

5

GFS V16: Major Upgrades to Data Assimilation

• Local Ensemble Kalman Filter (LETKF) with model space localization and linearized observation operator

to replace the Ensemble Square Root Filter (EnSRF)

• 4-Dimensional Incremental Analysis Update (4D-IAU)

• Turn on SKEB in EnKF forecasts• Update variational QC• Apply Hilbert curve to aircraft data• Correlated observation error for CrIS over sea surfaces and IASI over sea and land• Update aircraft bias correction with safeguard• Assimilate AMSU-A channel 14 and ATMS channel 15 w/o bias correction • Assimilate CSR data from ABI_G16, AHI_Himawari8, and SEVIRI_M08• Assimilate AVHRR from NOAA-19 and Metop-B for NSST• Use CRTM v2.3.0

6

I/O Change and Inline Post

• The size of each forecast history file in binary (nemsio) format increased from 17GB (64L GFS) to 34GB (127L GFS).

• A decision was made to write out history files in netCDF format with compression. compression ratio: 5.3x (33.6 GB to 6.3 GB), with information loss.

• The Unified Post Processor is now run inline with the model forecast to 1) reduce I/O footprint and, 2) retain UPP data precision and accuracy because forecast history files are now in compressed netCDF format with information loss.

Product Changes

○ Add more pressure levels (at 0.01, 0.02, 0.04, 0.07, 0.1, 0.2, 0.7 hPas) ○ new fields : include cloud ceiling, total column and low/mid/high cloud fractions, and

radar reflectivity at 1 km/4 km and 1st/2nd model level above ground.○ Replace filtered Shuell SLP with unfiltered SLP using same ID PRMSL○ Replace legacy synthetic nadir GOES 12/13 with synthetic nadir ABI GOES-R products○ Remove terrain adjustment of bufrsnd profiles○ FSU storm genesis verification○ See PNS for GFS V16 product removal

○ New WAVE Multi_1 computational grid grib2 files■ Northern hemisphere 0p16 to Replace global 30 arc min and all regional grids■ Southern Hemisphere 0p25 to Replace global 30 arcmin■ Arctic polar-stereographic 9km to Replace Arctic 18km■ New global extended 0p25 to Replace global extended 30 arcmin

○7

GFS V16 Implementation and Evaluation Schedule

● NCEP OD science brief: 6 Oct 2020

● Code Handoff to NCO: 9 Oct 2020

● SPA begins work: 12 Oct 2020

● NCO 30-day parallel: 21 Dec 2020

● NCO OD IT brief: 27 Jan 2021

● Implementation: 3 Feb 2021

● Real-time parallel started: May 19, 2020● Retrospective forecasts period: June 2019 ~ May 2020

● EE2 Coordination meeting: 11 June 2020● PNS sent to HQ: 18 June 2020

● Retrospective forecasts complete: August 31, 2020● MEG Evaluation Kickoff webinar: July 23, 2020● Field Evaluation: August 3, 2020 - September 25, 2020● MEG Evaluation Briefing: September 24, 2020● Field Recommendations due: September 28, 2020

8

GFS.v16 Preliminary Results-- Success and Challenges

C768/C384 (13/25km) L127 Configuration Cycled with data assimilationJune 2019 ~ July 2020

Note: EMC Verification and Product Generation Branch Model Evaluation Team, NWS forecast offices, OAR Labs and the user community will provide more comprehensive evaluation of GFSv16, and provide recommendations to the NCEP Office of Director for implementation

9

500-hPa HGT Anomaly Correlation

NH Day-5

NH Die-off

Increase is significant up to day 6

SH Day-5

SH Die-off

10

Precip ETS and BIAS SCORES over the Continental US00Z Cycle, verified against gauge data

11

1 hPa Temperatures : 60N-90N : Dec 2019 – Jan 2020

GFSv15

GFSv16-Beta

GFSv16

GFSv15

GFSv16-Beta

GFSv16

Credit: Craig Long12

Hurricane Track Forecast -- Still a Grand Challenge

AtlanticEastern Pacific

WesternPacific

Dorian Humberto Gabrielle Lorenzo

13

Temperature Biases in the NH Lower Troposphere 5-Day Forecasts fit to ROBS

DJF 2019/20

GFSv15

GFSv16

The large cold bias found in GFS.v15 in NH winter is removed in GFS.v16

14

June-July 2020

However, New cold bias developed in GFS.v16 in NH summer

GFSv16GFSv15

15

Climate Run:QBO, SAO -- w/ Unified GWD

• Operational model (GFS.v15) cannot simulate the QBO

• A QBO-like feature is captured in GFS.v16 “climate” run with the non-stationary GWD physics (v0) included; However, the periodicity is too short, appears to be an downward propagating SAO.

• GWD needs further improvement. The updated parameterization (v1) further improved the upper atmosphere forecast (see Valery Yudin’s presentation). It will be tested and transitioned to operation in GFS.v17/ GEFSv13

Ensemble Spread at Model Top

It is still not well understood why the spread is so large

Increasing damping reduced ensemble spread in short-lead forecast but makes jet streams too weak in medium-range forecast

16

Spurious edge imprints

Gird Imprinting Issue

“Cubic grids based on a separate gnomonic projection for each face of the cube possess a sharp angular discontinuity across each edge and corner. This can lead to unsightly spurious imprints in fields derived from forecasts”

U V

Derived Aviation Product: GTG

Presentation by R. James Purser, and Miodrag Rancic: Eliminating grid imprinting with a Conformal Cubic Overset Grid

Thank You

17