The Hurricane Analysis and Forecast System (HAFS)- Stand-Alone

The Hurricane Analysis and Forecast System (HAFS)- Stand-Alone Regional model (SAR) real-time experiments for the 2019 North Atlantic hurricane season and recent developments

2020 UFS users’ workshop

Jili Dong1,4, Bin Liu1,4, Zhan Zhang1,4, Weiguo Wang1,4, Avichal Mehra1, Andrew Hazelton2,3, R. James Purser1,4, Tom Black1, Jeff Beck5,6, Henry Winterbottom1,4, Lin Zhu1,4, Keqin Wu1,4, Chunxi Zhang1,4, Biju Thomas1,4, Dan Iredell1,4, Vijay Tallapragada1, Ghassan Alaka3, Xuejin Zhang3, Sundararaman Gopalakrishnan3, Frank Marks3

1NWS/NCEP/EMC, 2UM/CIMAS, 3OAR/AOML/HRD,4IMSG, 5OAR/GSL, 6CSU/CIRA

1 Outline

2019 HAFS-SAR HFIP real-time demonstration for the North Atlantic - Objectives - Model configuration and real-time set up - seasonal statistics and two high-impact storms (Barry and Dorian)

Recent developments - Dynamics and physics experiments - Increase horizontal resolution to improve intensity forecast - A more uniform grid for HAFS-SAR: Extended Schmidt Gnomonic (ESG) grids

2 HAFS-SAR real-time experiments: objectives

• Demonstrate the capability of the next generation FV3-based TC prediction system HAFS within the UFS framework on convective scale hurricane forecasts in a real-time environment

• Systematically evaluate the performance of the FV3-based stand-alone regional model on Atlantic hurricane forecasts • Diagnose, assess and compare real-time results with operational HWRF/HMON/GFS

• HAFS stand-alone regional model • Help with configuration selection of HAFS system (stand-alone regional and global- nest) • Potential to provide high resolution forecasts with less computing resource

HAFS - SAR HAFS - Global-nest

3 real time in 2019 HAFS-SAR set up: dynamics and physics

• Dynamics: – Timestep: 90 seconds • In each timestep: vertical remapping freq=4 • In each vertical remapping: acoustic freq=5 • Horizontal advection scheme: hord=6 for KE gradient, vorticity+w, pt – More diffusive scheme than global GFS (hord=5) to keep model stable

• Physics HAFS - SAR – PBL: EDMF+HWRF (over water) – Surface: GFS surface+HWRF (over water) – Cumulus convection: None – Microphysics: GFDL MP – Land-surface: GFS – Radiation: RRTMG LW and SW – Microphysics and radiation interaction: on – GWD: convective off; orographic on – NSST

• No DA, vortex init and ocean coupling - One single domain with ~ 3km resolution • IC and BC - Lat: 4.8º S -- 54.6º N – IC is from GFS nemsio - Lon: 115.1º W -- 8.6º W – BC is from GFS grib2 every 3 hours - 64 vertical hybrid levels (sigma-pressure) - model top at 0.2 hPa - lowest model level on ~25m above sfc4 HAFS-SAR track and intensity forecast for 2019 AL season

Better track skill from HAFS

Relative to HWRF Smaller track error from HAFS

Case #

HAFS-SAR real-time: 4 cycles (00, 06, 12, 18Z) daily for 126 hours forecasts; Dong et al. 2020, 02L Barry to 19L Rebekah Atmosphere 5 HAFS-SAR size forecast for 2019 AL season

● HAFS models have relatively low bias in 50 kts and 64 kts wind radii forecast ● Large size error and bias in 34 kt wind Large r34 error radii means overprediction of storm size; from HAFS similar behavior observed in Barry and Dorian (2019)

Dong et al. 2020, Atmosphere

6 HAFS-SAR forecast: Hurricane Barry (2019)

Track composites

Track improvement from HAFS

OBS HAFS-SAR 60-hr fcst

Capture the asymmetric structure when Barry made landfall 7 HAFS-SAR forecast: Hurricane Dorian (2019)

Smaller error from HAFS

Track composites

No Florida Landfall

8 HAFS-SAR 2019 real-time forecast: summary

• For 2019 Atlantic hurricane season • HAFS-SAR performs better in track forecasts than GFS/HWRF/HMON in most of forecast hours (~20% skill improvement to HWRF)

• HAFS-SAR has large initial intensity error; error quickly decreases in 6 -12 hours (possibly spin-up), then increases until 72 hours; relatively small intensity error on D5

• 34 kts radii are consistently overpredicted while 64 kts radii are fine

• HAFS-SAR provides better track forecasts in Barry (02L) and Dorian (05L) (two US-landfalling storms); large track error for Humberto (09L)

• HAFS-SAR and HAFS-global-nest are comparable in track, intensity and size forecasts

Dong, J.; Liu, B.; Zhang, Z.; Wang, W.; Mehra, A.; Hazelton, A.T.; Winterbottom, H.R.; Zhu, L.; Wu, K.; Zhang, C.; Tallapragada, V.; Zhang, X.; Gopalakrishnan, S.; Marks, F. The Evaluation of Real-Time Hurricane Analysis and Forecast System (HAFS) Stand-Alone Regional (SAR) Model Performance for the 2019 Atlantic Hurricane Season. Atmosphere 2020, 11, 617.

9 Recent developments on HAFS-SAR: Dynamics and physics experiments

Horizontal advection schemes Hord = 5 vs. hord = 6

Hord=5 reduces wind radius error

Hord=5 Hord=6 Hurricane Dorian 2019

Hurricane Michael 2018: wind spd at 900 hPa (72 hrs fcst) - Hord=5 degrades track forecast after D3 - Hord=5 improve intensity fcst between D1-D4 Scale-aware cumulus convection

Conv increase track Conv reduces wind error radius error

Hurricane Dorian 2019 Hurricane Dorian 2019

10 Recent developments on HAFS-SAR: Increasing horizontal resolution for HAFS-SAR

• Explore the sensitivity of HAFS intensity forecasts to horizontal resolution • Increasing horizontal resolution from ~ 3km (2019 HAFS) to 2 km and 1.5 km with static HAFS SAR domains: stronger storm predicted with higher resolution

Best track Best track

Hurricane Michael (2018) init at 2018/10/07/18 Hurricane Lorenzo (2019) init at 2019/09/24/18

Stronger storm with higher Stronger storm with higher resolution (e.g. 1.5 km) resolution (e.g. 1.5 km)

Composite reflectivity for Hurricane Michael valid at 2018/10/10/12

Stronger storm with more fine detail convective scale structure

11 HAFS 3 km HAFS 2 km HAFS 1.5 km Recent developments on HAFS-SAR: ESG (Extended Schmidt Gnomonic) grid for HAFS-SAR

Horizontal resolution (m)

Min:2.56 km Min:3.13 km Max: 3.61 km Max: 3.17 km Avg: 3.16 km Avg: 3.15 km

A more uniform grid

Original GFDL grid 3km ESG grid 3km

HAFS-SAR ESG in 2020 HFIP real-time demo: SLP; cycle=2020/07/24/12 30 hrs fcst

- Running for 7 days (168 hrs) forecast for N. Atlantic Hanna - the latest FV3 dycore (ufs-weather-model) and CCPP

- Cumulus convection off; GWD off

- LBC boundary blending to control numerical instability

- no ocean coupling and DA (will be available soon)

12 Please see more in R James Purser talk in Wednesday 4:00 pm (plenary session) Recent developments on HAFS-SAR: ESG (Extended Schmidt Gnomonic) grid for HAFS-SAR

HAFS-ESG: smaller intensity error, better Hurricane Dorian 2019 intensity skill and reduced negative wind bias

HAFS-ESG: stronger storms

CAT5 CAT4 CAT3 kts CAT2 CAT1 TS

Hours since 2019/08/24/12 Hours since 2019/08/24/12 • Track forecasts close to each other • ESG improve on intensity forecast and reduce negative wind bias • Size similar; ESG tend to reduce size error in longer lead times 13 • ESG predicted stronger TC: more cat4 and cat5 HAFS-SAR: future work

- Moving nest for HAFS-SAR (Xuejin Zhang’s talk on Wednesday; potentially vertical nesting in xthe future) - Ocean coupling capability for HAFS-SAR (Bin Liu’s talk next) - Inner-core/satellite data assimilation to improve TC initialization and forecast - Physics and dynamics experiments to find optimized combination of physical parameterization xand numerical schemes - A basin-scale HAFS-SAR ESG for East Pacific and North Atlantic