Christina Holt 1,2,3, Ligia Bernardet 1,2,3, Mrinal Biswas 1,4 1 Developmental Testbed Center 2 NOAA ESRL Global Systems Division, Boulder CO 3 University.

Christina Holt1,2,3, Ligia Bernardet1,2,3, Mrinal Biswas1,4 1Developmental Testbed Center

2NOAA ESRL Global Systems Division, Boulder CO3University of Colorado CIRES , Boulder CO

4National Center for Atmospheric Research, Boulder, CO

Advancing Microphysics Parameterizations in the Hurricane Weather Research and Forecasting

(HWRF) System

WWOSC 2014Thursday, August 21, 2014

Montreal, Canada

Acknowledgements:

Rob Fovell, UCLA

Outline

The role of the DTCThe Community HWRFImplementation tests performed at

DTCCurrent collaboration with research

communityOngoing & preliminary findings

Future work

2

The Developmental Testbed Center…is a distributed facility (NCAR and NOAA)

where the NWP community can test and evaluate models and techniques for use in research and operations

DTC’s GoalsLink Research and Operational CommunitiesSpeed transition of research results in operationsAccelerate improvement in NWPDevelop and test promising new NWP techniquesProvide an opportunity for NWP community to

perform cycled tests of model and data assimilation systems

3

DTC Strategies to promote HWRF R2OCode Management

Create and sustain a framework for NCEP and the research community to collaborate and keep HWRF code unified

User and developer support Support the community in using an operational hurricane model

DTC Visitor Program – some approved projects involving HWRF

Development of an HWRF diagnostics module to evaluate intensity and structure using synthetic flight paths through tropical cyclones (J. Vigh - NCAR)

Diagnosing tropical cyclone motion forecast errors in HWRF (T. Galarneau - NCAR)

Improving HWRF track and intensity forecasts via model physics evaluation and tuning (R. Fovell - UCLA)

Evaluation of two HWRF microphysics/radiation configurations with remote-sensing data (S. Bao – CCU)

Testing and Evaluation Perform tests to assure integrity of community code and

evaluate new developments for potential operational implementation 4

HWRF: NOAA operational hurricane model

HWRF Components•WRF-NMM model (NMM)•Pre-Processor (WPS and prep_hybrid)•Vortex improvement•Data assimilation (Gridpoint Statistical Interpolation)•Coupler (NCEP)•Ocean (Princeton Ocean Model for Tropical Cyclones-TC)•Post-Processor (UPP)•Vortex Tracker (GFDL)

Community version released each year in August

Reflects the current year’s operational HWRF capabilities 5

Physics Parameterization

Cumulus (d01 and d02)

Simplified Arakawa Schubert with shallow convection

Microphysics Ferrier for the tropics

Planetary Boundary Layer

GFS (Hong and Pan 1996,Vickers and Mahrt 2004, Gopal et al 2013) (90:90:30 s)

Surface Layer GFDL (modified)

Land Surface Model GFDL slab model

Radiation GFDL (60:60:60 min)

6

27 km

3 km

9 km

POM-TCPOM-TC

Introduced an empirical

dependency of BL height on Critical

Richardson Number

Introduced an empirical scaling factor, gfs_alpha,

that reduces mixing by limiting the

momentum eddy viscosity

HWRF v3.5a Atmospheric Configuration

Advects individual species

Advects total condensate

Partial Double Moment

Single Moment

6-classes w/ graupel

Cloud, rain, and snow

Coupled with RRTMG

(mixing ratio & concentration)

Supplies only MR to

radiation scheme

Tested for HWRF:

red. int. bias at long lead times

in ATL

Used in Operational

HWRF

HWRF T&E: Thompson/RRTMG

Thompson vs. Ferrier

RRTMG vs. GFDLUses internal assumptions to compute

concentrations given MR

Uses assumptions consistent

with Thompson to

compute concentrations given MR

Interacts with clouds

Interacts with clouds*

Tested for HWRF:

red. int. bias at long lead times in the

ATL

Used in Operational

HWRF

Thompson mp for Hurricane Sandy improved track forecasts

-A. Chakraborty, India CAOS,

NCAR/RAL visitor

Thompson mp for Hurricane Sandy improved track forecasts

-A. Chakraborty, India CAOS,

NCAR/RAL visitor

DTC: Coupling of the Thompson with RRTMG

DTC: Coupling of the Thompson with RRTMG

7

* Inconsistencies between Ferrier and GFDL radiation: clouds too transparent to radiation

-Rob Fovell, UCLA, HFIP Participant

* Inconsistencies between Ferrier and GFDL radiation: clouds too transparent to radiation

-Rob Fovell, UCLA, HFIP Participant

T/RRTMG improves track for AL but degrades for EP

T/RRTMGF/GFDLTrack error

North Atlantic Eastern North PacificIntensity Bias

T/RRTMG increases intensity for shorter lead

times, decreases for longer lead times 8

T/RRTMG creates negative intensity bias in EP

Case study: Daniel 04E

Thompson too fast and northward; reason under investigation

Ferrier runs similar at 5-day lead time

Bulk statistics for Daniel 04E and case study show that Thompson takes track to N

Storms are in area of strong SST gradientNorthern tracks leads to cool SST under storm and low bias

SS

T (

C)

9

DTC VSP: Rob Fovell and Peggy Bu, UCLAHWRF’s GFDL radiation scheme is deficient with

respect to cloud-radiative forcing (CRF)… see Bu et al. (2014)With proper CRF, the horizontal extent of the near

surface wind field increasesStorms with larger wind fields advect more

planetary vorticity northward (beta drift) and move fasterWesterly moving storms pull northward, and are

ahead of the observed stormWorking Hypotheses:

CRF and gfs_alpha work together to broaden the wind field too much, causing increased beta drift northward

Once the storms are over cold water to the north, they lose intensity 10

Results from DTC visit

T/RRTMG has narrower/weaker storm

F/RRTMG has slightly weaker, broader storm

Courtesy of Rob Fovell

11

30-36 hr 72-96 hr

T/RRTMG has broader/weaker storm

gfs_alpha = 0.7

New DTC Physics ExperimentsSeries of forecasts started from the same

initial conditions using HWRF v3.5a Varying microphysics scheme and

gfs_alphaAssess storm size and track changes

Are the forecasts with real storm settings consistent with the working hypotheses from the ideal studies?gfs_alpha Microphysics Radiation

0.7 Ferrier GFDL0.4 Thompson RRTMG0.4 Ferrier RRTMG0.7 Thompson RRTMG0.7 Ferrier RRTMG 12

Storm Tracks for Daniel (EP 2012)

Thompson 0.4Thompson 0.7Ferrier 0.4Ferrier 0.7Ferrier/GFDL/0.7Best Track

• Tracks diverge around 60 hrs• Thompson fcsts take a northward

track• gfs_alpha = 0.4 has a more

accurate track• Ferrier is less sensitive to gfs_alpha

13

14

Averaged over 60 -126 hrs

Thompson 0.4Thompson 0.7Ferrier 0.4Ferrier 0.7Ferrier/GFDL/0.7Observed

Average Radial Velocity

Averaged over first 60 hrs

SummaryThe findings in the real case study reflect

the idealized findings obtained by DTC Visiting ScientistsCRF + larger gfs_alpha broaden the wind

field at shorter lead timesBroader wind field (all else equal) leads to

northward trackInvestigation of other factors is needed to

determine the intensity/size relationships with physics

Further testing would be necessary to identify potential alternatives for operational implementation

15

Moving forwardDTC will continue to work

with visiting scientistsAssess the impact of

gfs_alpha, radiation, and microphysics combinations on forecasts

Provide resources to assess the potential for future R20

Look at radiation budgets, and the sensitivity of track, intensity, and structure to CRF

Physics for high-resolution

Framework: multiple

moving nests

LSM, storm surge,

inundation

Ongoing

development

DA / Initialization

16

Community Support by DTC-Model freely available and supported -Upcoming release this month•2014 operational capability•Idealized capability•Support for all basins•Unified Python scripts with EMCImproved model for 2014•Updated moving nests, ocean, and initialization improved forecast•Opened opportunities for major future developments