1

HWRF PHYSICS

Young C. KwonEMC/NCEP/NOAA

Hurricane WRF TutorialNCWCP College Park, MD

Jan 14 2014

2

Contents

1.Overview2.Land surface model3.Surface layer physics (air-sea interaction)4.Planetary Boundary Layer5.Convective parameterization6.Micro-physics7.Radiation8.Physics upgrade plan for FY2014

3

Overview

1. At the initial operational implementation, HWRF physics suite was closely following as GFDL hurricane model physics.

2. Some physics are from GFS (PBL, convection), some are originated from NCEP mesoscale model (Micro-Physics) and others are from GFDL (radiation, surface physics, Land surface), and modify to tropical environment.

3. Many aspects of physics have been upgraded, and the 2013 HWRF physics will be covered in this presentation. The proposed 2014 physics upgrades will also introduced briefly.

4

Scheme Descriptions

Ocean model POM-TC(Princeton Ocean Model) is coupled to Atm. Model, HWRF3D POM in ATL; 1D POM in EP and uncoupled other basins

Land model GFDL slab model*

Surface layer physics M-O similarity theory. GFDL based but Cd and Ch upgraded

Planetary Boundary Layer GFS scheme with modification of diffusivity and Ric

Convective parameterization Simplified Arakawa-Schubert scheme with modifications

Explicit MP Ferrier scheme*

Radiation GFDL LW/SW radiation scheme*

*: plan to upgrade at 2014

HWRF model Physics suite

5

6

𝜕𝑇𝜕𝑡

=−𝑢𝜕𝑇𝜕 𝑥

−𝑣 𝜕𝑇𝜕 𝑦

+ 𝑃𝑅𝜔𝜎 +𝐹𝑇+ �̌�

𝐶𝑃

where,

Thermodynamic equation

Time tendency horizontal advection vertical advec. + adiabatic heating

diabatic heating

Diabatic heating: phase change of water – convection, microphysics Radiative absorption/emission – radiation Subgrid vertical mixing – PBL, convection Surface fluxes – air-sea interaction, land surface

H. diffusion

7

𝑇subgrid scale mixing

micro-physical processes

Radiative cooling/warming

subgrid scale convection

Horizontal/vertical advections, horizontal diffusion

Dynamics

physics

8

Land surface modelGFDL hurricane model slab

∂T*/∂t = (-σT*4 - Shfx - Levp + (S+F ))/ρscsd)

9

Verification of HWRF Skin temperature over CONUS(compare to GFS analysis)

10OCEAN

Hurricane

Low level inflow

Upper level outflow

Energy gain from sea surface (sensible and latent heat) Ch

Energy loss by surface friction Cd

Hurricane intensity is proportional to sqrt(Ch/Cd) over ocean – Emanuel(1995)

Air-sea interactions

11

Modified GFS SchemeOriginal GFS

Scheme

Cd: Surface exchange coefficient for momentum

Ck: Surface exchange coefficient for moisture & heat

Km: Eddy diffusivity for momentum

Km: Eddy diffusivity for momentum

Same Observations

Ch before modification

10m Wind speed (m/s)

CdX1

03

10m Wind speed (m/s)

ChX1

03Cd and Ch profile in the current HWRF model

(gray dots)

12

HEXOS data(1996, Decosmo et al)

CBLAST data (2007)

Original HWRF/GFDL model

13

PBL scheme: Parameterize subgrid-scale vertical turbulence mixing of momentum, heat and moisture in the boundary layer. There are two main categories of PBL schemes: local vs non-local mixing scheme.

Local mixing scheme: vertical mixing is proportional to the local gradient., e.g. Mellor-Yamada-Janjic scheme, Blackadar schemeNon-local mixing scheme: vertical mixing is not only proportional to local gradient but also counter-gradient mixing due to large scale eddy, e.g., GFS scheme, YSU scheme.

HWRF model uses GFS PBL scheme, which is non-local mixing scheme. GFS PBL has shown to good performance outside of hurricane regions while PBL height in hurricane area is too deep and too strong mixing compare to observational data. Recent PBL scheme upgrades address this issue significantly.

14

PBL1. First guess PBL height

2. Update using the first guess PBLh

3. Enhance PBLh using updated

4. Momentum diffusivity (Km) is calculated under PBLh

z (1 - z/h) p

5. Moist diffusivity (Kt) is calculated using Prandtl number

Procedures in the operational HWRF PBL scheme

15Hong and Pan (1996)

16

Cd Ch

Km Km

z (1 - z/h) pIntroduce to match Km to obs

Gopalakrishnan et al (2012)

17

α=1.0 α=0.25

Reduction of momentum diffusivity led to shallower PBL height and inflow depth

Gopalakrishnan et al (2012)

18

Variable Critical Richardson number (Vickers & Mahrt, 2003)

PBL z (1 - z/h) p

Motivation: The GFS PBL scheme used in HWRF model has been known to produce too diffusive boundary layer in hurricane condition. Thanks to HRD’s effort to improve the hurricane PBL in HWRF model, the diffusivity and PBL height of HWRF model greatly improved based on composite dropsonde observations (e.g., Gopalakrishnan et al. 2013, JAS; Zhang et al. 2013, TCRR)

However, outside of hurricanes, the GFS PBL behaves quite well and some underestimation of PBL height is reported (Jongil Han, personal communication). Therefore, it may worth trying to revise the current PBL scheme to work well in both inside and outside of hurricane area seamlessly.

19

Critical Richardson number function of Ro (Vickers and Mahrt, 2003)

Hurricane cases

Vickers and Mahrt(2003) Critical Richardson number is not a constant but varies with case by case.

Ric = 0.16(10−7 )−0.18

The magnitude of Ric modifies the depth of PBL and diffusivity, so the Ric varying with conditions would fit both hurricane condition and environments.

20

PBL height difference (new PBL scheme with var Ric – PBL scheme in 2012 HWRF with constant Ric=0.25)

PBL height over the ocean and hurricane area becomes shallower while that over land area becomes deeper

Both configurations have set to 0.5

Hurricane Katia (20110829018+96hr)

21

Convective parameterization:

When grid resolution of a numerical model is too coarse to resolve individual convection, there are need to parameterize the impact of convection to grid scale.Convection does stabilized the atmospheric column by vertical transportation of heat, moisture and momentum. There are two main categories in convective parameterization scheme. One is an adjustment scheme and the other is a mass flux scheme. HWRF model uses Simplified Arakawa Shubert (SAS) which is one of the mass flux scheme.

Grid point value

22

SAS deep convection scheme

SL

DL

LFC

CTOP

h hs

Environmental moist static energy

120-180mb

A

hs

hc

0.1A

Updated SAS scheme

Courtesy from Jongil Han (EMC)

23

Spurious? No momentum mixing

momentum mixing

analysis

Too intense ?

Han and Pan 2006

Mean sea level pressure (hPa)

132-

h fo

reca

sts

with

var

ious

am

ount

s of

mom

entu

m m

ixin

g27 Sep 2000, 12 UTC

24

Microphysics scheme:

While convective parameterization scheme is parameterizing subgrid/unresolvable moist processes, microphysics scheme predict the behavior of hydrometeo species explicitly. Hence, microphysics scheme are called explicit moisture scheme, grid scale precipitation scheme or large scale precipitation scheme .

There are bulk microphysics schemes (which are widely used in NWP models) and bin microphysics scheme. HWRF uses Ferrier microphysics scheme which is a single moment bulk scheme.

25

Cloud Microphysics Tropical Ferrier scheme

• mp_physics=85• Very similar to current NAM general Ferrier scheme Differences in RH condensation onset, number concentration, etc• Designed for efficiency• Advection only of total condensate (CWM) and vapor• Diagnostic cloud water, rain, & ice (cloud ice, snow/graupel) from storage arrays (F_*)• Assumes fractions of water & ice within the column are fixed

during advection• Supercooled liquid water & ice melt• Variable density for precipitation ice• (snow/graupel/sleet) – “rime factor” (F_rime)

26

Qt=Qi+Qr+Qc

(CWM = ice mixing ratio+rain mixing ratio + cloud water mixing ratio)

Qi = Fice * Qt

Ql = (1-Fice) * Qt Qr = Ql * Frain = (1-Fice) * Frain * Qt

Qc = (1-Frain)* Ql =(1-Fice) * (1-Frain) * Qt

F_rime =

F_rime should be always bigger than 1. Based on F_rime value, ice species are defined like, snow/sleet/grauple

27

RACW

CloudWater

GROUND

RE

VP

Rain

WaterVapor

RAUT

Sfc Rain

CND

ICN

D

DEP

Sfc Snow/Graupel/Sleet

Cloud Ice

PrecipIce

(Snow/Graupel/

Sleet)

IACWR

IEVP

IACW

IACR

IMLT

T < 0oC

T > 0oC

Flowchart of Ferrier Microphysics

From Ferrier, 2005

28

SW

Clear sky: net ~ -2o/day

SW

SW

SW

SWSW

LW

LW

LW

LW

absorption

reflection

sensible

latent

emissivity

albedo

Land ..low heat capacity, rapid temperature Changes… diurnal variability

Sea … high heat capacity, slow changesexcept for TC wake effects

~ -10+o/day

TC’s & Radiation Effects

Low clouds

29

GFDL radiation

Long wave• ra_lw_physics=98• Used in Eta/NMM• Default code is used with Ferrier

microphysics• Spectral scheme from global

model• Also uses tables• Interacts with clouds (cloud

fraction)• Ozone profile based on season,

latitude• CO2 fixed

Short wave• ra_sw_physics=98• Used in Eta/NMM model• Default code is used with Ferrier• Microphysics (see GFDL

longwave)• Interacts with clouds (and cloud

fraction)• Ozone/CO2 profile as in GFDL

longwave

30

Potential physics upgrades

31

Noah land model

32

Upgraded Land Surface model (GFDL slab to NOAH)

1. GFDL slab has shown large negative temperature bias over SW CONUS2. NOAH LSM has more down-stream application potential (e.g. storm surge, inland

flooding) on top of reducing negative temperature bias3. Track errors of land-falling storms seem to be improved according to preliminary

tests

~18% improvement with NOAH LSM

GFS anl HWRF fcst

HWRF - GFS

Cold bias of HWRF sfc T

33

Upgraded Ferrier Microphysics

1. New ice nucleation scheme to reduce no. concentration of small ice crystals

2. New, simpler closure for diagnosing small ice crystals and large, precipitating ice particles from ice mixing ratios

3. Advection of mass-weighted rime factor (i.e. “graupel”)

4. Slightly slower fall speeds of rimed ice

5. Increase the maximum (minimum) number concentration of small (large) ice in order to simulate better anvil cloud

operation

upgradedobs

Before upgrades

34

Upgraded SW/LW radiation schemes(GFDL radiation to RRTMG)

1. GFDL radiation schemes have problems of proper representations of cloud-radiation interactions, especially net cloud top cooling and net cloud base warming.

2. Although the use of RRTMG radiations degraded the intensity forecast skills of HWRF model, we are going to test again with tuning of some key parameters.

Cloud top cooling due to radiation

35

MESO SAS convection scheme

convective updraft area

fundamental assumption of SAS

The convective updraft area(Ac) is much smaller than grid box(Ae) σ = Ac/Ae << 1.0 : updraft fraction

When grid resolution becomes finer, the assumption will not be valid anymore (<~10km). The explicit MP scheme may also have a problem 10km or finer resolution to create moist adiabatic profile smoothly, which lead to grid-point storms.Meso- SAS scheme is designed to resolve this issue of the original SAS scheme by removing the assumption of σ << 1.0 (Hualu Pan)

36

IMPORTANT:We need closure assumption for the MESO SAS, which is the specification of the convective updraft fraction σ.

The current MESO SAS scheme determines σ based on the ratio of grid point vertical velocity and convective updraft vertical velocity as followed:

σ = 0.91 + 0.09

(If , then σ=1 and convection is off.