Numerical Simulation and Prediction of Supercell Tornadoes

Ming XueSchool of Meteorology and

Center for Analysis and Prediction of StormsUniversity of Oklahoma

mxue@ou.edu

February 2008

Outline of Talk

LES-resolution simulation of supercell tornado

Prediction of real tornados with radar data assimilation

Sensitivity of tornado prediction to microphysics

Tornadogenesis and Tornado Dynamics as Revealed by LES-resolution Numerical

Simulations of Supercell Storm

Numerical Simulation of Supercell Tornado using up to 12.5 m Grid Spacing

Using the Advanced Regional Prediction System (ARPS, Xue et al 2000, 2001, 2003) of CAPS

1977 Del City, OK sounding (~3300 J/kg CAPE) 2000 x 2000 x 83 point uniform resolution covering 50 x 50 km2. x = 25 m, zmin = 20 m, dt = 0.125 s.

x = 12.5 m in a 20 x 20 km subdomain, dt = 0.05 s.

Warmrain microphysics with surface friction at the later stage Simulations up to 5 hours Using 2048 Alpha Processors at Pittsburgh Supercomputing Center 60TB of data generated by one 25m simulation over 30 minutes, output

at 1 second intervals

Sounding for May 20, 1977 Del City, Oklahoma tornadic supercell storm

CAPE=3300CAPE=3300J/kgJ/kg

Full Domain Surface Fields of 50m simulation

t =3 h 44 mint =3 h 44 min

Red – positive Red – positive vertical vorticityvertical vorticity

Near surface vorticity, wind, reflectivity, and temperature perturbation from 25-m run

2 x 2 km2 x 2 km

Vort ~ 2 sVort ~ 2 s-1-1

Movie

Near surface vorticity, wind, reflectivity, and temperature perturbation from 12.5 m grid

1.5 x 1 km1.5 x 1 kmdomaindomain

Vort > 4 sVort > 4 s-1-1

Movie

Near surface vorticity, wind and p' felds- evolution from single to multiple vortices

t=13447 s t=13661s

Vort_max=3.27 /s Vort_max=3.28 /s

Movie of Cloud Water Field25 m, 7.5x7.5km domain, 30 minutes

Movie of Cloud Water Field. dx=25m 7.5x7.5km domain, 30 min.

130m/s

-100mb

>120m/s max >120m/s max surface windssurface winds

>90mb p drop>90mb p drop +60m/s speed +60m/s speed increase in ~2minincrease in ~2min

220min 236min220min 236min

Max sfc wind speedMin. sfc perturb. p220min 236min220min 236min

Maximum surface wind speed and pressure drop in 12.5 m simulation

What is the main source of air parcel and vorticity feeding the tornado?

Trajectory calculations based on 1-s model output

View from SouthView from Southt=13250st=13250sbeginning of beginning of vortex intensificationvortex intensification

z = 3 kmz = 3 km

View from NortheastView from Northeast

3km3km

RFD of1st cell

RFD of2nd cell

Inflowfrom east

Low-level jump flowLow-level jump flow

East West

Diagnostics along Trajectories

Orange portion t=13250-500s – 13250+200s

t=13250sBeginning of low-level spinup

14km14km

X Y Z

8km8km

WVh

Streamwise Vort.Cross-stream Vort.Horizontal Vort.

Vertical Vort.Vertical Vort.Total VortTotal Vort..

13250132501275012750 1345013450

Vorticity components along trajectory

Force along trajectoryForce along trajectory

BuoyancyBuoyancyVert. PgradVert. PgradSum of the twoSum of the two

Perturbation pressurePerturbation pressure-76mb-76mb

55

-5-5

1325013250

~2 m s~2 m s-2-2

+b' due to -p'+b' due to -p'

Forces along trajectory

Can we numerically predict real tornadoes?

May 8th, 2003 OKC tornado

OKC tornado2210-2238 UTC

30 km long path

F4

(Hu 2005; Hu and Xue 2007)

DA cycles on 1-km Grid3DVAR+Cloud Analysis Forecast

2030 UTC 2140 UTC

4 nested grids

Observed v.s. Predicted Z and Vr at 1.45° of the supercell storm

Observation 1 km Forecast

From 2140 to 2240 UTC every 5-min

Reflectivity

Radial velocity

What about the prediction of embedded tornado?

50-m Grid Forecast v.s. Observation

Forecast Low-level Reflectivity Observed Low-level Reflectivity

Movie

43 minute forecast

50-m Grid Forecast v.s. Observation

Forecast Low-level Reflectivity Observed Low-level Reflectivity

Movie

43 minute forecast

43 min. forecast on 100m grid

t=34 min

t=40 min

Sfc vert. vort., and p’ E-W x-sections of vert. vort. and w

A case from CASA 2007 Spring Experiment

CASA – an NSF ERC for Collaborative Adaptive Sensing of Atmosphere

- Low cost, high density, adaptively scanning radars

© KSWO TV

© Patrick Marsh

May 8-9, 2007

A series of low-levelcirculations.

NWS TornadoWarnings: 7:16pm,7:39pm, 8:29pm

7:21pm (0021Z)

8:30pm (0130Z)

9:54pm

10:54pm (0354 Z) Minco Tornado

A Case from 2007 CASA Spring Experiment

dx = 400 m 115-min. prediction of sfc winds, Z (color), and vertical vorticity at 0355 UTC. Both WSR-88D and CASA IP1 data were assimilated very 5 min. for 1 h. The black triangle indicates the location of observed Minco tornado.

0:00Z 0:30Z 1:00Z 1:30Z 2:00Z

0Z Analysis1 hr. spin-up period 1 hr. assimilation window

with 5 min assimilation intervals

Forecast to 0500 UTC

0:00Z 0:30Z 1:00Z 1:30Z 2:00Z

0Z Analysis1 hr. spin-up period 1 hr. assimilation window

with 5 min assimilation intervals

Forecast to 0500 UTC

Predicted sfcVort. max

115-min sfc forecastMinco tornado

Importance and/or Uncertainties of Microphysics?

Daniel Dawson’s Poster Yesterdayusing multi-moment microphysics

Impact of Microphysics on Prediction of Tornadic Supercell Storm

May 3, 1999 Moore – OKC F-5 Tornado Case

Daniel Dawson’s Poster Yesterdayusing multi-moment microphysics

Surface ’gray shading), Z (blue contours), vertical vorticity (color shading), and wind vectors at the time of largest vertical vorticity using 100 m resolution and with MY1 (a) and MY2 (b) schemes.

HP storm LP storm

100 m simulations with MY1 and MY2 schemes

Vis5D visualization of the cloud field (gray surface), and 0.3 s-1 vertical vorticity iso-surface (yellow) from the 100

m simulations using MY1 (left) and MY2 (right) schemes.

MY Single-moment MY two-moment

Greensburg, Kansas Tornado, 5 May 2007

Numerical prediction of tornados - has its time come?

What is the predictability of tornadoes?