DYMECS
The three-dimensional structure of convective stormsRobin
HoganJohn NicolRobert PlantPeter Clark
Kirsty HanleyCarol HalliwellHumphrey LeanThorwald Stein
([email protected])www.met.reading.ac.uk/~dymecs
(UK Met Office)
The three-dimensional structure of convective stormsNWP models
run at km-scale: errors in timing, location, structure of
convective precipitation.
Storm analysis of 2D fields (surface rainfall rate, OLR)
highlights errors, but not underlying processes.
Use high-resolution (300m) radar observations for many storms to
evaluate model storm morphology and dynamics.
The three-dimensional structure of convective stormsUKV
1500m200mAnimations by Robin Hogan
The DYMECS approach: beyond case studies
Met Office 1km rainfall compositeTrack storms in real time and
automatically scan Chilbolton radarDerive properties of hundreds of
storms on ~40 days:Vertical velocity3D structureRain & hailIce
water contentTKE & dissipation rateEvaluate these properties in
model varying:ResolutionMicrophysics schemeSub-grid turbulence
parametrization
25m diameter S-band (3 GHz)Steerable (2 degrees per second)0 dBZ
out to 150 km
Storm structure from radarDistance east (km)Distance north
(km)Radar reflectivity (dBZ)40 dBZ0 dBZ20 dBZ
ShallowDeepObservationsUKV 1500m200m Median storm diameter with
height
500m
Convergence?
Lack of anvils?
Drizzle from nowhere?
Vertical profiles ofreflectivity
1.5-km 1.5-km + graupel 1.5-km no
crystalsObservationsConditioned on average reflectivity at
200-1000m below 0oC.
Reflectivity distributions forprofiles with thismean Z 40-45 dBZ
are shown.Model:High rainfall rate from storms lacking ice or have
ice cloud dBZ0 within 90 km of the radarEstimated vertical
velocityTruevertical velocity(slide courtesy John Nicol)
Vertical velocity distribution between 7-8 kmTrue model
velocityEstimated model velocityRadar estimated velocityRadar
mapped true velocity
map 500m simulation compares well with radar using 2D flow
assumption (dashed lines)(slide courtesy John Nicol)Evaluation of
width of updraftsModel updrafts shrink with resolution200-m model
has about the right widthDoes 100-m model shrink further or stay
the same?How does Smagorinsky mixing length affect model?
Observations200-m model500-m model1.5-km modelRetrieval in both
observations and model:wmin=0.5 m/s; wmax>3.0m/s True model
versus mapped observations: wmin=1.0 m/s; wmax>5.0m/s The
three-dimensional structure of convective stormsThorwald Stein
([email protected])www.met.reading.ac.uk/~dymecs
Models with smaller grid length producenarrower storms, similar
to observations.
Models associate shallow ice cloudwith high rainfall too
frequently.
500m grid length model has verticalvelocity distribution
comparable toobservations.
Future work: Study the E of DYMECS.Mixing-length sensitivityin
200m storm structures
100m mixing length300m mixing length40m mixing length500m
model1500m model200m simulation can approximate storm structures in
coarser grid-length simulations by varying mixing length
