Dynamical and Microphysical Evolution of Convective Storms (DYMECS) University: Robin Hogan, Bob Plant, Thorwald Stein, Kirsty Hanley, John Nicol Met Office:

Dynamical and Microphysical Evolution of Convective Storms

(DYMECS)University: Robin Hogan, Bob Plant, Thorwald Stein, Kirsty Hanley, John NicolMet Office: Humphrey Lean, Emilie Carter, Carol Halliwell, Andy Macallan

• Forecast models now “resolve” convection… or do they?– Significant problems remain with intensity and scale of storms– Urgent need to evaluate different model configurations – Case studies useful but might not be representative generally

• In DYMECS we aim to observe hundreds of storms through their lifecycle to evaluate the model statistically– Develop adaptive scanning algorithm for Chilbolton to track storms as they

evolve– Run unmanned during 40 days over an 18-month period including two

summers– Derive macrophysical, microphysical and dynamical properties vs time in

lifecycle– Compare variables with assumptions made in some convective

parameterizations– Track the same properties in storms in the operational 1.5-km Met Office

model– Rerun model, testing microphysics, sub-grid mixing and resolution

Work Packages (WPs) Yr 1 Yr 2 Yr 3

PDRA 1: Dr Thorwald Stein 1. Design and test storm tracking algorithm 2. Apply storm tracking algorithm over 18 months 3. Derive properties from radar scans 4. Statistical analysis of observed storm cells PDRA 2: Dr Kirsty Hanley 5. Cell tracking in operational forecast model 6. Modelling case studies and sensitivity tests 7. Long-term comparison of model and observations 8. Synthesis and hypothesis testing

Work plan

Start of project: April 2011

We are here: Jan 2012



1. Design & test storm tracking algorithm

1. Nimrod rain rates received at Chilbolton in near real time and 400x400 km region around Chilbolton extracted (John)

2. Contiguous regions above a certain threshold identified, labelled and tracked from one 5-min to next, allowing for merger and break-up (Thorwald)

3. Storms are prioritized automatically based on rain rate, area, distance to radar etc (although can “log in” and manually set priority), then scans commands are issued to the radar (Robin & Thorwald)

Done.



2. Apply storm tracking over 18 months

1. Applied to 8 cases so far (4 in August, 2 in Nov and 2 in Dec)

2. Funding for total of 40 days, focussing on summer 20123. Note that winter convection also of interest since snow

showers often poorly forecast; microphysics is suspect

Ongoing.



3. Derive properties from radar scans

• Cloud area, cloud-top height versus time into cell lifecycle• Surface rain rate, drop size, hail intensity from polarization variables (Hogan

2007)• Ice water content using radar reflectivity and temperature (Hogan et al. 2006)• TKE and dissipation rate from Doppler spectral width (Chapman and Browning

2001)

• Updrafts…

Ongoing.

Updrafts?• Hogan et al. (2008)

– Track features in radial velocity from scan to scan

• Chapman & Browning (1998)– In quasi-2D features (e.g.

squall lines) can assume continuity to estimate vertical velocity



4. Statistical analysis of observed storms

Alan Grant (2007) suggested the following “testable relationships” in convection parameterization:

where

• up is the mean in-cloud dissipation rate

• wup is the cumulus vertical velocity scale

• Lup is the horizontal length scale of the updrafts

• Aup is the fractional area of some horizontal domain occupied by cumulus updrafts (equal to the cloud-base mass flux in a convection scheme divided by wup)

• Dcld is the depth of the convective cloud layer

• CAPE is the convective available potential energy

About to start.

upupup Lw /3 cldupup DAL 2/1 CAPE/2upup wA



5. Cell tracking in operational model

• Humphrey and Andy have provided operational model data– 5 minutely for surface rain rate, hourly for 3D fields

• Thorwald has tracked storms successfully from rain rate• Kirsty will analyse 3D fields to derive variables for

comparison with radar over storm lifecycle

Partially done.



6. Modelling case studies & sensitivity tests

• mOnSoOn application successful: Emilie & Kirsty can share jobs

• Horizontal resolution (Emilie already started)– Down to 100 m; model currently predicts smaller cells as resolution

increases

• Sub-grid mixing scheme (first task for Kirsty)– Test 2D & 3D Smagorinsky, prognostic TKE and a stochastic backscatter

scheme– Evaluate rate of change of cloud size with time, and TKE

• Microphysical scheme (can do shortly?)– Test single- and double-moment liquid, rain, ice, snow, graupel and

possibly hail, as well as interactive aerosol-cloud microphysics

Just starting.



7. Long-term comparison

• Choose a suitable resolution, microphysics and sub-grid setting and run model over all observed cases and perform long-term evaluation against radar

Not yet started.



8. Synthesis & hypothesis testing

• Tackle key scientific questions for convective-scale forecasting– Can models distinguish single cells, multi-cell storms & squall lines?– What about location of daughter cells formed by gust fronts?– Does BL scheme “diffuse away” gust fronts necessary to capture

triggering of daughter cells and if so how can this be corrected?– What is cause of quasi-stationary storms (often cause of flooding)?– Can we diagnose parameters that should be used in convection

scheme from observations? – Can we estimate entrainment profile in convection from

observation-model combination (key uncertainty for climate)?

Not yet started.

Forecast 3D storm structure

3D structure observed by Chilbolton

Early 3D reconstructions

Met Office 1.5 km model

National radar network rainfall

16.00 on 26 August 2011

Ra

in r

ate

(m

m h

-1)

Radar observations

Forecast plan-view of rainfall

Dynamical and Microphysical Evolution of Convective Storms (DYMECS) University: Robin Hogan, Bob Plant, Thorwald Stein, Kirsty Hanley, John Nicol Met Office:

Documents

resolution slide

rain rate kirsty

convection scheme

vertical velocity slide

d structure

d features

hundreds of storms

met office model rerun