Click to add Text © Crown copyright Met Office Statistical Analysis of UK Convection and its...

Click to add Text

© Crown copyright Met Office

Statistical Analysis of UK Convection and its representation in high resolution NWP ModelsHumphrey Lean, Kirsty Hanley, Carol Halliwell

MetOffice@Reading, Reading, UK

Robin Hogan, Thorwald Stein, Bob Plant, Peter Clark

University of Reading, Reading, UK


Introduction

• “Convection permitting models” (i.e. those with explicit convection) have provided a step change in our ability to forecast on smaller scales.

• Here describe work to improve our understanding of representation of convection in these models.

• Not a trivial question. Just running at higher resolution is not necessarily a panacea. Although there is no convection scheme we have to think about the representation of other subgrid processes.


Model resolution

• Current highest resolution operational UK configuration 1.5km (UKV model)

•For research we are running higher resolution versions of the Unified Model down to 100m (55m for London)

• Helps understanding of convection permitting versions of model to look as function of gridlength.

• Also looking to future. Higher resolution very expensive but also interest in high res models of small areas (e.g. Weymouth 333m for Olympics, London 333m).


Heathrow downscaler from UKV

300 x 200

333m resolution

70L

10 second timestep

Run twice a day at 03Z and 09Z

Run to T+36

Clive Wilson, Anke Finnenkoetter


“UKV” 1.5km UK Model convection1.5km model Radar

15 UTC 12th April 2012

•Convective cells too large and too intense. •Not enough light rain.


Average cell diameter (km)(averaged over 22 convective cases)

Threshold (mm/hr)

Radar 1.5km UKV

0.125 7.81 16.04

0.25 6.32 13.21

0.5 5.58 11.71

1.0 4.42 9.93

2.0 3.28 7.95

4.0 2.57 5.96

16.0 2.13 3.37

6

UKV Under-resolvedCurrently investigating shallow scheme rather thantrying to resolve smaller showers.


Gridscale structure in 750 hPa w13UTC 12/05/2010 – case of light scattered showers

4km 1.5km 500mEmilie Carter


100m UM versions

• Cold pooling in valleys COLPEX (Clark, Vosper, Carter)

• London (Lean)

• Convection (Carter, Halliwell, Hanley)

•Tornadoes (Hanley)

• StCu (Boutle)

• Fog inc nesting in ensemble (Porson)

• (Also 333m Weymouth model for Olympic sailing and 333m London model (fog)).


Comparison with observations

Temp RH

Wind speed Wind dir

Emilie Carter

COPE IOP 2 July 5 2013 sea breeze convergence

Different scale!

Vertical velocity at 500m amsl

Below 500m starting to resolve BL turbulence

K Hanley


DYMECS project

• DYnamical and Microphysical Evolution of Convective Storms (DYMECS).

• Statistical analysis of convective cells with Chilbolton Radar over a number of cases.

• Compare to UM configurations between 1.5km and 100m

• R Hogan, Plant, Stein, Nicol, Clark University of Reading

• Lean, Halliwell, Hanley MetOffice@Reading

Also in summer 2013 had field phase of COPE (COnvective Precipitation Experiment) which will be used for similar analysis and model comparisons.


The DYMECS approach: beyond case studies

Met Office 1km rainfall composite

Track storms in real time

and automatically

scan Chilbolton

radar

Derive properties of hundreds of storms on ~40 days:Vertical velocity3D structureRain & hailIce water contentTKE & dissipation rate

Evaluate these properties in model varying:Resolution (down to 100m)Microphysics schemeSub-grid turbulence parametrization

25m diameter S-band (3 GHz)Steerable (2 degrees per second)

0 dBZ out to 150 km


Model Differences


2.2km 1.5km

200m 100m 1km RADAR

500m4km

© Crown copyright Met Office© Crown copyright Met Office

Convective cell life cycle

Thorwald Stein


20 April 2012 25 Aug 2012

100m model best

500m model best

1.5km model best

11 UTC - radar 15 UTC - radar

Surface rainrate statistics


Cloud widths for different reflectivity thresholds.

Radar Distance (km) 1.5km 500m

200m 100m

Thorwald Stein

Median width of deep storms25th Aug 2012


25 Aug 2012

Cutaway: reflectivity Surface: rainrateShading: extent of cloud

Robin Hogan

3D visualisation of data


Reflectivity

Actual model vertical velocity

Estimated vertical velocity

Observations UKV 1500m Updraft Updraft retrievalretrieval

10 km height

20 km width

40 dBZ

+10 m/s

-10 m/sEstimate vertical velocity from vertical profiles of radial velocity, assuming zero divergence across plane.

Quantify errors due to 2D flow assumption

John Nicol


Vertical distribution of vertical velocity and reflectivity

Observations

1500m

500m

200m

100m

John Nicol



John Nicol

Vertical velocity and reflectivity distributions as function of radius from centre of updrafts. Set to zero where: w<1m/s, Z<20dBZ.

1200-1600UTC 25/08/12. Black traces - mean widths for 1m/s and 20dBZ.

N.B. higher values of Z than before - rain



John Nicol

Vertical velocity and reflectivity distributions as function of radius from centre of updrafts. Set to zero where: w<1m/s, Z<20dBZ AND where values start increasing again.“primary profile”

1200-1600UTC 25/08/12. Black traces - mean widths for 1m/s and 20dBZ.


Vertical velocity and reflectivity as a function of gridlength.

John Nicol

Observations

1500m

500m

200m

100m



John Nicol

Observations

1500m

500m

200m

100m

Updrafts fit best at about 200m – definitely too narrow at 100m



John Nicol

Observations

1500m

500m

200m

100m

Primary (monotonic) cloud widths proxy for updraft width



John Nicol

Observations

1500m

500m

200m

100m

Non monotonic profiles of Z too narrow in 100m and 200m (tallies with cloud and rain data).



John Nicol

Observations

1500m

500m

200m

100m

Obs show bigger increase from dropping the monotonic condition than any model. Fits with cloud not correctly filling in between updrafts.


25 Aug 2012

Cutaway: reflectivity Surface: rainrateShading: extent of cloud

Robin Hogan

3D visualisation of data


Mixing length dependence

John Nicol

radar

200m l=300m

200m l=100m

200m l=40m

Smaller mixing length lower vertical velocity (but more cells)


• Can we make the 200m model produce larger storms on 25th Aug 2012 by changing the mixing length?

• Need to reduce the timestep to 3s to avoid hitting stability limit.

• Can produce larger storms but at the expense of the smaller storms.

• Increasing the mixing with height may be a solution i.e. make Smagorinsky scheme dependent on Δz.

Varying the mixing length

l=300m – 2491 stormsl=100m – 3583 stormsl=40m – 4359 storms

200m

Kirsty Hanley


Variation with microphysics

John Nicol

Observations

aggregates

graupel

default

Melting level at 2.6km.

Hypothesise differencedue to fall speed.


Conclusions• UKV clearly under-resolves many small showers in UK

• High res models (~100m) improve some aspects but also some problems.

• Models below 500m tend to produce too narrow showers (measured by surface rain or cloud) in cases where showers are large (for small showers 200m or 100m fits well). Cloud widths roughly the same in 200m and 100m.

• Updraft widths good in 200m model but too narrow at 100m.

• Above two together imply there may be an issue about how the model fills in cloud between updraft cores.

• Representation very sensitive to mixing. Also sensitivity to microphysics (fall speed).


Future work

• What can be done to understand lack of convergence and too narrow updrafts/clouds in 100m/200m models?

• Suspect problem is turbulence “grey zone”. Would better resolution of turbulence solve these problems at higher resolution? Try higher vertical/horizontal resolution and see if updrafts/clouds get wider (or stop collapsing). Work with LES community.

• Can we improve models with more appropriate subgrid mixing schemes?

• Effect of microphysics?


DYMECS References

Hanley, K. E., Plant, R. S., Stein, T. H. M., Hogan, R. J., Nicol, J. C., Lean, H. W., Halliwell, C. and Clark, P. A. (2014), Mixing-length controls on high-resolution simulations of convective storms. Q.J.R. Meteorol. Soc.. doi: 10.1002/qj.2356

Stein, Thorwald H. M., Hogan, Robin J., Hanley, Kirsty E., Nicol, John C., Lean, Humphrey W., Plant, Robert S., Clark, Peter A., Halliwell, Carol E. (2014), The three-dimensional morphology of simulated and observed convective storms over southern England. Mon. Wea. Rev.. doi: 10.1175/MWR-D-13-00372.1

Stein, Thorwald H. M., Hogan, Robin J., Clark, Peter A., Halliwell, Carol E., Hanley, Kirsty E., Lean, Humphrey W., Nicol, John C., Plant, Robert S., (2014), The DYMECS project: A statistical approach for the evaluation of convective storms in high-resolution NWP models. (submitted to Bull. Amer. Meteorol. Soc.).

Click to add Text


Thank you for listening.Any questions?

Click to add Text © Crown copyright Met Office Statistical Analysis of UK Convection and its...

Documents

Transcript of Click to add Text © Crown copyright Met Office Statistical Analysis of UK Convection and its...