580 likes | 775 Views
High Resolution UM and Orography: Research at the Met Office. Richard Forbes October 2004 Rachel Capon, Peter Clark, Humphrey Lean, Clive Pierce, Nigel Roberts, Samantha Smith, Simon Vosper. Contents. The Unified Model at high resolution Orography and the hi-res UM
E N D
High Resolution UM and Orography: Research at the Met Office Richard Forbes October 2004 Rachel Capon, Peter Clark, Humphrey Lean, Clive Pierce, Nigel Roberts, Samantha Smith, Simon Vosper
Contents • The Unified Model at high resolution • Orography and the hi-res UM • Orographic rainfall and smoothed orography • Rain advection and orography • Lee waves • MAP Case Study • Channel flow • Boscastle Flash Flood • Summary
Unified Model • UM Configurations: • Operational 12km 38 level model • Research mode 4km 38 level and 1km 76 level models (High Resolution Trial Model, HRTM) • Relevant Model Parametrizations: • Mixed-phase cloud and precipitation scheme (Wilson and Ballard 1999, Smith 1990) + prognostic ice/snow/rain/graupel option for high resolution • Mass flux convection scheme (Gregory and Rowntree 1995) + CAPE dependent CAPE closure timescale for 4km • Diffusion (targetted to high w regions) + Smagorinsky-Lilly type turbulence parametrization (available soon)
High Resolution Trial Model Configuration 1 km 76 levels Resolved convection Prognostic rain Initial 12km T+1 4 km 38 levels Mass-limited convection Prognostic rain Initial 12km T+1 4 km 1 km
Orographic Rainfall Humphrey Lean, Peter Clark
Seeder-feeder model • Seeder-feeder model taking into account condensation, advection and accretion of water cloud to form precipitation. dqcl/ds = αC0 –(qcl/U)A0P0.68 dP/dz = –ρqclA0P0.68 • Equilibrium solution
Surface Precipitation Along the Slope Predicted from Simplified Model h = 1.5km Ph=0.5mm/hr = 0.02 Red = without rain drift, black=with
Does the seeder-feeder reach equilibrium ? • For a long slope (e.g. >50km), the solution reaches equilibrium and the total precipitation enhancement hill height • For a short slope (e.g. <50km), the solution has not reached equilibrium and the total precipitation enhancement hill volume • This has implications for representation of orography in models and the impact of smooth (filtered) orography on the precipitation from the seeder-feeder process.
Effect of Smoothing Orography • Effect of smoothing orography depends on scale compared to L (~50km) • Recall for s<<L enhancement proportional to volume of hill - hence smoothing on these scales just redistributes rain (applicable to grid resolutions of ~10km and less). • For s>>L enhancement proportional to height of hill - in this case smoothing will reduce total amount of rain (applicable to grid resolutions >10km).
Effect of smoothing orography 12km Orography Smoothed 12km Orography
Case Study: 29 Nov 2001Frontal rainband with significant orographic enhancement Met Office analysis chart for 00z on 29/11/2001
Effect of smoothing orography on total precipitation:11-12Z 29th Nov 2001 Accumulated Rainfall from 12km Unified Model with unsmoothed orography (top) and smoothed (bottom). Rain over South Wales mountains changes by only ~1%.
Seeder-Feeder Model Conclusions • A simplified seeder-feeder model has been used to investigate magnitudes and spacial dependencies in idealised cases. • This above model has been used to investigate scale dependence for long and short hill limits. • Smoothing the orography has no effect on the total rainfall in the 12km model but a large effect in the 60km one as would be expected from the limit arguments. • The drift of rainfall is expected to redistribute rainfall on scales of order 10km
Advection of Rain Richard Forbes
Advection of Rain in NWP Models • For high resolution NWP models (< 5 km) the advection of rain by the wind becomes more significant (smaller grid boxes, shorter timesteps). • Is it important to include this in NWP models ? • Location of seeder-feeder rainfall over orography can be crucial for river catchment rainfall accumulations (catchment boundaries are associated with orography). • Compare the Unified Model at 2km with and without rain advection (prognostic vs. diagnostic rain).
Impact of including rain advection on rainfall distribution. 10hr model forecast. Rainfall rate difference (advection-no advection) Rainfall rate (mm/hr) Orography (m)
Vertical cross section of snowfall/rainfall rate across Dartmoor Vertical cross sections across Dartmoor Snowfall rate No Rain Advection Rainfall rate Snowfall rate With Rain Advection Rainfall rate
VerificationDartmoor River Catchment Rainfall 9 Hour Accumulation Model Forecast (Diagnostic Rain) NIMROD (Radar Network) Model Forecast (Prognostic Rain) Exe Teign Dart Tamar Avon & Erme
Correlation between model and NIMROD radar-derived accumulated rainfall for Dartmoor river catchments No Rain Advection With Rain Advection
Rain Advection Summary • Rainfall forecasts for fast-response river catchments are important for flood prediction. • A difference in the location of orographically enhanced rainfall of only a few km can result in a large difference in the rainfall prediction for river catchments. • Including the advection of rain in a 2km version of the Unified Model significantly improved the spatial distribution of surface rainfall over orography and associated river catchments.
Lee Waves Simon Vosper
Pennine Lee Wave Observation Campaign • Region east of the Pennines in the Vale of York, is known to have significant forecasting problems associated with orographic flows. • Obs. field campaign for 1 year (2004). • Study lee waves, rotors and downslope windstorms. • Compare models (3DVOM, BLASIUS and UM) with observations.
Pennine Lee Waves: Model Comparison Plan view of vertical velocity on 2km model level 3DVOM (linear model) BLASIUS UM 1km horizontal grid resolution
Pennine Lee Waves: Radiosonde comparison Radiosonde (black) 3DVOM (red) BLASIUS (green) UM (blue) • UM has best phase but low amplitude • 3DVOM/BLASIUS have good amplitude but poor phase
MAP Case Study Samantha Smith
MAP Case Study: 8 Nov • How does the UM represent flow over high complex mountains, especially for low Froude number problems ? • MAP case 8 Nov 1999 • Northerly flow with upstream flow splitting
MAP Case Study: 8 Nov • UM 4km reproduces large scale flow, flow splitting, cold pool in Po valley, lee-side Foehn and Froude number well. • Comparison with Payerne sonde.
MAP Case Study: 8 Nov • Comparison of model vertical velocity with aircraft at different heights. • Wavelengths are predicted but amplitude underpredicted.
Total surface drag as a function of model resolution • Magnitude of surface drag increases with increasing resolution • No sign of convergence down to 4km
Channel Flow Rachel Capon
The Levanter Wind: Straits of Gibraltar • Low level weak Easterly flow capped by strong inversion. • Substantial acceleration through Straits of Gibraltar. • Regular feature – quite predictable given large scale setup.
Surface Observations Around Gibraltar 25 knots 25 knots
‘Levanter’ Wind through Straits of Gibraltar – Impact of Model Resolution 12 km L38 (part) 4 km L38 (part) 1 km L38 18 UTC 26/03/2002 from Global analysis 12 UTC 25/03/2002 30 Hour Forecast
1 km Forecast 26/03/2002 • Peak 10 m wind speed ~ 20 m/s • Very steady • Note ‘side bands’ – are they realistic?
Side Bands Side Bands Second Levanter Case 21st May 20031 km forecast 08 UTC (T+14) 14 UTC (T+20) How do we verify?
Sun Glint Dark=low reflection =more waves =higher wind speed Some serendipitous data! AVHRR Visible Image 13:45 pass on 21/05/2003 Greatly Contrast Enhanced
Sun Glint Anemometry AVHRR Visible Image 13:45 pass on 21/05/2003 14 UTC (T+20) Greatly Contrast Enhanced
Comparison of Dover and Gibraltar Straits • Straits of Gibraltar Jet occur with a strong capping inversion to the boundary layer. • Gibraltar appears to be genuine ‘gap flow’. • Note scale of channel flow - ~100 km, and some upwind impact reflecting stable flow dynamics. • “Side bands” represented in the model. Cause ?
Boscastle Flash Flood Peter Clark, Humphrey Lean, Clive Pierce, Nigel Roberts, Richard Forbes
Boscastle Flash Flood • Boscastle (South-West England), 16th August 2004 • Large amount of precip >60mm over a few hours • Small river catchment • ~£500,000,000 damage • Fortunately, no one killed • No warnings (even 2 hour warning would be useful)
Rainfall Accumulations 12-18 UTC 16th August 2004 Forecasts from 03 UTC 4 km 12 km NIMROD radar Peak Accumulations >60mm On 4 km grid Positional error and false alarm 20 km radius from Boscastle
Comparison of 00 UTC and 03 UTC 4 km forecasts 12-15 from 03 UTC 12-15 from 00 UTC 12-18 from 03 UTC T+15 run from 00 UC does not cover full period of actual rain 00 UTC run better than 03 UTC
Mechanism for 16th August 2004 Storm Triggering 10 m wind convergence 11 Z 10 Z Persistent Convergence Due to coast and orography Orography in white
Boscastle Summary • Explicit convection in 4 km gave better peak accumulations than parametrized in 12 km • 00 UTC run better than 03 UTC • High predictability but not high enough to focus attention on individual catchments • Triggering mechanism appears to be orography + land/sea roughness + surface heating • Forecast would have justified flash flood warnings for N Devon & N Cornwall 6-9 hours ahead