Download presentation
Presentation is loading. Please wait.
Published byAnis Eunice Carr Modified over 9 years ago
1
Page 1© Crown copyright 2005 Progress with high resolution modelling with the Unified Model Peter Clark Group Leader Mesoscale Modelling Met Office Joint Centre for Mesoscale Meteorology University of Reading
2
Page 2© Crown copyright 2005 Talk Outline 1.Met Office operational models. 2.High Resolution model configurations. 3.Rainfall verification and model products.
3
Page 3© Crown copyright 2005 Organisation JCMM (Reading) Mesoscale Modelling Peter Clark + 6 Mesoscale Data Assimilation Sue Ballard+3 General Parametrization Data Assimilation Satellite Applications Dynamics Research Evaluation and Diags UM Systems Met Office (Exeter)
4
Page 4© Crown copyright 2005 Met Office Regional NWP Strategy North Atlantic and European (NAE) limited area model Europe/Storm tracks, T+48…. 12 km 4D Var Data Assimilation main forecast 24 km ensemble system embedded in global ensemble New UK Model quasi-operational April 2005 T+36 4 km, UK weather especially surface impacts ‘Spin-up’ from NAE analysis through summer 2005. Full 3DVAR/MOPS assimilation cycle from end 2005 (tomorrow!). Vertical resolution enhancement 2006 Experimenting with 1 km since 2002. ‘On-demand’ small area 1.5 km model by 2007. Expect UK model to move to 1.5 km in 2009.
5
Page 5© Crown copyright 2005 Future UM Operational Configurations Global 40 km North Atlantic & European 12 km Old UK 12 km Retiring New UK 4 km Levels: 38 50 Deep Strat 70+
6
Page 6© Crown copyright 2005 Other UM high resolution areas S Africa and Ethiopia 4 km New Zealand and Alps (MAP) 60km, 40km, 20km, 12km, 4km, 2km and 1km – studies of stress convergence (Stuart Webster) TOTAL RESOLVED SSO OR 2km 60km
7
Page 7© Crown copyright 2005 Motivation for high resolution forecasting Severe convective storms can lead to flash flooding. strong winds associated with storms. Boscastle (SW England), 16th August 2004 ~£500,000,000 damage Fortunately, no one killed Even 2 hours warning useful Forecasting of convective precipitation is primary public safety need
8
Page 8© Crown copyright 2005 Emphasis for UK modelling Main emphasis on ~1 km very short range model. 4 km very useful for temperature/visibility/wind forecasting. Uncomfortable about precipitation. Nevertheless, more useful than expected. Much depends on intelligent upscaling in post- processing.
9
Page 9© Crown copyright 2005 Unified Model at 4 km and 1 km resolution Non-hydrostatic, compressible, deep atmosphere, semi-Lagrangian, semi-implicit dynamics. Arwakawa C horizontal rotated lat/long, Charney Philips vertical flexible terrain following height based. Philosophy has been to start with existing UM physics and enhance only where evidence shows need. Main physics developments are microphysics and turbulence Still taking conservative approach Additional developments: Enhanced urban scheme Surface slope in radiation
10
Page 10© Crown copyright 2005 New UK 4 km Model Broad Leaf Trees Needle Leaf Trees C3 Grass C4 Grass Shrubs Urban Lakes Bare Soil Land Ice Post-processing products area
11
Page 11© Crown copyright 2005 Microphysics Operational UM Wilson and Ballard microphysics– prognostic total ice+snow, diagnostic ice/snow split, diagnosic rain. Since UM6.0 there is a prognostic representation of: 3D SL advection initially Separate 3D advection by wind and 1D SL transport relative to air Useful for single column Cheaper and no significant impact on solution (we think!) Developed from standard UM Wilson and Ballard microphysics New microphysics fully flexible Working towards (optional) convergence with Met Office CRM. Working on improvements to numerics. Cloud liquid water Water vapour Graupel Ice crystals Snow aggregates Rain
12
Page 12© Crown copyright 2005 10 th July 2004 1 km UM 0700 Z
13
Page 13© Crown copyright 2005 10 th July 2004 1 km UM 0800 Z Convergence Line
14
Page 14© Crown copyright 2005 10 th July 2004 1 km UM 0900 Z
15
Page 15© Crown copyright 2005 10 th July 2004 1 km UM 1000 Z
16
Page 16© Crown copyright 2005 10 th July 2004 1 km UM 0900 Z Convergence Line
17
Page 17© Crown copyright 2005 Impact of reduced snow fallspeed and enhanced sublimation Standard Run Reduced ice fallspeed Double evaporation rate Convergence Line
18
Page 18© Crown copyright 2005 Quantifying Systematic and Local Impacts Graupel → very small systematic increase in rainfall, small local impact. Increasing rain evap. rate → small systematic decrease in rain, larger local impact. Water loading → small systematic decrease in rainfall, significant local impact Decreasing the snow fall speed → large systematic impact, significant local impact.
19
Page 19© Crown copyright 2005 Two alternative turbulence treatments Standard UM 1D boundary layer (Lock et al, 2000) Non local eddy diffusivity Moist Multiple regime Not using shallow convection (future work) Implicit solution Fixed Horizontal hyper- diffusion (del-4). Arbitrary chosen to give most reasonable power spectra. Explicit solution Smagorinsky-Lilly 3D with stability dependent length scale. Stability functions same as local part of standard UM scheme. Basic length scale proportional to horizontal grid length. Same numerical solution method as standard scheme.
20
Page 20© Crown copyright 2005 Impact of turbulence scheme on convective forecast (4 th July 2005) UM BL3D Smagorinsky. 1km UM 6 hour forecast surface rainfall rate.
21
Page 21© Crown copyright 2005 Impact of turbulence scheme on convective forecast (4 th July 2005) Number of cells Reference With Turbulence Histogram of cell sizes Average cell size Time →
22
Page 22© Crown copyright 2005 Summary – Turbulent Mixing 3D sub-grid turbulent mixing parametrization introduced into the UM (based on Smagorinsky-Lilly). Tested in idealised and real case studies and can have a very significant impact on convective initiation and evolution. Reduces over-prediction of small convective cells at 1km. Reduces excessive rain rates in larger storms. BUT not appropriate for all situations (e.g. very stable). Work is ongoing into most appropriate formulation for different resolutions, and enhancing the scheme (e.g. stochastic backscatter).
23
Page 23© Crown copyright 2005 Convection at 4 km We don’t know how to parametrize convection at 4 km. We don’t all agree what a correct solution would look like! We have decided not to try to develop a ‘4 km’ convection scheme. Gregory-Rowntree ‘hands over’ to explicit smoothly but not correctly Depends on parameter choice Some modes of behaviour not necessarily physical Behaviour different for boundary forced domain compared with periodic, homogeneous (CRM) Nested has additional sink of small scale energy Domain and problem dependent Pragmatism (fudges!) necessary CRM equilibrium behaviour used for guidance
24
30 Convection scheme closure CAPE (J/Kg)Mass flux Mass flux CAPE / CAPE closure timescale 3/9/02 Nigel Roberts, JCMM CAPE (J/Kg)
25
Page 25© Crown copyright 2005 Severe Organised Convection 3 rd August 2004 NIMROD Radar Rainrate 5/2/1 km Composite
26
Page 26© Crown copyright 2005 Severe Organised Convection 3 rd August 2004 Operational 00Z 03/08/2004 12 km Mesoscale Total Rainfall rate (Part of domain) Every timestep
27
Page 27© Crown copyright 2005 Severe Organised Convection 3 rd August 2004 As Operational 00Z 03/08/2004 4 km UK Total Rainfall rate (Part of domain) Every timestep
28
Page 28© Crown copyright 2005 Severe Organised Convection 3 rd August 2004 Radar 12 km4 km 1600 UTC T+16 Forecasts
29
Page 29© Crown copyright 2005 Why (and when) 4 km can be useful for precipitation Convergence lines that trigger new convection are quite well resolved by 4 km model. This can give good spatial indication of rain. Especially true with cold pool dynamics – (our) parametrized convection poor to useless. (Probably true of any quasi-equilibrium scheme).
30
Page 30© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 Network radar – 1/2/4 km Composite 09 UTC
31
Page 31© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 Network radar – 1/2/4 km Composite 10 UTC
32
Page 32© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 Network radar – 1/2/4 km Composite 11 UTC 19C 11C
33
Page 33© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 Network radar – 1/2/4 km Composite 12 UTC
34
Page 34© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 Network radar – 1/2/4 km Composite 13 UTC
35
Page 35© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 – 12 UTC 4 km model 10 m wind and convergenceRainfall rate
36
Page 36© Crown copyright 2005 CSIP IOP 18 – 25 th August 2005 – 12 UTC 4 km 12 km Screen Temperature
37
Page 37© Crown copyright 2005 Orographic features important to radiation Oliphant et. al. 2003, ‘Spatial variability of surface radiation fluxes in mountainous terrain’ Characteristics in order of importance: slope aspect, slope angle, elevation, albedo, shading, sky view factor, leaf area index The most important factor is the area presented by each grid-box to the incoming direct SW radiation
38
Page 38© Crown copyright 2005 Grid-box mean slope aspect and angle Slope aspect: Slope angle:
39
Page 39© Crown copyright 2005 4km Mesoscale Unified Model: 8 hr forecast Extra direct SW surface flux:Temperature difference at 1.5 metres:
40
Page 40© Crown copyright 2005 4km Mesoscale Unified Model: 16 hr forecast Extra direct SW surface flux:Temperature difference at 1.5 metres:
41
Page 41© Crown copyright 2005 Summary – Orography and Radiation Included slope aspect and angle into the incoming direct short-wave radiation scheme. Tested in UM with grid resolutions ranging from 60km (global) to 1km (over the southern UK). At high resolution, small (0.5K) surface temperature changes resulting from the scheme can lead to differences in convective initiation and evolution.
42
Page 42© Crown copyright 2005 Variable Resolution An alternative approach to 1-way nesting. Grid varies from coarse resolution at the outer boundaries smoothly to a uniform fine resolution in the interior of the domain Benefits close to hires domain boundary, e.g. reduces spin-up of convection at inflow boundaries Uniform High Res zone Var-Res 2 Var-Res 1 Uniform Coarse Res 1 Uniform Coarse Res 2 Typically, there are 3 regions, and inflation ratio R 1 = R 2 = 5~10% R2R2 R1R1
43
Page 43© Crown copyright 2005 May 3 2002 Case - Variable Resolution Model Rainfall at 14 UTC. The three regions of the variable resolution domain are shown
44
Page 44© Crown copyright 2005 Summary – Variable Resolution Variable resolution grid capability implemented in the UM. Tested in idealised and real case studies with a nesting ratio of 1 : 4 and results look promising. Currently working on the model parametrizations to make them depend more appropriately on the local grid-length in different parts of the domain (e.g. grid-length dependent convection scheme). (More fudges!)
45
Page 45© Crown copyright 2005 Known UM problems 4 km convective cells much too large, too few. (Expected and forecasters have adapted). Latest physics produces acceptable area average precipitation but.. Non-fatal grid-point storms at 4 km. Solutions have run into problems. (No problems at 1 km). Valley cooling problem at all scales. Caused by vertical non-interpolating SL advection. Fixed (we hope!) by selective fully interpolating mod.
46
Page 46© Crown copyright 2005 Model Physics (at present) 12 km/L384 km/L381 km/L76 Timestep300 s100 s30 s Convection Scheme Full Gregory- Rowntree Gregory-Rowntree with restricted mass flux None MicrophysicsPrognostic icePrognostic ice and rain Prognostic ice, rain. Two ice+graupel under test. Surface9 Tile MOSES DiffusionDel 4 theta + Targeted moisture Del 4 To be replaced by 3D turbulence. Boundary Layer / Turbulence Standard 1D (3D Local likely)
47
Page 47© Crown copyright 2005 Precipitation Verification Techniques Philosophy based on assumption that small scales less skilful than large. Rather than doing point by point verification use fractions (~probabilities) over a certain area surrounding each grid point. Calculate various probability and categorical scores based on accumulation thresholds. Basis of products and investigation of skill as function of scale.
48
Page 48© Crown copyright 2005 Radar12 km forecast1 km forecast 0.125 0.5 1 2 4 8 16 32 mm The problem we face 0100 km Six hour accumulations 10 to 16 UTC 13th May 2003
49
Page 49© Crown copyright 2005 Schematic example - different scales
50
Page 50© Crown copyright 2005 1-km forecastRadar 0.125 0.5 1 2 4 8 16 32 mm Six hour accumulations 10 to 16 UTC 13th May 2003
51
Page 51© Crown copyright 2005 4 mm threshold, Fractions at grid scale (1 or 0) ModelRadar > 4 mm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Fraction
52
Page 52© Crown copyright 2005 ModelRadar > 4 mm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Fraction 4 mm threshold, Fractions within 35x35 km squares
53
Page 53© Crown copyright 2005 ModelRadar 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Fraction 4 mm threshold, Fractions within 75x75 km squares
54
Page 54© Crown copyright 2005 ModelRadar 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Fraction 4 mm threshold, Fractions within 105x105 km squares
55
Page 55© Crown copyright 2005 Brier score for comparing fractions Skill score for fractions/probabilities - Fractions Skill Score (FSS) A score for comparing fractions with fractions
56
Page 56© Crown copyright 2005 Graphical behaviour of the Fractions Skill Score
57
Page 57© Crown copyright 2005 Summer 2004 Trial Seven cases from 2004 period (mostly convective) For each case run 4 forecasts at 3 hour intervals Run one suite with 4km, 1km assimilation and a second initialising 4km, 1km from 12km analyses. Forecasts out to T+7 for 1km model Aggregate statistics over forecasts and cases.
58
Page 58© Crown copyright 2005 HRTM Domains Note that operational UK 4km model uses larger (whole UK) domain
59
Page 59© Crown copyright 2005 Area average rain rates over 2004 summer trial Solid: Assim Dotted: Spinup Blue 12km Green 4km Red 1km Black Radar
60
Page 60© Crown copyright 2005 Scores for 6 hour accums 1, 4 and 16mm thresholds Solid: Assim Dotted: Spinup Blue 12km Green 4km Red 1km 1mm / 6hr threshold 4mm / 6hr threshold 16mm / 6hr threshold
61
Page 61© Crown copyright 2005 Scores for 1 hour accums 1 and 4mm threshold Solid: Assim Dotted: Spinup Blue 12km Green 4km Red 1km
62
Page 62© Crown copyright 2005 Intensity/Scale verification Barbara Casati PhD in conjunction with Met Office. Similar ideas to Nigel Roberts – Haar wavelet transform similar to successive ‘box averages’ Summary methods useful for comparison between models.
63
Radar Model forecast from Casati (2004) Radar > 1 mmForecast > 1 mmBinary error image X > uX < u Y > u Hits a False Alarms ba+b Y < uMisses c Correct Rejections d c+d a+cb+da+b+c+d=n
64
wavelet decomposition of the binary error 1 0 Scale from Casati (2004)
65
Page 65© Crown copyright 2005 MSE skill score 1 0 -2 -3 -4 threshold (mm/h) spatial scale (km) [from Casati (2004)] Axes multiples of 2
66
12-18Z 4 km 00Z 6 hr rainfall 12 km 00Z 6 hr rainfall4 km 00Z avg 12 km 6 hr rainfall Error scale (km) 2x2x 16 x 1 mm 64 mm 2x2x 16 x Max radar = 44 mm 68 mm 7 mm 46 mm Rainfall threshold (mm)
67
Page 67© Crown copyright 2005 Distribution-free test as normality of errors can’t be assumed. B = number of +ve skill scores for a given scale and intensity during a given time interval, e.g. 1 month. Hypotheses: H 0 : SS >= 0 (implicit positive and skillful) H 1 : SS < 0 (less skill than a random forecast) H 0 is rejected if b <= b n, where B ~ bi(n, 0.5) for small samples (n < 40), = 0.025 The value of (n – B) / n is shaded in intensity-phase space for each scale and intensity where H 0 is rejected. Modified sign-test statistic
68
Page 68© Crown copyright 2005 Added benefit: comparison of prevalent errors at the monthly time scale (sub-)“grid” scale errors are more prevalent at trace rainfall totals for the 4 km model prevalent errors at twice and four times the 12 km grid length for thresholds > 16 mm are less for the 4 km model (captures large totals better) May 2005 12 km vs radarMay 2005 4 km avg vs radar X X X X X X X X X X X X X X X X 48 km 32 mm
69
Page 69© Crown copyright 2005 Questions?
70
Page 70© Crown copyright 2005 18 UTC 09/12/2003 1km L76 Forecast 24 h loop
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.