Grupo de Meteorologia e Climatologia na Universidade de Aveiro Alfredo Rocha, Tiago Luna, Juan Ferreira, Ana Carvalho and João Sousa Departamento de Física Predictability of Extreme Weather Events
Some research centres do research in meteorology (i.e. weather forecasting) and climate. This effort should produce added value to better forecast of extreme events (amongst others) as a public service. This requires strong and well defined colaboration between Universities and State Laboratories. It is also needed more critical mass of high quality in meteorology/climate research. Model validation Ensemble forecasting Forecast verification Data assimilation Major tasks which require collaboration
How can research (at Unis and RUs) contribute to better prediction of Extreme Weather Events 1.Flash floods and land slides in Madeira on the 20 th February Wind storm in the Torres Vedras region on the 23 rd December Precipitation event in the Lisbon/Setúbal region on the 18 th February case-studies: These events have not been (fully) predicted by operational weather forecast institutions
Questions: 1. Was the event predictable? 2. If yes: How long before? With maxima correct? Without phase error? 3. Origin: Synoptic or ographic? 5. Relevance of model horizontal resolution and parametrizations? 1. Extreme precipitation event - Madeira – 20th February 2010
The cause of the catastrophe was not just meteorological There have been similar precipitation amounts in the past without catastrophe Model WRF-ARW 3.1 Initial and Boundary conditions Forecasts (3h-3h) and Analyses (6h-6h) from GFS 0.5ºx0.5º
Initial Conditions Run 12h Run6 12h Run5 12h Run4 12h Run3 12h Run2 12h Run1 00h Run0 06h Run00 00h Run_prev Run_Anals Run7 - S/Cum 00h Run8 – C/Cum 00h Run_stopo D1 – 25 x 25 km D2 – 5 x 5 km D3 – 1 x 1 km D1 – 25 x 25 km D2 – 5 x 5 km D1 – 25 x 25 km D2 – 5 x 5 km operational Simulations with Weather Research and Forecasting (WRF) model
Madeira ‘as seen’ by WRF 1 km horizontal resolution max ~ 270 m – A model is always a model! z max ~ 270 m – A model is always a model! IM stations with hourly prcp data
Were ICs and BCs (GFS forecasts) good? Diference (%) of 3-D integrated precipitable water GFS forecasts and GFS Analyses ‘Break’ of predictability? Some predictability at 3 days?
Accumulated Prcp (mm/day), on the 20.02, D03 – 1 x 1 km ICs - 00h Topography determinant Similar maxima for 5 x 5 km
Phase errors ICs 00h Average PRCP (mm/day) - south coast 2-3 h
Errors – D2 (5 km) ICs 12h 13 12h 14 12h 15 12h 16 12h 17 00h h 18 06h 19 00h 19
Errors – D3 (1 km) ICs 00h 20.02
Local PRCP run 7 (D03) / IM (without cumulus parametrization) Phase and amplitude +/-correct! ICs 00h 20.02
Local PRCP run 8 (D03) / IM (with cumulus parametrization.) Phase and amplitude +/- correct! ICs 00h 20.02
2d 3d 5d 4d 6d 7d 36h 24h 12h
PRCP (mm/h) - Funchal ICs from 12h of (7 days) till 06h of (1 day) Obs - IM Daily amounts +/- OK, but intensity wrong!
Conclusions: Some predictability from 3 days. ICs e BCs (GFS forecasts) are determinant Horizontal resolution not important for regional PRCP but important for local PRCP Precipitation of orographic origin Local PRCP +/- correct (phase and amplitude) only for ICs 12h prior to the event Cumulus parametrization not important
2. Wind storm in the Torres Vedras region on the 23rd December 2009 Max wind gust at IM Stations: Torres Vedras – km/h – 4:40 am Cabo Carvoeiro – km/h – 4:50 am Max wind gust estimated by IM using Doppler Radar: ~ 200 km/h propagating SW to NE
Simulations with WRF: 1 x 1 km horizontal resolution 25 to 85 vertical levels ICs 00h Parametrization of wind gusts (ECMWF Newsletter, 119, Spring 2009, 15-18). Operational at WCMWF since Sep Total gust = turbulent gust + convective gust Applied to WRF output
m/s Max. (WRF) inst. Windspeed 00h-06hMax. Turbulent wind gust 00h-06h X X X X Max ~70 km/h
m/s Max. Total wind gust 00h-06hMax. Convective wind gust 00h-06h Max >160 km/h Max ~20 km/h
3. Precipitation event in the Lisbon/Setúbal region on the 18th February mm/day - Record daily PRCP at Lisboa/Geofísico 36 mm/h am in Lisbon 40 mm/h 11 – 12 am in Setúbal
Observations
Two WRF operational spatial configurations/resolution OP1 and OP2 Two different sets of parametrizations Assimilation of radiosonde and surface data for one domain/param. configuration 5 WRF simulations
OP2 L (d01)= 21 km L (d02) = 7 km Vertical levels = 26 OP1 L (d01) = 25 km L (d02) = 5 km Vertical levels = 27
WRF Set 1 – PRCP UTC
WRF Set 2 – PRCP UTC
WRF Set 1 + Assim. – PRCP UTC…..not good!