Model verification and it’s relevance for forecasters

Slides:



Advertisements
Similar presentations
Validation of Satellite Precipitation Estimates for Weather and Hydrological Applications Beth Ebert BMRC, Melbourne, Australia 3 rd IPWG Workshop / 3.
Advertisements

Diana-Corina BOSTAN National Meteorological Administration ROMANIA.
Monitoring the Quality of Operational and Semi-Operational Satellite Precipitation Estimates – The IPWG Validation / Intercomparison Study Beth Ebert Bureau.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz Institut für Physik der Atmosphäre On the Value of.
Assessment of hailstorms in WRF weather simulations over Switzerland in summer Sensitivity, climatology, comparison with observation data Andrey.
1 COSMO GM, Athens/2007 | E. Zala; D. Leuenberger Update on Weather-Situation Dependent COSMO-7 Verification against Radar Data Results of 2006 Emanuele.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Quantitative precipitation forecasts in the Alps – first.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz New automatic weather type classifications at MeteoSwiss.
Rapid Update Cycle Model William Sachman and Steven Earle ESC452 - Spring 2006.
COSMO General Meeting Zurich, 2005 Institute of Meteorology and Water Management Warsaw, Poland- 1 - Verification of the LM at IMGW Katarzyna Starosta,
COSMO General Meeting – Moscow Sept 2010 Some results from operational verification in Italy Angela Celozzi - Federico Grazzini Massimo Milelli -
The National Environmental Agency of Georgia L. Megrelidze, N. Kutaladze, Kh. Kokosadze NWP Local Area Models’ Failure in Simulation of Eastern Invasion.
Verification Summit AMB verification: rapid feedback to guide model development decisions Patrick Hofmann, Bill Moninger, Steve Weygandt, Curtis Alexander,
How can LAMEPS * help you to make a better forecast for extreme weather Henrik Feddersen, DMI * LAMEPS =Limited-Area Model Ensemble Prediction.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Report on Workshop « Stratified verification by weather.
Latest results in verification over Poland Katarzyna Starosta, Joanna Linkowska Institute of Meteorology and Water Management, Warsaw 9th COSMO General.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Priority project « Advanced interpretation and verification.
Page 1© Crown copyright Scale selective verification of precipitation forecasts Nigel Roberts and Humphrey Lean.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss WG4 activities Pierre Eckert, MeteoSwiss, Geneva.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Quantitative precipitation forecast in the Alps Verification.
Priority project Advanced interpretation COSMO General Meeting, 18. September 2006 Pierre Eckert.
GPS GPS derived integrated water vapor in aLMo: impact study with COST 716 near real time data Jean-Marie Bettems, MeteoSwiss Guergana Guerova, IAP, University.
Deutscher Wetterdienst Fuzzy and standard verification for COSMO-EU and COSMO-DE Ulrich Damrath (with contributions by Ulrich Pflüger) COSMO GM Rome 2011.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz The challenge to verify operational weather warnings.
Observed & Simulated Profiles of Cloud Occurrence by Atmospheric State A Comparison of Observed Profiles of Cloud Occurrence with Multiscale Modeling Framework.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss A more reliable COSMO-LEPS F. Fundel, A. Walser, M. A.
U. Damrath, COSMO GM, Athens 2007 Verification of numerical QPF in DWD using radar data - and some traditional verification results for surface weather.
General Meeting Moscow, 6-10 September 2010 High-Resolution verification for Temperature ( in northern Italy) Maria Stefania Tesini COSMO General Meeting.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz Automatic weather classification at MeteoSwiss Tanja.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Verification results at MeteoSwiss in the year 2010 Francis.
NCAR, 15 April Fuzzy verification of fake cases Beth Ebert Center for Australian Weather and Climate Research Bureau of Meteorology.
VERIFICATION Highligths by WG5. 2 Outlook Some focus on Temperature with common plots and Conditional Verification Some Fuzzy verification Long trends.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz Weather type dependant fuzzy verification of precipitation.
A study on the spread/error relationship of the COSMO-LEPS ensemble Purpose of the work  The spread-error spatial relationship is good, especially after.
10th COSMO General Meeting, Cracow, Poland Verification of COSMOGR Over Greece 10 th COSMO General Meeting Cracow, Poland.
Eidgenössisches Departement des Innern EDI Bundesamt für Meteorologie und Klimatologie MeteoSchweiz Weather type dependant fuzzy verification of precipitation.
WG4 Oct 2006 – Sep 2007 plans COSMO General Meeting, 21 September 2006 Pierre Eckert.
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss WG4 activities Pierre Eckert MeteoSwiss, Geneva.
New results in COSMO about fuzzy verification activities and preliminary results with VERSUS Conditional Verification 31th EWGLAM &16th SRNWP meeting,
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology MeteoSwiss Fuzzy Verification toolbox: definitions and results Felix.
COSMO General Meeting 2011 WG5 Parallel Session 5 September 2011
I. Sanchez, M. Amodei and J. Stein Météo-France DPREVI/COMPAS
A few examples of heavy precipitation forecast Ming Xue Director
LEPS VERIFICATION ON MAP CASES
Intensity-scale verification technique
Fuzzy verification using the Fractions Skill Score
Systematic timing errors in km-scale NWP precipitation forecasts
Spatial Verification Intercomparison Meeting, 20 February 2007, NCAR
Multi-scale validation of high resolution precipitation products
Dan Petersen Bruce Veenhuis Greg Carbin Mark Klein Mike Bodner
Tuning the horizontal diffusion in the COSMO model
A Review of the CSTAR Ensemble Tools Available for Operations
WG5-Report from Switzerland: Verification of aLMo in the year 2005
Daniel Leuenberger1, Christian Keil2 and George Craig2
Application of satellite-based rainfall and medium range meteorological forecast in real-time flood forecasting in the Upper Mahanadi River basin Trushnamayee.
WG4: interpretation and applications A success story… to be continued… Pierre Eckert MeteoSwiss, Geneva.
Sub-daily temporal reconstruction of historical extreme precipitation events using NWP model simulations Vojtěch Bližňák1 Miloslav.
COSMO General Meeting 2009 WG5 Parallel Session 7 September 2009
A. Topographic radiation correction in COSMO: gridscale or subgridscale? B. COSMO-2: convection resolving or convection inhibiting model? Matteo Buzzi.
Thomas Gastaldo, Virginia Poli, Chiara Marsigli
Quantitative verification of cloud fraction forecasts
Linking operational activities and research
Numerical Weather Prediction Center (NWPC), Beijing, China
Christoph Gebhardt, Zied Ben Bouallègue, Michael Buchhold
2007 Mei-yu season Chien and Kuo (2009), GPS Solutions
Some Verification Highlights and Issues in Precipitation Verification
Some ideas on verification targeted to the use of the forecasters Pierre Eckert, MeteoSwiss, Geneva Some ideas on the presentation to the forecasters.
Weather Dependent Verification and link with forecasters
Short Range Ensemble Prediction System Verification over Greece
VERIFICATION OF THE LAMI AT CNMCA
Presentation transcript:

Model verification and it’s relevance for forecasters Dublin 1.10.2010

Goals Models are becoming more accurate, however; The forecasters have access to many models without knowing their weaknesses and limitations.. The roles of the forecaster and the modeller are still very divided. The actual model verification is performed over long periods or averaged over large area. This only provides general global information, but doesn’t meet the specific needs of the forecaster. The forecaster would like to know specifically when to trust the accuracy of the model. In COST733, the evaluation of the different weather types was done by looking at the ability to represent different precipitation patterns in the Alpine domain. For verification purposes we are more interested in differentiating weather classes where the models have difficulties from those where it performs well.

History Alps orography T213 and T106 In 1987 only 1 model was available in Switzerland (ECMWF) The orography was coarse, a good knowledge of the local climate was crucial to provide a forecast. The forecaster knew the weakness of the model.

Which model to trust?

COSMO2 Design IFS/ECMWF, 16 km, synoptic scale 4 daily updates COSMO2 Design COSMO 7km, 385x325x45, regional scale Own assimilation cycle (nudging) 2 daily 72h forecast COSMO 2.2km, 480x350x60, local scale Own assimilation cycle (nudging) 8 daily 18h nested forecast forecast range clock time +18h 00 UTC 01 UTC 3h 03 UTC 04 UTC 06 UTC 07 UTC

Verification available (SYNOP, spring)

Verification available (Upper air)

Stratification by weather type

New weathertype classification for verification OLD: Manual classification by E. Zala based on 500 hPa and surface charts NEW: automatic classification based on IFS analyses (Internship D. Jean-Mairet)  11 classes: 8 wind directions + flat, high, low pressure  63 % agreement with Zala classification for the years 2002-2008

automatic classifikation gives more N, less H Weather types for the years 2002-2009 (Zala classification and new automatic classification in comparison) automatic classifikation gives more N, less H

Neighborhood Verification „multi-scale, multi-intensity approach“ calculation of score for growing windows and various intensity thresholds suitable for precipitation verification „Fuzzy Verification Toolbox“ of Beth Ebert (http://www.bom.gov.au/bmrc/wefor/staff/eee/beth_ebert.htm) fcst obs

Fuzzy Verification - Methods Upscaling - ETS Fraction Skill Score - FSS Roberts & Lean (2005) Score: FSS for fractions / probabilities (0 = mismatch, 1 = perfect match) Score: ETS (Equitable Threat Score) (-1/3 = mismatch, 1 = perfect match) „Good verification approach to evaluate how well the forecast mean value agrees with the mean value of the observations.“ „Spatial averaging damps out areas with high rain rates and spreads areas of lower rain rates.“ (B.Ebert,2009) „FSS is ideally suited for considering the overall performance and aggregating results over many forecasts” (Mittermaier,2009) „The FSS is a good measure of the spatial accuracy of precipitation forecasts.“ (Mittermaier,2009)

How to read Neighborhood plots… best skill for low thresholds and large spatial scales numbers / colors  Value of the score, hier: FSS good bad increasing spatial scale bold numbers  useful scales (only FSS) low skill for high thresholds and small spatial scales increasing threshold

Results for 2009 Model Observations 3 h acc.: leadtimes 04 – 06 h for each model run of both models (00,03,06,12,15,18,21 UTC), „fairer comparison“ Observations 3 h acc.: Swiss radar composit (NASS product); in the case of missing data, the whole day is omitted (26 days)

Skill of COSMO models, 2009 Fractions Skill Score (FSS) Spatial scale = 100 km (COSMO-7: 15 grdpts, COSMO-2: 45 grdpts) COSMO-2 thresholds

Differences of FSS, 2009 COSMO-2 minus COSMO-7 NE S N SW F H NW E SE W COSMO-7 better COSMO-2 better N SW F H NW E SE W L

Example NW, 2009 COSMO-2 better especially for high thresholds… COSMO-2 - COSMO-7 COSMO-7 better COSMO-2 better good bad

Other application of weathertypes Radar verification (only COSMO-7 at the moment)

Guideline

Conclusions of the COST meeting in June: Recommendations for guidelines The guidelines should be self-contained (without links). They could look like a cookbook, for example for the use of a parameter or for the treatment of a specific situation. A light version can be at the disposal of the forecaster on duty (usually under time stress) when a longer version can be studied offline. This longer version can also be used as an education tool for newcomers. The shorter version can also be implemented as a seasonal factsheet. The seasonal factsheets should include (if possible) the expected changes of the current model version with respect to the version which was running the previous season. Generally speaking the guidelines should be short, attractive and meaningful.

Forecasters feedback Forecaster feedback should be organized either by mailbox, a forum or regular discussions. At the end of each season a debriefing could be organized and a synthesis written. This could form a good base for the following corresponding season.

Suggestions for verification

Day to day verification (for rainfall and sunshine duration)

Synthetic map

Advice Northerly situation: Precipitation generally overestimated.

Conclusions The verification should be simple and easy to access. It is important to have a targeted verification of models, not averaged over long periods and large areas such as: Verification by weather situation. Verification by season. Verification by climatic region. A day to day verification could allow analysis a posteriori of a forecast considering the elements available at the time. The forecaster would like to know for which situation one model is better than another; which model is better to predict cut-off or ending an Omega block (IFS, GFS)? With the stratify verification the forecaster can judge the capability of the model according to the weather situation. A feedback from the forecasters plays a crucial role. The model physics is frequently upgraded or modified; making it difficult for the forecaster to keep pace with the model’s bias.

Thank you

… Upscaling (UP) 1. Principle: (Zepeda-Arce et al., 2000) 1. Principle: Define box around region of interest and calculate the average of observation and forecast data within this box. 2. Contingency Table Rave observation 3. Equitable Threat Score (ETS) yes no Hit False Alarm Miss Correct negative forecast Q: Which fraction of observed yes - events was correctly forecast?

… Fraction Skill Score (FSS) (Roberts and Lean, 2005) 1. Principle: Define box around region of interest and determine the fraction pj and oj of grid points with rain rates above a given threshold. 2. Probabilities X x 3. Skill Score for Probabilities Q: On which spatial scales does the forecast resemble the observation?

… Fraction Skill Score (FSS) (Roberts and Lean, 2005) 4. Useful Scales useful scales are marked in bold in the graphics

Zusammenfassung Was lernen wir aus den Fuzzy-Ergebnissen? beide COSMO - Modelle haben Skill gute Vorhersage der räumlichen Struktur auf grösseren Skalen (hohe FSS-Werte) der Skill der Modelle variiert stark für verschiedene Wetterlagen und auch der Unterschied zwischen COSMO-2 und COSMO-7 ist unterschiedlich stark ausgeprägt Bester Skill: Frühsommer und Herbst bzw. Süd- und Westlagen Grösste Differenz COSMO-2 minus COSMO-7: Sommer (Mai bis September) bzw. Nord- und Westlagen sowie konvektive Lagen Was bleibt unklar? Interpretation der Egebnisse des Intensity Scale: Gibt es einen räumlichen Shift im COSMO-2?

Example Case 06.06.2009, 03-06 UTC

Example Case 01.07.2009, 12-15 UTC

3. Useful Scales, 2009 (FSS) COSMO-2, gridpoints* 2.2 km COSMO-7, gridpoints* 6.6 km # cases dx 0.1 0.2 0.5 1.0 2.0 5.0 10 20 NE 9 15 27 / N 63 NW 3 SE 45 S 1 SW E W F H L dx 0.1 0.2 0.5 1.0 2.0 5.0 10 20 NE 5 9 15 / N NW 3 SE 21 S SW E W F H L 10 Determination of minimum spatial scale for each threshold und weathertype, where the forecast can be considered „useful“ according to the definition of the FSS; i.e. the forecast is better than a constant forecast of the wet area ratio. The lower limit of the useful scales is defined by L(FSS lt 0.5+f/2) where f is the observed fractional rainfall coverage over the domain (wet area ratio). 41 41 4 11 59 4 74 Bestimmung der kleinsten Skala [gridpoints] für die die Vorhersage nach der Definition der FSS „useful scales“ nützlich gewesen ist… 59 31 31 17 14 10 7 4 1 0.4 <0.1 17 14 10 7 4 1 0.4 <0.1 % obs gridpts >= thresh (whole period)

Bias correction map against rain gauges (long term corrections) Data sets and methodology RADAR: 24h precipitations sums from 06 to 06 UTC MeteoSwiss network composit "rain" product horizontal resolution 2x2km visibility-, clutter- and bias- corrected                       !!! Note that this bias correction is currently based on precipitation data of 2004! An updated verification with corrections valid for 2008 will follow soon. !!! Bias correction map against rain gauges (long term corrections)

COSMO-7 model: +06h to +30h  precipitations sums from 00UTC runs horizontal resolution 7x7km Weather Type classification: ele classification subjective mainly based on 500hPa geopotential/winds classes: north, northeast, east, southeast, south, southwest, west, northwest, high, low, flat new, automatic weather type classification (mo classification) based on IFS analyses valid at 12 UTC and with a resolution of 1° first step: distinction between advective and convective classes based on 500hPa wind speed averaged over the alpine domain. > 11 m/s : advective, < 11m/s convective the advective classes are further divided into the main 8 wind directions based on the GWT (GrossWetterTypes)-Method developed in the COST-Action733. To this end the 500hPa geopotential is correlated with three prototype patterns (Figure below). Based on the correlation coefficients of the zonal and meridional patterns the wind direction is determined. The vorticity pattern is not used. the convective classes are divided into flat, high and low pressure situations according to the following rules: average pressure over Alpine domain >= 1019 hPa : high average pressure over Alpine domain < 1012 hPa: low average pressure over Alpine domain between those thresholds and all types with 500hPa wind speeds lower than 5 m/s: flat.