Federal Department of Home Affairs FDHA Federal Office

  • Slides: 28
Download presentation
Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss WG 4: interpretation and applications overview Pierre Eckert Meteo. Swiss, Geneva

Topics • Sochi Olympic games PP CORSO • FIELDEXTRA presentation by JM Bettems •

Topics • Sochi Olympic games PP CORSO • FIELDEXTRA presentation by JM Bettems • Postprocessing • • CORSO Kalman filter COSMO-MOS CAT diagnostics Use of chekclist • Guidelines • Plans COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 2

Priority project CORSO • Task 1: implementation of high resolution model • Task 2:

Priority project CORSO • Task 1: implementation of high resolution model • Task 2: postprocessing and usability • Task 3: development of EPS COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 3

Priority project CORSO • Task 1: implementation of high resolution model • Task 2:

Priority project CORSO • Task 1: implementation of high resolution model • Task 2: postprocessing and usability • Task 3: development of EPS COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 4

RESULT FROM KALMAN FILTERING OF T 2 M I. Rozinkina, S. Cheshin, M. Shatunova,

RESULT FROM KALMAN FILTERING OF T 2 M I. Rozinkina, S. Cheshin, M. Shatunova, I. Ruzanova Hydrometeorological Research Center of Russia

Kalman Filter The temperature T observed at the station at time t is represented

Kalman Filter The temperature T observed at the station at time t is represented as where Tp is the window width used for expanding the temperature forecasts (4 -7 days) and Np is the number of harmonics used (Tp multiplied by (1, 2, or 3)), The difference D between the observed temperature and the averaged forecast at time t is represented as where Td is the window width used for expanding D (1 day) and Nd is the number of harmonicas used (1, 2, or 3), The forecast at the time t is calculated using the formula COSMO General Meeting 2012, Lugano, September, 10 -13

Results Corrected 2 m temperature for Tp=7 days, Td=1 day, and various Np and

Results Corrected 2 m temperature for Tp=7 days, Td=1 day, and various Np and Nd The 2 m temperature forecast at Krasnaya Polyana station was corrected over February, 2012 by applying the described method, The errors in the initial forecasts: average deviation: 2, 86 K root-mean-square deviation: 3, 89 K For Np=7 and Nd=1, the errors of the corrected forecasts: average deviation: 0, 18 K root-mean-square deviation: 2, 55 K For Np=14 and Nd=2, the errors of the corrected forecasts: average deviation: 0, 40 K root-mean-square deviation: 2, 3 K For Np=21 and Nd=3, the errors of the corrected forecasts: average deviation: 0, 39 K root-mean-square deviation: 2, 19 K • observed data • T 2 forecasts • revised T 2 forecast COSMO General Meeting 2012, Lugano, September, 10 -13

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss Local forecasts with COSMO-MOS Concept, Performance and Implementation ECAC & EMS, September, 14 th 2010 COSMO-GM, Lugano, 10. 09. 2012 Vanessa Stauch

Objectives of statistical PP ü To complement the NWP forecasts with the information in

Objectives of statistical PP ü To complement the NWP forecasts with the information in observations ü To reduce systematic NWP forecast errors, e. g. due to simplified (small scale) processes, incorrect (smoothed) local forcing, … ü To calibrate (ensemble) forecasts such that they are reliable and sharp ü To derive forecasts for variables that are not predicted by the NWP model taken from Wilks 2005 COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 9

Dilemma global MOS Length of training period ~ “MOSMIX” + insensitive to model error„MOSMIX“:

Dilemma global MOS Length of training period ~ “MOSMIX” + insensitive to model error„MOSMIX“: multiple linear regression based changes COSMO- simple error model, little MOS discrimination MOS complexity + sampling of many cases, Updateable good MOS discrimination → long lead times, rare events » correction mainly of the systematic errorupdate online “KF” “UMOS” - inert when model error changes on global NWP models “UMOS”: ‘updateable’ MOS of Canadians (and Austrians), weighting of model versions “KF”: Kalman Filter based online update of systematic error correction » reduction of the mean error and Temporal flexibility (e. g. its variability change of model version) COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 10

Implemented statistical approaches ü Multiple linear regression with stepwise forward model selection ü Logistic

Implemented statistical approaches ü Multiple linear regression with stepwise forward model selection ü Logistic regression (returns probability of exceedance for one threshold q) ü Extended logistic regression (Wilks, 2009, returns entire probability distribution of forecast) COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 11

Data sampling & estimation strategies COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 12

Data sampling & estimation strategies COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 12

10 m wind speed: setup comparison 01. 12. 2010 – 28. 02. 2011 COSMO-7

10 m wind speed: setup comparison 01. 12. 2010 – 28. 02. 2011 COSMO-7 COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 13

10 m wind speed: setup comparison 01. 12. 2010 – 28. 02. 2011 COSMO-2

10 m wind speed: setup comparison 01. 12. 2010 – 28. 02. 2011 COSMO-2 COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 14

Summary multiple linear regression ü MOS forecasts reduce forecast error variance and systematic error

Summary multiple linear regression ü MOS forecasts reduce forecast error variance and systematic error ü In comparison to Kalman filter approach, effect on the error variance much higher ü Comparison COSMO-2 with COSMO-7 shows positive effect of higher resolved (=better) inputs ü Recommendation for production setup: • training period: 50 days for temperature, 90 days for wind speed • daytime dependent coefficients, all runs. • update once a day COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 15

COSMO-MOS: Performance and recommendations RESULTS WITH EXTENDED LOGISTIC REGRESSION COSMO-MOS | COSMO-GM, 10. 09.

COSMO-MOS: Performance and recommendations RESULTS WITH EXTENDED LOGISTIC REGRESSION COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 16

Simulation setup 10 m wind gusts Verification period: 01. 09. 2010 – 02. 11.

Simulation setup 10 m wind gusts Verification period: 01. 09. 2010 – 02. 11. 2010 Hourly wind gust observations from the Swiss automatic measurement network (~70 stations used) Thresholds for estimation: 25, 50, 75 % quantiles COSMO-2 time lagged ensemble “eps”: median and std as predictors “lag”: all members separate predictors COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 17

Overall comparison CRPSS COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 18

Overall comparison CRPSS COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 18

Summary 10 m wind gusts ü Extended logistic regression is a suitable statistical model

Summary 10 m wind gusts ü Extended logistic regression is a suitable statistical model for deriving PDFs from deterministic model output ü COSMO-2 time lagged ensemble does contain useful ensemble information for statistical post-processing ü Leadtime dependency of “eps” approach apparent but might be alleviated with longer runs (→ COSMO NEx. T? ) ü Training periods need to be seasonal → maybe include more years in order to improve the distributions COSMO-MOS | COSMO-GM, 10. 09. 2012 Vanessa Stauch 19

IACETH Clear Air Turbulence over Europe: Climatology, Dynamics and Representation in COSMO-7 Masterthesis of

IACETH Clear Air Turbulence over Europe: Climatology, Dynamics and Representation in COSMO-7 Masterthesis of Lysiane Mayoraz Supervised by Michael Sprenger and Vanessa Stauch

IACETH Turbulence indices: • TI 2 (Ellrod & Knapp Index 2) → deformation, shearing

IACETH Turbulence indices: • TI 2 (Ellrod & Knapp Index 2) → deformation, shearing und divergence • RI (Gradient Richardson Number) → rate between the static stability and the vertical windshear. If RI < 1: instable • EDR (Eddy Dissipation Rate) → rate at which turbulent kinetic energy is converted into heat → Turbulent spot well visible with the three indices calculated from the COSMO-7 forecasts! → But signal too low (~ 1'000 m) 15/05/2012 Clear Air Turbulence over Europe / Masterthesis / Lysiane Mayoraz 21

IACETH Turbulence indices: • TI 2 (Ellrod & Knapp Index 2) → deformation, shearing

IACETH Turbulence indices: • TI 2 (Ellrod & Knapp Index 2) → deformation, shearing und divergence • RI (Gradient Richardson Number) → rate between the static stability and the vertical windshear. If RI < 1: instable • EDR (Eddy Dissipation Rate) → rate at which turbulent kinetic energy is converted into heat Without extended turbulence parametrisation Extended turbulence parametrisation: brings a significant amelioration compared to the operational forecasts 15/05/2012 Clear Air Turbulence over Europe / Masterthesis / Lysiane Mayoraz 22

IACETH Observations Data Flight Data Monitoring Data from Swiss (year 2011) Selection criteria 50

IACETH Observations Data Flight Data Monitoring Data from Swiss (year 2011) Selection criteria 50 turb. events (out of 100'000 flights) 15/05/2012 Clear Air Turbulence over Europe / Masterthesis / Lysiane Mayoraz 23

IACETH Comparison Observations / Model Results: All clear detected events are associated with a

IACETH Comparison Observations / Model Results: All clear detected events are associated with a large and long-lasting event from the model! 15/05/2012 Clear Air Turbulence over Europe / Masterthesis / Lysiane Mayoraz 24

Check list «risk of thunderstorms» COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss.

Check list «risk of thunderstorms» COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 25

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss Guidelines http: //www. wmo. int/pages/prog/ www/manuals. html

 • 2. WHY SHOULD WE USE EPS? • 3. TYPES OF EPS •

• 2. WHY SHOULD WE USE EPS? • 3. TYPES OF EPS • 3. 1 Global EPS • 3. 2 Regional EPS • 3. 3 Convective-scale EPS • 6. USE OF EPS IN DETERMINISTIC FORECASTING • 6. 1 Decision-making from deterministic forecasts • • 7. SCENARIOS 8. FULL PROBABILISTIC FORECASTS 9. POST-PROCESSING 10. USE OF EPS IN PREDICTION OF SEVERE WEATHER AND ISSUE OF WARNINGS • 11. SEVERE WEATHER IMPACT MODELLING • 13. FORECASTER TRAINING COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 27

Plans Aviation • COSMO-MOS: visibility, ceiling, wind direction • Improve and operationalise CAT forecasts

Plans Aviation • COSMO-MOS: visibility, ceiling, wind direction • Improve and operationalise CAT forecasts • Other applications First guess into forecast matrix • «Best» deterministic input temperature, wind, sunshine duration, precipitation, … • Estimates for probabilities (compatible with deterministic) Guidelines • Strenghts and weaknesses of the various models • Use of O(1 km) models, use of O(2 km) EPS Exchange of experiences and methods COSMO General meeting ¦ Lugano, September 2012 Pierre. Eckert[at]meteoswiss. ch 28