Federal Department of Home Affairs FDHA Federal Office

  • Slides: 12
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 Quantitative precipitation forecasts in the Alps – first results from the Forecast Demonstration Project MAP D-PHASE Felix Ament, Marco Arpagaus, and Mathias W Rotach Meteo. Swiss COSMO-2 Radar Lyss (Lyss-Hochwasser, 29. 8. 2007)

Verification – rules of the game Models apply warnlevels domain averages Most recent forecast,

Verification – rules of the game Models apply warnlevels domain averages Most recent forecast, but starting not before +03 h hourly accumulations OBS RR time series 6 times a year Twice a year • Period: Summer 2007 (June, July and August) • Spatial resolution: 18 target regions in Switzerland • Temporal resolution: 3 hour intervals • Forecast range: Use most recent forecast, but ignore a certain cut-off time at the beginning of each forecast (default cut-off: 3 h) First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch Every 10 years Alert time series 2

Observational data RADAR time series Swiss Radar composite • Warn regions averages • 3

Observational data RADAR time series Swiss Radar composite • Warn regions averages • 3 Radar stations • 5 min scans accumulated to hourly estimates • Hourly accumulations Spatial average Multiplicative correction to achieve match of daily accumulations • 1 km resolution Gridded rain gauge data Spatial average RADAR_CAL time series • Statistical interpolation + elevation correction • Warn regions averages • Daily accumulations • Daily sums equivalent to gridded gauge data First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch • Hourly accumulations 3

Verification of precipitation amount (RADAR_CAL reference) resolved conv. global model param. conv. RADAR Whole

Verification of precipitation amount (RADAR_CAL reference) resolved conv. global model param. conv. RADAR Whole Switzerland, summer 2007, relative BIAS Single target region, 3 hourly resolution, correlation First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 4

Fuzzy Verification on coarser scales than model scale: “Do not require a point wise

Fuzzy Verification on coarser scales than model scale: “Do not require a point wise match!“ Method Raw Data Fuzzyfication Score Example result Average X Upscaling X X X Equitable threat score x X X X x Fractional coverage Fraction Skill Score (Roberts and Lean, 2005) X X X x X X X Skill score with reference to worst forecast x First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 5

Fuzzy Verification COSMO-2 – COSMO-7 90 Fraction skill score COSMO-2 (2. 2 km) =

Fuzzy Verification COSMO-2 – COSMO-7 90 Fraction skill score COSMO-2 (2. 2 km) = 20 7 Difference COSMO-7 (7 km) - 33 90 = 58 33 20 7 Threshold (mm/3 h) bad Threshold (mm/3 h) good COSMO-7 better First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch COSMO-2 better 6 Spatial scale (km) Upscaling - 58 Spatial scale (km) JJA 2007, Verification against Swiss Radar Composite, 3 hourly accumulations

Fuzzy Verification COSMO-DE – COSMO-EU 90 Fraction skill score COSMO-DE (2. 8 km) =

Fuzzy Verification COSMO-DE – COSMO-EU 90 Fraction skill score COSMO-DE (2. 8 km) = 20 7 Difference COSMO-EU (7 km) - 33 90 = 58 33 20 Spatial scale (km) Upscaling - 58 Spatial scale (km) JJA 2007, Verification against Swiss Radar Composite, 3 hourly accumulations 7 Threshold (mm/3 h) bad Threshold (mm/3 h) good COSMO-EU better First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch COSMO-DE better 7

Alerts – level „yellow“, 3 h intervals (Alert level yellow = return frequency of

Alerts – level „yellow“, 3 h intervals (Alert level yellow = return frequency of 6 times per year) Relative frequency of an alert (frequency bias) and but Probability to detect an event (probability of detection) Probability to issue a false alarm (false alarm ratio) First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 8

Concept of “Relative Value” Economic point of view: Having no protection results in Losses

Concept of “Relative Value” Economic point of view: Having no protection results in Losses Precautions causes Costs Event Yes No Yes C C No L 0 Relative Value Total Cost 0. 0 - useless 0. 25 0. 75 1. 0 + useful no real perfect forecast First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 9

Relative value – Alert level „yellow“ (03 h, 06 h and 12 h accumulations,

Relative value – Alert level „yellow“ (03 h, 06 h and 12 h accumulations, cut-off +03 h) + useful param. all models conv. resolved conv. RADAR global model - useless insensitive … … against false alarms First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 10

Calibration versus Ensemble Simple calibration of COSMO-2 D-PHASE poormen's ensemble Multiply COSMO-2 precipitation forecasts

Calibration versus Ensemble Simple calibration of COSMO-2 D-PHASE poormen's ensemble Multiply COSMO-2 precipitation forecasts by a factor of Issue an alert, if a certain fraction of all models gives a warning 2. 0 1. 25 1. 0 COSMO-2 0. 8 0. 5 First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 10% 20% 30% 40% 50% 60% 70% 80% 90% 11

Conclusions • Observational uncertainties (QPE) are not significantly smaller than forecast errors (QPF) –

Conclusions • Observational uncertainties (QPE) are not significantly smaller than forecast errors (QPF) – at least for extremes! • High resolution models resolving deep convection tend to perform better than models with parameterized convection. This applies for all international models. • Probabilistic forecasts are useful for customers. However, the method of choice is still unclear: Simple static recalibration and an uncalibrated ensemble forecasting system perform equally! First results of D-PHASE | 6 th COPS Workshop, 27 -29 February 2008, Hohenheim Felix. [email protected] ch 12