How can LAMEPS * help you to make a better forecast for extreme weather Henrik Feddersen, DMI...
Transcript of How can LAMEPS * help you to make a better forecast for extreme weather Henrik Feddersen, DMI...
How can LAMEPSHow can LAMEPS** help you to make help you to make a better forecast for extreme weathera better forecast for extreme weather
Henrik Feddersen, DMIHenrik Feddersen, [email protected]@dmi.dk
*LAMEPS = Limited-Area Model Ensemble Prediction System
OutlineOutline
MotivationMotivation Ensemble methodologyEnsemble methodology Rainfall case studiesRainfall case studies Probability upscalingProbability upscaling Verification, summer 2011Verification, summer 2011 Wind case studyWind case study SummarySummary
MotivationMotivation
Assess forecast uncertaintyAssess forecast uncertainty Assess risk of high-impact weatherAssess risk of high-impact weather
Heavy rain (>24mm/6h)Heavy rain (>24mm/6h) Cloudburst (>15mm/30min)Cloudburst (>15mm/30min) Heavy snowfall (>15mm/6h)Heavy snowfall (>15mm/6h) Snowstorm (>10mm/6h and >10m/s)Snowstorm (>10mm/6h and >10m/s) Storm (mean and gust) (>24m/s)Storm (mean and gust) (>24m/s) Hurricane (mean and gust) (>32m/s)Hurricane (mean and gust) (>32m/s)
Ensemble methodologyEnsemble methodology
Sample uncertainty in Forecast initial conditions Model formulation Lateral boundary conditions
Initial conditionsInitial conditionsPerturbation of analysisPerturbation of analysis
ObservationsObservations AnalysisAnalysisObservationsObservations ForecastForecast
Perturbed analysesPerturbed analyses Ensemble membersEnsemble members
Initial conditionsInitial conditionsPerturbation of observationsPerturbation of observations
ObservationsObservations AnalysisAnalysisObservationsObservations ForecastForecast
Ensemble data assimilationEnsemble data assimilation Ensemble membersEnsemble membersPerturbed Perturbed observationsobservations
Model uncertaintyModel uncertaintyMulti-model ensembleMulti-model ensemble
NWP Model A
Model B
Model C
Model uncertaintyModel uncertaintyStochastic physicsStochastic physics
Initialization Dynamics
Postprocessing
Physics
NWP model
Model uncertaintyModel uncertaintyMulti-scheme ensembleMulti-scheme ensemble
Turbulence Surface 1
Surface 2
RadiationConvection 1
Convection 2
NWP model physics
Model uncertaintyModel uncertaintyMulti-parameter ensembleMulti-parameter ensemble
Turbulence Surface
Radiation
NWP model physics
Convection
Limited-area ensembles vs Limited-area ensembles vs global ensemblesglobal ensembles
Global ensembles Uncertainty in synoptic development in the medium-
range Limited-area ensembles
Uncertainty in mesoscale development in the short-range
DMI-HIRLAM Ensemble DMI-HIRLAM Ensemble Prediction SystemPrediction System
Resolution = 0.05Resolution = 0.05° horizontal / 40 vertical levels° horizontal / 40 vertical levels
Members = 25Members = 25
Forecast length = 54hForecast length = 54h
Forecast frequency = 4 times per dayForecast frequency = 4 times per day
Initial and lateral boundary conditions = 5Initial and lateral boundary conditions = 5
Scaled Lagged Average Forecast (SLAF) error perturbationsScaled Lagged Average Forecast (SLAF) error perturbations
Cloud schemes = 2Cloud schemes = 2
STRACO and KF/RKSTRACO and KF/RK
Stochastic physics = yes/noStochastic physics = yes/no
Surface schemes = 2Surface schemes = 2
ISBA and ISBA/NewsnowISBA and ISBA/Newsnow
Independent of ECMWF's ensemble prediction systemIndependent of ECMWF's ensemble prediction system
Short-range ensemble spreadShort-range ensemble spread
Short-range ensemble spreadShort-range ensemble spread
Case study, 2 July 2011Case study, 2 July 2011Precipitation stamp mapPrecipitation stamp map
Case study, 2 July 2011Case study, 2 July 2011Probability mapProbability map
Probability = 10-20%: Only 3-4 members Probability = 10-20%: Only 3-4 members predict the event!?predict the event!?
Case study, 2 July 2011Case study, 2 July 201150 50 mm/6h contoursmm/6h contours
More than 4 members predict the event!More than 4 members predict the event!
Different members Different members in different coloursin different colours
Probability upscalingProbability upscaling
Conventional probabilityConventional probability— In every grid point: Fraction of members that predict In every grid point: Fraction of members that predict
the eventthe event Upscaled probabilityUpscaled probability
— In every grid point: Fraction of members that predict In every grid point: Fraction of members that predict the event the event in a neighbourhood in a neighbourhood of the grid pointof the grid point
— Probability that the event will happen Probability that the event will happen somewheresomewhere near grid pointnear grid point
Probability upscaling exampleProbability upscaling example
Prob = 1/25
Prob = 3/25
Prob = 8/25
Upscaled probabilitiesUpscaled probabilities
Max probability > 40%Max probability > 40%
Upscaling diameter = 15 grid cells ~ 80 kmUpscaling diameter = 15 grid cells ~ 80 km
Verification of 2 July 2011 caseVerification of 2 July 2011 case
Note the agreement between locations of Note the agreement between locations of max probability and max observed rainfall!max probability and max observed rainfall!
Observed
Alternative verificationAlternative verification
Location of ensemble member maximaLocation of ensemble member maxima
Where is max precip most likely?Where is max precip most likely?
Where density of ensemble members is highest!Where density of ensemble members is highest!
Upscaling method will show just that!Upscaling method will show just that!
Heavy rainfall examplesHeavy rainfall examples
Some things to consider...Some things to consider...
At what probability threshold should you take At what probability threshold should you take action?action?
How does forecast skill depend on forecast How does forecast skill depend on forecast range?range?
How many false alarms can you expect?How many false alarms can you expect?
Verification, JJA 2011Verification, JJA 2011Relative operating characteristicRelative operating characteristic
Hit ra
te =
events correctly foreacast / events occurred
False alarm rate = events falsely foreacast / events non-occurred
Perfect forecastPerfect forecast
(FAR,HR) if forecast, when prob > 1/25(FAR,HR) if forecast, when prob > 1/25
(FAR,HR) if forecast, when prob > 2/25(FAR,HR) if forecast, when prob > 2/25
Relative operating characteristicRelative operating characteristicUpscaling vs No upscalingUpscaling vs No upscaling
NB. False alarms are acceptable, if they are NB. False alarms are acceptable, if they are accompanied by nearby hits for the same forecast!accompanied by nearby hits for the same forecast!
Relative operating characteristicRelative operating characteristicForecast skill as a function of lead timeForecast skill as a function of lead time
Relative operating characteristicRelative operating characteristicEnsemble vs DeterministicEnsemble vs Deterministic
Threat scoreThreat score
TS= hitshitsmissesfalse alarms
If probability If probability threshold = 50%...threshold = 50%...
False alarmFalse alarm
HitHitMissMiss
Simplified threat scoreSimplified threat score
HIT = 1 if at least one hitHIT = 1 if at least one hit FA = 1 if at least one false alarm and no hitsFA = 1 if at least one false alarm and no hits MISS = 1 if at least one missMISS = 1 if at least one miss
Count for each forecast...Count for each forecast...
Purpose: Find optimal probability threshold Purpose: Find optimal probability threshold for which an event should be forecastfor which an event should be forecast
Simplified threat scoreSimplified threat score
The threat score is maximized if warnings are The threat score is maximized if warnings are issued when the forecast probability issued when the forecast probability ≥ 20%≥ 20%
Simplified threat scoreSimplified threat score
The threat score is maximized if warnings are only The threat score is maximized if warnings are only issued when the max forecast probability issued when the max forecast probability >> 45% 45%
Issue a warning only if the max forecast probability exceeds Issue a warning only if the max forecast probability exceeds a certain threshold (will reduce false alarms)a certain threshold (will reduce false alarms)
Guidelines to forecastersGuidelines to forecasters
20-50% probability: Pay attention20-50% probability: Pay attention > 50% probability: Take action> 50% probability: Take action
Why not use ECMWF's ensemble Why not use ECMWF's ensemble prediction system?prediction system?
LAM-EPS 50mm contoursLAM-EPS 50mm contours ECMWF-EPS 50mm contoursECMWF-EPS 50mm contours
Why not use ECMWF's ensemble Why not use ECMWF's ensemble prediction system?prediction system?
LAM-EPS 50mm contoursLAM-EPS 50mm contours ECMWF-EPS 15mm contoursECMWF-EPS 15mm contours
50mm/10km50mm/10km22 vs 15mm/1000km vs 15mm/1000km22
Wind case, 8 Feb 2011Wind case, 8 Feb 2011Deterministic model vs ensemble meanDeterministic model vs ensemble mean
24h forecast24h forecast
Wind case, 8 Feb 2011Wind case, 8 Feb 2011Deterministic model vs ensemble meanDeterministic model vs ensemble mean
12h forecast12h forecast
Wind case, 8 Feb 2011Wind case, 8 Feb 2011Deterministic modelDeterministic model
““Truth”Truth”
SummarySummary
Limited-area, high-resolution, short-range ensemble forecasts Limited-area, high-resolution, short-range ensemble forecasts can provide guidance for extreme weather for can provide guidance for extreme weather for localizedlocalized events events
Particularly useful for forecasting the Particularly useful for forecasting the locationlocation of extreme of extreme events, e.g. convective rainfall events (using the upscaling events, e.g. convective rainfall events (using the upscaling method)method)
Guidelines must be provided for the usage of probabilistic Guidelines must be provided for the usage of probabilistic forecasts of extreme events, e.g.forecasts of extreme events, e.g.
20-50% probability: pay attention20-50% probability: pay attention > 50% probability: take action> 50% probability: take action
Upscaled probability forecasts have been used frequently by Upscaled probability forecasts have been used frequently by DMI forecasters for guidance this summerDMI forecasters for guidance this summer
Focus on rainfall events, but also potential for wind eventsFocus on rainfall events, but also potential for wind events