Current Products and Future Plans for the Expanded UW Short-Range Ensemble Forecast System

Eric P. Grimit

Department of Atmospheric Sciences, University of Washington

Seattle, Washington

Current Products and Future Plans for the Expanded UW Short-Range Ensemble Forecast System

Mean & Std. Dev. for sea-level pressure at F24

5 mb4 mb

Probability of surface wind speed > 21 kt (SCA)

Probability of 12h accum. precip. > 0.01” (rain/no-rain)

PSCZ

Future Research Plans

Evaluate the expanded UW MM5 SREF system and investigate multimodel applications

Develop a mesoscale forecast skill prediction system

Additional Work mesoscale verification probability forecasts deterministic-style solutions additional forecast products/tools (visualization)

•Evaluate the expanded UW MM5 SREF systemCompare skill scores, spread-error correlations,verification rank histograms, ROC curves, and errorvariance diagrams using these ensemble systems asbenchmarks:

•Old UW MM5 SREF systems (2000-01; 5-member)•“Poor Man’s” ensemble (PEPS; 7-member)•NCEP SREF system (Eta-RSM; 10-member)

•Multimodel applicationsCombine UW MM5 ensemble with:

•PEPS•NCEP SREF system

•Stochastic OR model/field parameter perturbationsIdeas not fully developed; try this next year

Spread-Error Scatter DiagramsALL CASES HIGH & LOW SPREAD CASES

2(,E) = ; =std(ln )2 1-exp(-2)

1- exp(-2)2

(Houtekamer 1993; Whitaker and Loughe 1998)Spread-Error Correlation Theory

Spread-error correlation depends on the time variation of spread

For constant spread (=0) = 0. Spread is the most useful predictor of

skill when it is extreme (large or small)

•Are spread and skill well correlated for other parameters?ie. – wind speed & precipitationUse sqrt to transform data to be normally distributed.

•Do spread-error correlations improve after bias removal?

•How do the correlations compare to the theory?

Developing a Prediction System for Forecast Skill

•What is “high” and “low” spread?need a spread climatology, i.e.- large data set

•What are the synoptic patterns associated with “high” and “low” spread cases?

Use NCEP/NCAR reanalysis data and compositing software

•How do the answers change for the expanded UW MM5 ensemble?

•Is forecast skill correlated with the spread of a temporal ensemble?

Temporal ensemble = lagged forecasts all verifying at the same time

Spread of a temporal ensemble ~ forecast consistency

Temporal Short-range Ensemble

F36 F24 F12F48

with the centroid

BENEFITS:•Yields mesoscale spread

•Less sensitive to one synoptic-scale model’s time variability

•Best forecast estimate of “truth”

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

CENT-CENT-MM5MM5

00 UTCT - 48 h

12 UTCT - 36 h

00 UTCT - 24 h

12 UTCT - 12 h

00 UTCT

F00* Does not have mesoscale features* “spun-up” CENT-MM5 analysisM = 4

verification

Mesoscale VerificationWill verify 2 ways:•At the observation locations (as before)•Using a gridded mesoscale analysis

SIMPLE possibilities for the gridded dataset:

•“adjusted” centroid analysis (run MM5 for < 1 h)Verification has the same scales as the forecastsUseful for creating verification rank histograms

•Bayesian combination of “adjusted” centroid withobservations (e.g.- Fuentes and Raftery 2001)Accounts for scale differences (change of support problem)Can correct for MM5 biases

TRUEVALUES

OBSERVATIONSCENT-MM5“adjusted”

OUTPUT

Bias parameters

Noise

Measurement error

Large-scale structure Small-scale structure (after Fuentes and Raftery 2001)

NEW verification methods/scores?

•gradient-magnitude•pattern recognition•event-based scoring

•Expanded UW SREF probability of precip forecastsCompare to:

•Sample climatology•NGM MOS•NCEP SREF•Old ensemble

•Calibrate using weighted ranks(Eckel and Walters 1998)

•Calibrate using Bayesian Model Averaging (BMA) weights(Hoeting et al. 1999)

•Look at probability forecasts for other parameters

Probability Forecasts

Deterministic-Style Solutions•Centroid

Compare to mean & members using both verification approaches

•Bayesian Model Averaging (BMA)i.e.- Weighted mean

Probability of Warning Criteria at McGuire AFB Bas e d o n 1 5 /0 6 Z MM5 En s e m b le

0

10

20

30

40

50

60

70

80

90

100

Date/T ime

Pro

ba

bili

ty (

%)

T S torm

W inds> 35k t

W inds> 50k t

S now> .5"/hr

Fzg Rain

15/06 12 18 16/00 06 12 18 17/00 06

Innovative Forecast Products/Tools

•Work with NWS-Seattle, Whidbey NAS forecasters(specialized products for warning criteria)

•Work with MURI visualization team at UW APL(ways to visualize uncertainty)

GOAL: VISUALIZING FORECAST UNCERTAINTYWITHOUT NEEDING A TON OF PRODUCTS