TDL WORKSHOP ON MOS INTERPRETIVE GUIDANCE

J. Paul Dallavalle

paul.dallavalle@noaa.gov

301-713-1065, ext. 174

Mary C. Erickson

mary.erickson@noaa.gov

301-763-8151

http://www.nws.noaa.gov/tdl/synop

March 31, 2000

OUTLINE

1. Why Interpretive Guidance?

2. Sample Guidance Products

3. MOS Defined

4. Developmental Strategy

5. Use of Guidance

6. New MOS Products

7. Verification - The Future

REFERENCES

Wilks,D.: Statistical Methods in the Atmospheric Sciences, Chap. 6, p. 159 -210.

Glahn, H.R., and D. Lowry, 1972: The use of model output statistics in objective weather forecasting, JAM, 11, 1203 -1211.

Draper, N.R., and H. Smith, Applied Regression Analysis, Chap. 6, p. 307 - 308.

Carter, G.M., et al., 1989: Statistical forecasts based on the NMC’s NWP System, Wx. & Fcst.,

p. 401 - 412.

Projection (Hour)6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

6

6.5

7

7.5

AVN MOS

AVN DMO

CONUS and Alaska 00z Apr.1 - Sept.30, 1999

Dew Point Mean Absolute Error

2

3

4

5

6

7

8

MA

E (d

eg F

)

24 48 72 96 120 144 168 192Projection (hours)

NEW MOS CLIMO DMO

MRF MOS Max Temp1999 Warm Season - CONUS/AK

WHY INTERPRETIVE GUIDANCE?

• Assist forecasters– “First Guess” for expected local conditions

– “Built-in” model/climo memory for new staff

• Add value to direct NWP model output– Objectively interpret model to remove systematic

biases and quantify uncertainty

– Predict what the model does not

– Produce site-specific forecasts

MODEL OUTPUT STATISTICS

Technique in which observed weather elements (predictands) are related statistically to appropriate predictor variables:

1. NWP Model Forecasts

2. Prior Surface Weather Observations

3. Geoclimatic Information

MODEL OUTPUT STATISTICS

• Current method: Multiple linear regression (forward selection)

• Mathematically simple, yet powerful• Non-linearity modeled by NWP variables, grid

binaries, variable transformations• Other statistical methods possible, e.g., logistic

regression, polynomial regression, neural networks, etc.

• Equations applied to similar NWP model

More Essentials

MODEL OUTPUT STATISTICS

• Advantages– Optimal predictor selection

– Recognition of model predictability

– Removal of some systematic bias

– Reliable probabilities

– Specific element and site forecasts

• Disadvantages– Short samples

– Changing NWP models

DEVELOPMENTAL STRATEGY

• Predictand Definition• Choice of Appropriate Predictors• Sample Size• Terms in Equations• Stratification of Developmental Data -

Seasons• Pooling of Developmental Data - Regions• Overfit of Data

PREDICTAND DEFINITIONS

• Daytime max temperature: 7am-7pm LST

• Nighttime min temperature: 7pm-8am LST

• Probability of precipitation (PoP): occurrence of 0.01 inches of liquid equivalent at the gauge within a specified period

Examples

PREDICTORS

• Describe physical processes associated with occurrence of predictand– PoP: use precipitable water, vertical velocity,

moisture divergence, model precipitation

• Mimic forecasters’ thought process– PoP: (vertical velocity) × (mean RH)

“REAL” REGRESSION EQUATIONS

MOS regressions equations are MULTIVARIATE:

Y = a0 + a1X1 + a2X2 + ... + aNXN, where

the “ai’s” represent coefficients, and

the “Xi’s” represent predictors

The maximum number of terms, N, can be QUITE large, e.g., N=15 for QPF, and N=20 for VIS

FORWARD SELECTION regression determines the predictors and the order of selection.

PTYPE DEVELOPMENT REGIONS

* forecasts will not be produced for stations in the hatched areas

CURRENT NGM/MRF MOS

• NGM MOS predictors valid at 6-h intervals to 48h; NGM MOS valid to 60h

• MRF MOS predictors valid at 12-h intervals to 192h; MRF MOS valid to 192h

• NGM MOS predictors on 190.5 km grid; MRF MOS predictors on 381 km grid

• NGM MOS predictors at 50-mb resolution in vertical; MRF MOS predictors at mandatory levels

A Summary

MOS USER TIPS

Inconsistent forecasts occur, especially with categorical elements - check spatial/temporal continuity

Seemingly inconsistent forecasts occur - do you know the predictand definition?

Guidance may be too conservative during persistent abnormal weather patterns

Systematic errors occur with certain “map types” Guidance does not predict mesoscale events Regionalized guidance may not account for local effects NGM/AVN MOS depend on different samples and different

models with different initial conditions and physics packages

MOS-2000

• Enhanced AVN MOS for Short-Range– First phase: NGM MOS look-alike– Guidance to 72 h– 1000+ stations

• Enhanced MRF MOS for Medium-Range– New forecast elements– Revised element definitions– Same stations as short-range

• Eta MOS for Storm-Related Elements

What Does It Mean to the Customer?

NEW AVN MOS GUIDANCE

• Available for 1000+ Sites in CONUS, AK, HI, PR

• Predictands from Current Observing System

• Predictands to 72 h

• Model Predictors from 95.25 km Grid

• Model Predictors valid every 3 h to 72 h

• 0600 / 1800 UTC AVN MOS Packages

A Different Look

NEW AVN VS. NGM MOS GUIDANCE

• AVN MOS may be more accurate with max/min, dew point, clouds, PoP, winds

• NGM MOS may be more accurate with some temperatures, ceiling height, precip. type, some QPF

• Differences are small

• http://www.nws.noaa.gov/synop/results.htm

What to expect?

NEW MRF MOS GUIDANCE

• Available for 1000+ Sites in CONUS, AK, HI, PR

• New Weather Element Definitions– Wind, clouds, precip. type

• Additional Weather Elements– Temperature, dew pt., QPF, thunderstorms,

snowfall

A Different Look

NEW MRF MOS GUIDANCE

• Forecast projections to 192 h

• Daytime max temp (7am - 7pm LST)

• Nighttime min temp (7pm - 8am LST)

• Probability of .01 inches in 12-h period (PoP)

Unchanged

NEW MRF MOS GUIDANCE

• Temp./dew pt. every 6 h from 18 to 192 h

• PQPF for 12- and 24-h periods– Categories: 0.10, 0.25, 0.50, 1.00, 2.00 in

• Prob. of thunderstorms for 12- and 24-h periods to 192 h

• Prob. of 24-h snowfall amounts– Categories: 1, 2, 4, 6, 8, 10 inches

Additions

NEW MRF MOS GUIDANCE

• Prob. of prevailing sky cover in 12-h period– Categories: clear, partly cloudy, overcast

• Prob. of max. sustained hourly wind speed in 12- h period– Categories: 10 kts, 11-21 kts, 22-34 kts, 35 kts

• Conditional prob. of precip. type in 12-h period– Categories: liquid, snow, freezing

Changes

MOS GUIDANCE

• May 2000 - AVN MOS Message (00Z / 12Z)• May 2000 - New MRF MOS Message • Oct. 2000 - Complete AVN MOS (No Snow)• Oct. 2000 - Complete MRF MOS (No Snow)• Oct. 2000 - Eta MOS Thunderstorm Guidance• Apr. 2001 - NGM MOS Discontinued ?• Apr. 2001 - Old AVN/MRF MOS Discontinued

Implementation Plans

MOS GUIDANCE

• Modify the WFO forecast according to my interpretation of MOS

• Modify MOS according to my analysis of all available information

• Compare MOS packages

• ABSOLUTELY NOT!!

• WITHOUT QUESTION!!

Do I use this tool?

MOS

• Frequent equation updates

• Increased accuracy

• More stations

• Additional predictands

• Dissemination in digital format

• Complete AVN MOS package 4x daily

• New techniques

Future Enhancements