Depicting LAMP probabilities and uncertainty in best category forecast N ew LAMP web page product

Preview:

DESCRIPTION

Depicting LAMP probabilities and uncertainty in best category forecast N ew LAMP web page product Judy E. Ghirardelli Scott Scallion National Weather Service Meteorological Development Laboratory. How we make our “best category” selection in LAMP. Development: - PowerPoint PPT Presentation

Citation preview

Depicting LAMP probabilities and uncertainty in best category forecast

New LAMP web page product

Judy E. GhirardelliScott Scallion

National Weather ServiceMeteorological Development Laboratory

How we make our “best category” selection in LAMP

• Development:– Thresholds are statistically developed by one of two techniques,

either:♦ Targeting unit bias (forecast the event as often as it occurs),

or♦ Maximizing the threat score within a bias range

– Thresholds vary by region, start time, projection, element, category, etc.

• Implementation:– LAMP compares the probability of the category to the threshold

for the category– LAMP starts with the rarest category and goes to the most

common category– The first category whose probability exceeds its threshold is the

chosen category

0

20

40

60

80

Clear Few Scattered Broken Overcast

Forecast Probability

Threshold Value

LAMP Categorical Forecast Selection Process

Pro

bab

ility

(%

)

Does the probability

equal or exceed the threshold?

The probability of “few” exceeds the threshold value for “few” – therefore LAMP categorical forecast is “few”

Category 1 Category 4Category 3Category 2 Category 5

Does the probability

equal or exceed the threshold?

Does the probability

equal or exceed the threshold?

Does the probability equal

or exceed the threshold?

Depicting Probabilistic Information

Take the “simple” example of Probability of Precipitation, which is a yes/no decision. If the probability exceeds the threshold, then the category chosen is “yes.” If not, the category chosen is “no.”

Depicting Probabilistic Information

Purpose: indicate to user the uncertainty associated with the Best Category forecasts given the probabilistic information

Threshold = dashed black line

Probability < thres = green line

Probability ≥ thres = red line

San Francisco – very small chance

of precip

St. Louis – slight chance of precip

Chicago – slight chance yes and slight

chance no precip

St. Cloud – high chance of precip

Depicting Probabilistic Information

Now take the more complicated example of visibility, which has multiple categories to choose.

The probability of vis < 1 mile (solid line) does not exceed the threshold (dashed line). Look to next

rarest category.

LAMP Probabilities and Thresholds for Flight CategoriesThreshold Plot Tab

Look at rarest of these categories

first.

LAMP Probabilities and Thresholds for Flight CategoriesThreshold Plot Tab

Look at next rarest of these

categories.

The probability of vis < 3 miles (solid line) exceeds the threshold (dashed line) only for the last hour;

therefore this condition is indicated for only that hour. Look to next rarest category to determine the

conditions for other hours.

LAMP Probabilities and Thresholds for Flight CategoriesThreshold Plot Tab

Look at next rarest of these

categories.

The probability of vis ≤ 5 miles (solid line) DOES exceed the threshold (dashed line) at times; therefore

this condition is indicated for those times.

Note that vis ≤ 5 is not chosen for the last hour, because a rarer condition (vis < 3 miles) was already

indicated.

LAMP Probabilities and Thresholds for Flight CategoriesThreshold Plot Tab

Show all together.

Looking at these categories, look at the rarest first, then the next rarest, etc. The condition indicated is the rarest probability which exceeds its threshold.

This is indicated by the red probability line.

Depicting Probabilistic Information

In addition to the previous kinds of plots, we are also offering a web page that shows one category’s probability and threshold at a time, and color codes the confidence in choosing that category by indicating how close the probability was to the threshold. One would have more confidence in a chosen category if the probability exceeded the threshold by a large amount, compared to the probability just barely exceeding the threshold.

Red=YesProbability exceeds threshold by more

than 10%

Orange=LikelyProbability exceeds

threshold but NOT by more than 10%

Yellow = ChanceProbability is less than threshold but

within 10%

Cyan = NoProbability is less than threshold by more than 10%

LAMP Probabilities and Thresholds for Flight CategoriesUncertainty Plot Tab – looking at vis ≤ 5 miles

Note that this shows you one condition (e.g., vis ≤ 5 miles). To determine the most likely condition, you

should consider the rarest conditions first.

Depicting Probabilistic Information

• New LAMP web pages coming soon:– Uncertainty plots - station selection page:

♦ http://weather.gov/mdl/gfslamp/uncertform.php♦ Accessible from the main GFS-LAMP Realtime Products

page:■ http://weather.gov/mdl/gfslamp/gfslamp.shtml

– Uncertainty plots - direct access:♦ http://weather.gov/mdl/gfslamp/uncertplots.php

– Threshold plots - direct access:♦ http://weather.gov/mdl/gfslamp/threshplots.php

Depicting Probabilistic Information

• Documentation on the new page:– http://weather.gov/mdl/gfslamp/docs/uncertplots_info.shtml

• For questions or comments, contact:– Judy.Ghirardelli@noaa.gov (LAMP Task Leader)– Scott.Scallion@noaa.gov (LAMP Developer and

Web Master)