The “User Metric” Concept cfab.eas.gatech/usermetric.html

Post on 31-Dec-2015

14 views 1 download

Tags:

description

The “User Metric” Concept http://cfab.eas.gatech.edu/usermetric.html. Contents. 1. Introduction. 2. Concept. 3. Problem. 4. Providers. Concept. User metric concept Combining probabilistic forecasts with knowledge from the user community to provide a metric for optimal choice. Problem. - PowerPoint PPT Presentation

Transcript of The “User Metric” Concept cfab.eas.gatech/usermetric.html

The “User Metric” Concept

http://cfab.eas.gatech.edu/usermetric.html

Contents

1. Introduction 1. Introduction

2. Concept 2. Concept

3. Problem 3. Problem

4. Providers 4. Providers

Concept

User metric concept

Combining probabilistic forecasts with knowledge from the user community to provide a metric for optimal choice.

Problem

Issues to address:

A user community is faced with making an absolute decision (yes or no).

Forecasts are in the form of probabilities. How can we use probabilistic forecasts to

provide the best information to the user community?

Can only occur with incorporation of user information.

Providers

1. Forecaster provider 1. Forecaster provider

2. User community provider 2. User community provider

Forecaster provides:

Probabilities of particular event occurring at a particular time at a particular intensity

ECMWF: Sept. 9-13, 2002ECMWF: Sept. 9-13, 2002

Discharge Probabilities as Computed in April, 1998

Discharge Probabilities as Computed in April, 1998

Do you have anything to say here?

ECMWF: Sept. 9-13, 2002ECMWF: Sept. 9-13, 2002

Discharge Probabilities as Computed in April, 1998

Discharge Probabilities as Computed in April, 1998

User community provides:

An assessment of the cost if a certain phenomena were to occur with a particular intensity.

This may be quantified to provide the cost of occurrence of a phenomena.

Community input provides an assessment of the cost of a strategy if an event were to occur.

100-0 75-2550-50 25-75

10-900-100

High Flood

Med. Flood

Light Flood

No Flood

0

10

20

30

40

50

60

70

80

90

100

High Flood

Med. Flood

Light Flood

No Flood

Do you have anything to say here?

100-0 75-2550-50 25-75

10-900-100

High Flood

Med. Flood

Light Flood

No Flood

0

10

20

30

40

50

60

70

80

90

100

High Flood

Med. Flood

Light Flood

No Flood

The user metric . . .

Takes the probabilistic forecast and the statistics provided by the user community to produce….

A statistical assessment of the aggregate risk of taking a particular action so that….

The user community can choose an optimal strategy given all of the available information.

User Metric Flowchart

ForecasterForecasterUser CommunityUser Community

User Metric User Metric

Flow of input

User Metric Flowchart

Scenarios

1. Simple example 1. Simple example

2. Typical situation 2. Typical situation

3. Scenario A: 3. Scenario A:

4. Scenario B: 4. Scenario B:

5. Scenario C: 5. Scenario C:

Scenarios

1. Simple example . . . 1. Simple example . . .

What is the best strategy for harvesting crops for a particular forecast?Are there some strategies better than others?We assume that the farming community may harvest all crops early with reduced yield, harvest a certain % early and the rest later, or all later - noting that the closer to maturation, the greater the crop yield.What is the risk in making a particular decision?

What is the best strategy for harvesting crops for a particular forecast?Are there some strategies better than others?We assume that the farming community may harvest all crops early with reduced yield, harvest a certain % early and the rest later, or all later - noting that the closer to maturation, the greater the crop yield.What is the risk in making a particular decision?

Scenarios

2. Typical situation may be . . . 2. Typical situation may be . . .

Over successive periods forecasts of rainfall indicate:

Initially decreases (A) Increases significantly (B) Decreases slightly (C)

Over successive periods forecasts of rainfall indicate:

Initially decreases (A) Increases significantly (B) Decreases slightly (C)

Scenarios A, B and C

AB C

Scenarios

Scenario A: Scenario A:

Rainfall probabilities suggest that there will be significant decreases in rainfall amounts

Rainfall probabilities suggest that there will be significant decreases in rainfall amounts

A

Scenario A

65%

25%

8% 2%

No Rain

Light rain

Mod. Rain

Heavy Rain

Scenario A

100-0 75-25 50-50 25-75 10-90 0-100

Heavy Rain

Light Rain0

10

20

30

40

50

60

70

80

90

100

Heavy Rain

Mod. Rain

Light Rain

No Rain

Scenario A

Yield Based on Aggregate Risk

74

76

78

80

82

84

86

88

100-0 75-25 50-50 25-75 10-90 0-100

Action (now-later)

Perc

enta

ge C

rop Y

ield

Scenarios

Scenario B: Scenario B:

Rainfall forecast probabilities suggest that rainfall will increase dramatically

Rainfall forecast probabilities suggest that rainfall will increase dramatically

B

Scenario B

5%

15%

50%

30%

No Rain

Light rain

Mod. Rain

Heavy Rain

Scenario B

Yield Based on Aggregate Risk

0102030405060708090

100-0 75-25 50-50 25-75 10-90 0-100

Action (now-later)

Perc

enta

ge C

rop Y

ield

Scenarios

Scenario C: Scenario C:

Forecast probabilities indicate that rainfall will decrease moderately

Forecast probabilities indicate that rainfall will decrease moderately

C

Scenario C

15%

55%

25%

5%

No Rain

Light rain

Mod. Rain

Heavy Rain

Scenario C

100-0 75-25 50-50 25-75 10-90 0-100

Heavy Rain

Light Rain0

10

20

30

40

50

60

70

80

90

100

Heavy Rain

Mod. Rain

Light Rain

No Rain

Scenario C

Yield Based on Aggregate Risk

0102030405060708090

100-0 75-25 50-50 25-75 10-90 0-100

Action (now-later)

Perc

enta

ge C

rop Y

ield

Click to edit title

Title1 Title2 Title3 Title4

A

B

C

D

Click to edit title

Text 1Text 1 Text 2Text 2 Text 3Text 3

Click to edit title

TextText

TextText

TextText

TextTextTextText

Add Title

Click to edit title

Text 1 Text 2 Text 3 Text 4

Chart Documents

12

34

5

8830

12

1030

30

2030

90

6523

15

2533

65

0

20

40

60

80

100

120

140

A B C

100-0 75-2550-50 25-75

10-900-100

High Flood

Med. Flood

Light Flood

No Flood

0

10

20

30

40

50

60

70

80

90

100

High Flood

Med. Flood

Light Flood

No Flood

Dissemination Process

Forecaster

Interpretation

Dissemination

End User

ModellerData

Your Company Slogan