SIMULATION AND ANALYSIS OF GOES-R GEOSTATIONARY LIGHTNING MAPPER (GLM) DETECTION ALGORITHM...

SIMULATION AND ANALYSIS OF GOES-R GEOSTATIONARY LIGHTNING MAPPER (GLM) DETECTION ALGORITHM

PERFORMANCE

Loren Sadewa Clark, Tom Dixon, Pete Armstrong, Ruth Cholvibul, Jared Clarke, Peter Silverglate and Don Chu

Agenda

1. NomenclatureLightning

Requirements

2. InputsLightning statistics

Design parameters

3. GLM ProcessingOnboard processing

Ground processing

4. SimulationEvent Generation

Detecting Lightning

Evaluating Results

AMS Annual Meeting 2014 - Atlanta, GA 2/5/14

http://www.goes-r.gov/spacesegment/glm.html

Nomenclature

Lightning1. Flashes are made up of pulses.

2. Pulses generate GLM events.

3. Events are exceedances of threshold in a pixel

4. Groups are events generated by the same pulse.

5. Radiance – energy (uJ) per aperture area (m2) per pixel steradian (sr).

Requirements1. Detection probability – fraction of flashes detected

2. False alarm probability – fraction of detected flashes that are not lightning


InputsLightning statistics (Grandell and Buechler)

1. Time: Flashes per second, mean and max time between pulses

2. Strength: event energy, pulse energy, max pulse energy 3. Area: pulse area, flash area

Design parameters (Edgington and Tillier)

1. Signal: aperture area, quantum efficiency, optical throughput

2. Noise: electronic and background3. Spacecraft: radiation shielding and attitude stability

(jitter)


Simulation – Event GenerationFor each frame, …

1. Shift background according to simulated attitude history.

2. Superimpose radiation and lightning.

3. Subtract tracked background and apply thresholds.

4. Update tracked background and go on to next frame.

Outputs1. Frame number

2. Pixel row

3. Pixel column,

4. Event strength,

5. Flash number or false event type flag

6. Background 5 most significant bits


Generated Events

AMS Annual Meeting 2014 - Atlanta, GA

2/5/14

Simulation – Lightning DetectionFor each frame, …

Calculate probability that the event was noise. If pixel was already active, …

Compute probability that event and parent were false.If that joint probability is less than limit, event is detected.

Assign ‘flash number’ of parent event to this event.Write to output file.

EndElse

Start new ‘flash number’ for this eventEnd

Update probability matrix.

Start event clock for the pixel.


Active Pixel Test


2/5/14

Simulation – Result EvaluationFor each frame, …

1. Match detected to simulated events.

2. Assign true flash number or flag to the detected event.

After all detected events are tagged, …

1. Compute the fraction of flashes detected to give detection efficiency.

2. Count number of ‘detected flashes’ that don’t include any real lighting events.

3. Divide this number by the number of true flashes to give false alarm proability.


Input and Output Events

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

1000

2000

3000

4000

5000

6000

7000

454 t

rue;

16242 f

als

e

pd = 0.76; far = 0/s

false

true

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

100

200

300

400

500

600

700

800

202 t

rue;

0 f

als

e

time (sec)

Evaluate.m


2/5/14

Acknowledgements This work was supported under NASA contracts NNG08HZ00C and

NNG14CR58C with Tom Dixon as the GLM instrument manager.

The authors also wish to thank:

Hugh Christian of Ryco who has provided instruction and has freely shared his experience and expertise over the years.

Steve Goodman, Doug Mach and Dennis Buechler of NOAA for their help generating lightning data.

Samantha Edgington and Clem Tillier of Lockheed Martin for their GLM performance analyses upon which much of this work was based.

James Bremer of Research Support Instruments who contributed heavily to the early development of our event and algorithm modeling.


2/5/14

Backup Slides

2/5/14AMS Annual Meeting 2014 - Atlanta, GA

Generated Events - Radiation


0 500 1000 1500 2000 2500 3000 3500 4000 45000

20

40

60

80

100

120

all events (counts)

mean:

457.7

std

ev:

577.8

22 hits; 181 events

0 500 1000 1500 2000 2500 3000 3500 4000 45000

5

10

15

brightest event (counts)

mean:

622.1

m

edia

n:

317

0 5 10 15 20 25 30 350

1

2

3

4

5

events/hit

mean:

8.2

m

edia

n:

6

0 20 40 60 80 100 120 140 160 1800

1

2

3

4

5

6

hit intervals (millisec)

mean:

43.9

m

edia

n:

28

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

1000

2000

3000

4000

5000

counts

time (sec)

radiation plots.m

Generated Events - Lightning


0 10 20 30 40 50 600

10

20

30

40

50

all pulses (uJ/m2sr)

mean:

23.3

std

ev:

10.5

19 flashes; 184 pulses

15 20 25 30 35 40 45 50 55 600

1

2

3

4

brightest pulse (uJ/m2sr)

mean:

37.5

m

edia

n:

39

0 10 20 30 40 500

2

4

6

8

pulses/flash

mean:

9.6

m

edia

n:

4

0 50 100 1500

10

20

30

40

50

60

pulse intervals (millisec)

mean:

22.8

m

edia

n:

14

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

puls

es (

uJ/m2 s

r)

time (sec)

lightning plots.m

SIMULATION AND ANALYSIS OF GOES-R GEOSTATIONARY LIGHTNING MAPPER (GLM) DETECTION ALGORITHM...

Documents

Transcript of SIMULATION AND ANALYSIS OF GOES-R GEOSTATIONARY LIGHTNING MAPPER (GLM) DETECTION ALGORITHM...