The University of Mississippi Geoinformatics Center NASA MRC RPC – 11 July 2007 Rapid Prototyping of NASA Next Generation Sensors for the SERVIR System of Fire Detection in Mesoamerica Joel Kuszmaul Henrique Momm Greg Easson The University of Mississippi Collaborators: Dan Irwin, NASA-MSFC Tim Gubbels, SSAI- Goddard Bob Ryan, SSAI-Stennis
23
Embed
Joel Kuszmaul Henrique Momm Greg Easson The University of Mississippi
Rapid Prototyping of NASA Next Generation Sensors for the SERVIR System of Fire Detection in Mesoamerica. Joel Kuszmaul Henrique Momm Greg Easson The University of Mississippi. Collaborators: Dan Irwin, NASA-MSFC Tim Gubbels, SSAI-Goddard Bob Ryan, SSAI-Stennis. Objectives. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Rapid Prototyping of NASA Next Generation Sensors
for the SERVIR System of Fire Detection in Mesoamerica
Joel KuszmaulHenrique Momm
Greg Easson
The University of Mississippi
Collaborators: Dan Irwin, NASA-MSFC Tim Gubbels, SSAI-Goddard Bob Ryan, SSAI-Stennis
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Objectives
• We do not seek to validate or evaluate the MODIS active fire detection algorithm
• This has been done by other scientists
• We seek to compare results from MODIS to results from VIIRS with the goal of identifying issues of active fire detection
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
MODIS Active Fire Product (SERVIR)
Fires in Mesoamerica
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
PIXEL VALUE
MEANING
0 not processed –missing data (black)
2 not processed – other reason (black)
3 water (blue)
4 Cloud (purple)
5 No fire (gray)
6 Unknown (black)
7 Low-confidence fire (orange)
8 Nominal confidence fire (yellow)
9 High confidence fire (red)
MODIS Active Fire Product (MOD14) Production Code, Version 4.3.2
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
The Kappa Statistic• Useful for assessing agreement
between two sets of classification• Corrects for chance agreement• An improvement on the proportion of
correct classification (simplest measure of agreement)
• Calculated in the general case as:
2..21..1
2..21..122111
ˆpppp
pppppp
= 0 chance agreement < 0 worse than chance > 0 better than chance
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
MODIS FIRE ALGORITHM
Channel Number
Central Wavelengh
(µm)Purpose
1 0.65 Sun glint and coastal false alarm rejection
2 0.86Bright surface, sun glint, and coastal false alarm rejection; cloud masking
7 2.10 Sun glint and coastal false alarm rejection
21 4.00 High-range channel for active fire detection
22 4.00 Low-range channel for active fire detection
31 11.00 Active fire detection, cloud masking
32 12.00 Cloud masking
Giglio et al (2003)
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparison between MODIS and VIIRS spectral and spatial resolution
Spectral Bands Spectral Range (um) Nadir HDR (m) Spectral Bands Spectral Range (um) Nadir HDR (m)M1 0.402-0422 750 8 0.405-0.420 1000M2 0.436-0.454 750 9 0.438-0.448 1000
The error matrix result comparing the MODIS andsimulated VIIRS fire products for March 20, 2003,using the Terra sensor data and the extended error modelfor the simulated VIIRS data.
Overall Accuracy: 0.999578148Kappa: 0.6989
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparing MODIS- and VIIRS-based Detection Tools (continued)
0.7371
0.7641
0.70110.6900
0.5434
0.6988
0.8791
0.7222
0.6686
0.7187
0.63570.6219
0.4735
0.6331
0.7945
0.6758
0.7028
0.7414
0.66840.6560
0.5084
0.6660
0.8368
0.6990
0.4500
0.5000
0.5500
0.6000
0.6500
0.7000
0.7500
0.8000
0.8500
0.9000
Results from the overall kappa calculations for the case of the point source error model
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparing MODIS- and VIIRS-based Detection Tools (continued)
Overall results comparison for the ability to detect fires using the four different definitions of fires
DAY TERRA AQUA DAY TERRA AQUA
March 20, 2003 0.7028 0.7414 March 20, 2003 0.7351 0.7869April 21, 2003 0.6684 0.6560 April 21, 2003 0.6920 0.6810April 28, 2003 0.5085 0.6660 April 28, 2003 0.5215 0.7014April 30, 2003 0.7335 0.6990 April 30, 2003 0.7503 0.7230
DAY TERRA AQUA DAY TERRA AQUAMarch 20, 2003 0.7369 0.7893 March 20, 2003 0.9134 0.9145April 21, 2003 0.6921 0.6811 April 21, 2003 0.8677 0.8075April 28, 2003 0.5220 0.7022 April 28, 2003 0.6922 0.8291April 30, 2003 0.7524 0.7238 April 30, 2003 0.9182 0.8326
Overall Low-confidence and higher
Nominal-confidence and higher High-confidence
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Low and Nominal Confidence Fires
• Nominal confidence fires found only 20% as often using VIIRS
• Low confidence fires not found at all
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparing Detection Tools Using Validation Data Sets: Results from the Aster Imagery
Fire location and the 25 nearest MODIS pixelscollected to investigate agreement with field data
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
MODIS - TERRA
0.0000
0.0500
0.1000
0.1500
0.2000
0.2500
0.3000
0.3500
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
Kap
pa
Sta
tist
ic
April, 21 2003 April, 30 2003
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
Pro
bab
ilit
y (O
mis
sion
Err
or)
MODIS - TERRA
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0.0030
0.0035
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
Pro
bab
ilit
y (C
omm
issi
on E
rror
)
VIIRS - TERRA
0.0000
0.0200
0.0400
0.0600
0.0800
0.1000
0.1200
0.1400
0.1600
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
Kap
pa S
tati
stic
April, 21 2003 April, 30 2003
VIIRS - TERRA
0.0000
0.1000
0.2000
0.3000
0.4000
0.5000
0.6000
0.7000
0.8000
0.9000
1.0000
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
VIIRS - TERRA
0.0000
0.0002
0.0004
0.0006
0.0008
0.0010
0.0012
0.0014
0 10 20 30 40 50 60 70 80 90 100
ASTER Threshold
Pro
babi
lity
(Com
mis
sion
Err
or)
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparing Detection Tools Using Validation Data Sets: Results from the Landsat Imagery
Examples of Landsat false color composite images showing active fires in Guatemala
Two independent image analysts
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Comparing Detection Tools Using Validation Data Sets: Results from the Landsat Imagery (continued)
Comparison of both large and small fires identified inLandsat-7 imagery and fires detected by the MODIS- and VIIRS-based DST
Small Fires OverallSensor No. Found Fires No. Found Fires No. Found Fires No. Found Fires No. Found Fires Accuracy
March 20, 2003 April 21, 2003 April 28, 2003 April 30, 2003 Overall
March 20, 2003 April 21, 2003 April 28, 2003 April 30, 2003 Overall
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Summary and Findings
• The highest values were obtained when the MODIS- and VIIRS-based assessments of high confidence fires
• The VIIRS-based fire detection system finds few nominal-confidence fires and no low-confidence fires
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007
Summary and Findings
• Previous researchers had identified the potential difficulty of the proposed VIIRS thermal band (3.95 m) in finding small and low intensity fires. Our results confirm their expectations.
• We recommend a change in the sensor-algorithm combination from what is currently planned.
The University of Mississippi Geoinformatics CenterNASA MRC RPC – 11 July 2007