Overview of Fire Occurence Accuracy Assessment Wilfrid Schroeder PROARCO – IBAMA University of...

Overview of Fire Occurence Accuracy Assessment

Wilfrid SchroederPROARCO – IBAMA

University of Maryland – Dept of Geography

Ground-based Accuracy Assessment for Fire and Deforestation Events Workshop

Grand Bittar Hotel - Brasília 26 July 2004

Main Goal

To address satellite derived fire product characteristics giving emphasis to MODIS data

To point out major elements affecting the resulting fire numbers

To suggest means of correcting/accounting for such influences

How can we improve data quality - validation


Are we representing the correct surface condition? (i) Qualitatively

Do hot spot numbers reflect the expected fire dynamics on a regional level? Temporal and spatial distribution

Do different satellite products reflect the same relative numbers?


Biome Characterization


Ecoregion Terra2001 Terra2002 Terra2003 Aqua2003 AVHRR2001 AVHRR2002 AVHRR2003Floresta Amazônica 19074 29452 24337 88027 50835 84338 61892Floresta Estacional 23887 41109 39835 74713 37736 61782 68240Cerrado 30205 39878 29378 50911 28747 47958 38766Caatinga 4143 4027 4921 23237 17377 22652 30028Campinarana 50 27 203 336 125 44 302Mata Atlântica 720 773 1168 4137 908 1942 2998Mata dos Pinheiros 251 284 717 9302 873 1392 3478Complexo do Pantanal 5601 7696 1928 3260 6723 11028 2415Vegetação Costeira 1111 900 813 2335 1563 2196 1970Campos 615 541 403 796 513 750 646Campos Campanha Gaúcha 67 55 72 274 70 152 52Campos de Roraima 227 224 89 173 95 126 109Ilha 0 0 0 0 0 0 0

Ecoregion Terra2001 Terra2002 Terra2003 Aqua2003 AVHRR2001 AVHRR2002 AVHRR2003Floresta Amazônica 0.0054 0.0083 0.0069 0.0248 0.0143 0.0238 0.0174Floresta Estacional 0.0145 0.0250 0.0242 0.0455 0.0230 0.0376 0.0415Cerrado 0.0192 0.0254 0.0187 0.0324 0.0183 0.0306 0.0247Caatinga 0.0056 0.0055 0.0067 0.0316 0.0237 0.0308 0.0409Campinarana 0.0002 0.0001 0.0007 0.0012 0.0004 0.0002 0.0011Mata Atlântica 0.0029 0.0031 0.0047 0.0168 0.0037 0.0079 0.0122Mata dos Pinheiros 0.0014 0.0015 0.0039 0.0505 0.0047 0.0076 0.0189Complexo do Pantanal 0.0348 0.0478 0.0120 0.0203 0.0418 0.0686 0.0150Vegetação Costeira 0.0077 0.0063 0.0057 0.0163 0.0109 0.0153 0.0137Campos 0.0056 0.0049 0.0037 0.0072 0.0046 0.0068 0.0059Campos Campanha Gaúcha 0.0011 0.0009 0.0011 0.0044 0.0011 0.0024 0.0008Campos de Roraima 0.0184 0.0181 0.0072 0.0140 0.0077 0.0102 0.0088Ilha 0 0 0 0 0 0 0

Absolute Numbers

Hot

Spot

Density


State Terra2001 Terra2002 Terra2003 Aqua2003 AVHRR2001 AVHRR2002 AVHRR2003Amazonas 1296 1741 2269 7117 1298 1858 2297Pará 9844 14747 8653 34471 28588 48163 28796Mato Grosso 26314 48515 41139 59436 33046 58457 55343Maranhão 6277 6370 5628 24733 18699 24134 24828Tocantins 7833 9584 6168 10573 9106 13625 9818Rondônia 3533 6164 7259 20944 5059 10463 9130Roraima 1515 995 2111 2690 2416 2018 4778Acre 664 616 1138 6845 828 1238 1229Amapá 759 551 522 1755 1301 1153 906

State Terra2001 Terra2002 Terra2003 Aqua2003 AVHRR2001 AVHRR2002 AVHRR2003Amazonas 0.0008 0.0011 0.0014 0.0045 0.0008 0.0012 0.0015Pará 0.0079 0.0118 0.0069 0.0276 0.0229 0.0386 0.0231Matro Grosso 0.0291 0.0537 0.0455 0.0658 0.0366 0.0647 0.0613Maranhão 0.0189 0.0192 0.0170 0.0745 0.0563 0.0727 0.0748Tocantins 0.0282 0.0345 0.0222 0.0381 0.0328 0.0491 0.0354Rondônia 0.0149 0.0259 0.0306 0.0882 0.0213 0.0440 0.0384Roraima 0.0068 0.0044 0.0094 0.0120 0.0108 0.0090 0.0213Acre 0.0044 0.0040 0.0075 0.0449 0.0054 0.0081 0.0081Amapá 0.0053 0.0039 0.0037 0.0123 0.0091 0.0081 0.0063

Absolute Numbers

Hot

Spot

Density

State Characterization

Are we representing the correct surface condition? (ii) Quantitatively

• Do absolute numbers reflect the ground truth?

• What are the major elements affecting statistics?

• How do such elements impact different satellite data?


Driving elements

• Satellite coverage• Viewing geometry (pixel size & view angle)• Fire size/temperature/duration• Clouds• Biome type• Diurnal Cycle


Satellite Coverage

MODIS Global Coverage15 June 2004

NASA-GSFC


Viewing geometry(pixel size & view angle)


0

2

4

6

8

10

12

14

16

18

0 5 10 15 20 25 30 35 40 45 50 55

Scan Angle (º )

Siz

e (k

m)

Are

a (k

m2)

Pixel Area

Cross Track

Along Track

AVHRR


0

2

4

6

8

10

12

0 5 10 15 20 25 30 35 40 45 50 55

Scan Angle (º )

Siz

e (k

m)

Are

a (k

m2)

Pixel Area

Cross Track

Along Track

MODIS


0

20

40

60

80

100

120

0 10 20 30 40 50 60 70

Scan Angle (o)

Hot

Spo

ts

NOAA-12

NOAA-14



0

50

100

150

200

250

0 10 20 30 40 50 60 70

Scan Angle (o)

Hot

Spo

ts

NOAA-12

NOAA-14




0

10000

20000

30000

40000

50000

60000

0 1 2 3 4 5 6

Pixel Size

Hot

Spo

ts

Terra_2001

Terra_2002

Terra_2003

Aqua_2003



0

50

100

150

200

250

1-jul

8-jul

15-jul

22-jul

29-jul

5-ago

12-ago

19-ago

26-ago

Hot

Spo

ts


AVHRR

0

50

100

150

200

250

300

1-jul

8-jul

15-jul

22-jul

29-jul

5-ago

12-ago

19-ago

26-ago

Hot

Spo

ts



MODIS

Fire size/temperature/duration


1.05

0.520.29 0.18 0.11 0.08 0.02

7.40

2.48

0

1

2

3

4

5

6

7

8

400K 450K 500K 550K 600K 650K 700K 750K 1000K

Flame Temperature

Hec

tare

s

Background: 300KSaturation (AVHRR): 322K

Pixel size: 1.1km

Fire size/temperature/duration

0

5

10

15

20

25

100 300 500 700 900 1100 1300 1500 1700 1900 2100 Mais

Meters

Fre

quency

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%


Clouds

MCT/INPE/CPTEC

Satélite Noaa-12 Data: 2004/06/22

Estado/State : Mato Grosso

Total Imageado : 99 % da Área

Total Imaged : 99 % of the area

Da Área Imageada : 0 % possíveis Nuvens

Of the Imaged Area : 0 % possible Clouds

Focos de Calor/Hot Pixels : 852


One can calculate hot spot density per area

imaged

Biome Type

• Different biome types will involve different biomass characteristics which influence fire intensity (thereby detectability)

• Climatology will also affect soil&vegetation moisture and atmospheric conditions

• Land use management will differ over biomes (timing and type of fires – conversion/maintenance)


Diurnal Cycle

• Satellite overpass time represent distinct fire situations

• Difference within image (eastern/western edges are imaged at different local times ~1.3 hour)

• Mid-afternoon overpass time taking advantage of usual higher fire frequency and increased local temperature (favoring detectability)


Validation Efforts

Satellite inter-comparison via GIS analysis

Airborne imaging (Vis/IR bands)

Ground sampling


Desktop Validation

Provides valuable information to assess product quality image navigation detectability [to some extent]

Pros Fire data is available online (PROARCO/INPE,

UMD/NASA, UW-ABBA) Almost costless (software license)

Cons May lead to wrong conclusions

**ASTER imagery can provide coincident high resol data to validate MODIS/Terra


Airborne Imaging

Provides high resol imagery Detectability performance (fire size) Area estimates

Pros Fastest way to collect large quantities of data for

validating different satellite data Good for both prescribed and opportunistic fires Can also be used for burn scar & deforestation product

validation Cons

High cost operation


Ground Sampling

Provides in situ measurements of fire characteristics Temperature, duration, size [to some extent] Detectability performance

Pros Best way to understand “what is in a pixel” Can serve to collect complementary data (fuel load,

combustion efficiency, etc.) Cons

Time consuming process Requires area selection (committing land owners)


Homework

Equipment Required

What can we do with that?

How can we accomplish

that?

Desktop PC + software Routine quality assessment (persisting errors); image navigation

Internet access:BdqueimadasRapidResponseABBA

Airborne Imaging system +airplane

Fire sampling: size/~intensity~~timing~/location

Large research projects (requires group coordination)

Ground GPS; digital camera; temperature sensor

Fire sampling: timing/durationlocation/~size~intensity/land history

Individuals doing field research/local fire community/~landowners


Fire product access:INPE (AVHRR; MODIS; GOES; DMSP – South America):http://www.cptec.inpe.br/queimadashttp://www.dpi.inpe.br/proarco/bdqueimadasNASA/UMD (MODIS – World Wide):http://rapidfire.sci.gsfc.nasa.govhttp://maps.geog.umd.eduWisconsin (GOES – North -> South America):http://cimss.ssec.wisc.edu/goes/burn/wfabba.html

E-Mail listStart new contacts from this workshop and strengthen old ones

Wire-up!!


Overview of Fire Occurence Accuracy Assessment Wilfrid Schroeder PROARCO – IBAMA University of...

Documents

Transcript of Overview of Fire Occurence Accuracy Assessment Wilfrid Schroeder PROARCO – IBAMA University of...