methods of detecting burnt area and estimating emissions...burned area (km2), number of scars and %...

Methods of detecting burnt area and estimating emissions Dr. Kevin Tansey ([email protected])

Why fire is important

•  Emitter of GHG and aerosols into the atmosphere –  Stohl, A. et al., 2006, … record high air

pollution levels in the European Arctic due to agricultural fires …, ACP, 7, 511-534, 2007

–  Page S.E. et al., 2002, Nature, 420, 61-65 •  Consequence of land cover/use change

–  Amazonia (trees and grasslands) –  Indonesian peatlands

•  Climate change impacts and feedbacks –  More fire-affected regions?

Current EO state of the art

•  Burned area –  MODIS, L3JRC, GlobCarbon + regional data –  No standards on validation/intercomparisons

•  Flaming fire detection –  MODIS, WFA, EUMETSAT, TRMM –  Limited detection capability

•  FRP –  MODIS FRP, SEVERI

•  Emissions databases –  GFED (mainly makes use of MODIS data)

•  Detecting fire is easy – disturbance less so •  Accuracy is certainly dependent on resolution

Validation activities

•  Effort being placed on validation of global product

• More an evaluation of existing products –  Geographically limited –  Reliance on secondary ground data –  Normally based on Landsat pairs (USGS) –  High-res = in situ in most cases

•  The community agrees on the need for validation protocols

•  Fuel loadings and burning efficiency –  Fuel type data needed –  Fuel load data collected

•  Burn severity –  Relationships between LAI and dNBR (Boer M. et al.,

2008, RSE, 112, 4358-4369)

•  Regional calculations –  Bottom-up and top down approaches –  Carbon flux estimates using daily climate data input

in SPITFIRE module in LPJ-GUESS model (Lehsten et al. 2008, BGD, 5, 3091-3122)

–  SAFARI 2000, GFED

Fire disturbance to emissions

•  Burned area –  MODIS, L3JRC,

GlobCarbon + regional data

Multi-year burned areas detected from 2000-2007 from SPOT-VGT satellite Tansey, K., et al. GRL, 35, L01401 doi:10.1029/2007GL031567

Burned area (km2), number of scars and % of each vegetation type per country burned (comma delimited).

Country Needleleaf forest Broadleaf forest Woodlands &

shrublands Grasslands &

croplands Angola - 2706,1290,6.2 271789,21077,25.5 22006,8459,20.6

Australia 345,249,1.7 4101,1219,1.7 525357,20303,8.3 28982,9083,3.3

Italy 84,54,0.2 48,21,0.4 421,318,0.4 1968,707,1.4

USA 5867,1344,0.5 146,115,0.0 17302,6739,0.4 11648,5496,0.4

L3JRC Reporting Example

Tansey, K., et al. (2004), J. Geophys. Res., doi:10.1029/2003JD003598.

Validation activities Validation tools and standards are being

planned under EC FP7 Geoland2 & NASA

CEOS WGVC

•  Biomass loss –  Fuel type and fuel load data are critical –  Burn severity can be directly derived from FRP

•  Emissions databases –  Global Fire Emissions Database (GFED)

Fire disturbance to emissions

Lehsten V. & Tansey, K. et al. Biogeosci. Disc., 5, 3091-3122

Intercomparison experiments

MODIS MCD45A1 burned area product

MONTHLY BURNED AREA MAPS

BA Month123456789101112

The MODIS Burned Area Product

Slides courtesy of Luigi Boschetti & David Roy

Global MODIS Burned Area Product •  Funded as part of NASA MODIS Fire Science Team

(Justice et al.) to complement the well established (Collection 1,3,5) MODIS 1km active fire product"

•  Global applications"–  Green house gas & aerosol emissions estimation "–  Applied users (e.g., natural resource management)"–  LCLUC research (e.g., Fire – Climate – People)"

"

•  Collection 5 processing now completed for " MODIS data sensed 2000+. New version (5.1)

scheduled for october-2009"

Algorithm

•  Rolling bidirectional reflectance distribution function (BRDF) based expectation change detection

•  Semi-Physically based; less dependent upon imprecise but noise tolerant classification techniques; very few thresholds

•  Automated, without training data or human intervention

•  Applied independently per pixel to daily gridded MODIS 500m land surface reflectance time series => globally map 500m location and approximate day of burning

The challenge: change detection of Burned Areas

BRDF Effects

gaps

Slides courtesy of L. Boschetti and D. Roy Algorithm Background

Bidirectional reflectance effect on a grass lawn observed under different angles (source University of Zurich, Department of Geography)

What is bidirectional reflectance?

bidirectional reflectance effect is evident when an object or image viewed or illuminated from different angles

backscattering forward scattering (sun behind observer) (sun opposite observer)

http://geography.bu.edu/brdf/brdfexpl.html Photographs by Don Deering

BRDF Effects

gaps

Day of burning

Persistence of the signal

The challenge: change detection of Burned Areas

Slides courtesy of L. Boschetti and D. Roy Algorithm Background

Conceptual Scheme (one pixel, time series)

Algorithm Background

!me

ρ

observed

Slides courtesy of L. Boschetti and D. Roy

Conceptual Scheme


!me

ρ

observed

t-‐1


Conceptual Scheme


!me

ρ

observed

t-‐1

BRDF Inversion window



!me

ρ

observed

t-‐1

ρ (t|t-‐1)

>

predicted BRDF Inversion window

Conceptual Scheme



!me

ρ

observed

t-‐1

ρ (t|t-‐1)

>

ρ (t|t-‐1)


Conceptual Scheme



!me

ρ

observed

t

ρ (t+1|t)

>

ρ (t+1|t)


Conceptual Scheme


Animation: 5 Months of burning, Okavango Delta, Botswana, 2002. Produced using multitemporal rolling BRDF-based change detection approach, Roy et al. 2005

•  Burned Area algorithm run globally for first Dme in MODIS C5 -‐ purposefully running to map burned areas conservaDvely

500m burned areas 5 months 2002 Zambia/Zimbabwe 650*500km

1km active fires 5 months 2002 Zambia/Zimbabwe 650*500km

Australia 500m burned areas 1 month 2002


Australia 1km active fires 1 month 2002


Brazil, Southern Para, 500m burned areas 1 month 2002

Brazil, Southern Para, 1km active fires 1 month 2002

Example refinement C5 monthly burned area (MCD45) product Greece August 2007

BoscheJ, Roy, Barbosa, et al, 2008

Active Fire Information Slides from Martin Wooster, King’s College London (KCL)

Terrestrial Fire Remote Sensing Products •  Burned Area Maps"

–  Identifies the location of burned ground, after fire event."

•  Active Fire Detections (“Hotspots”)"–  Identifies the location of fires that are burning at the

time of the satellite observation"

•  Fire Radiative Power (FRP)"–  A measurement of the rate of thermal radiative energy

release at the detected active fire pixels."

Terrestrial Fire Remote Sensing Products •  Burned Area Maps"

–  Identifies the location of burned ground, after fire event."

•  Active Fire Detections (“Hotspots”)"–  Identifies the location of fires that are burning at the

time of the satellite observation."

•  Fire Radiative Power (FRP)"–  A measurement of the rate of thermal radiative energy

release at the detected active fire pixels."

Observing Satellites "•  Geostationary

–  Near continuous view of Earth, Meteosat provides data of Africa every 15 minutes."

–  Lower spatial resolution (~ 3 to 5 km)"

•  Low Earth Orbit (~ Near Polar) –  Temporal resolution few hrs to few days"

–  Moderate to High spatial resolution""(usually around ~ 1 km)"

Active Fire Detections (“hotspots”)

The location of fires that are burning at the time of the satellite observation"

Active Fire Detections – Theory

•  Fires have very high temperatures (> 600 K) compared to their ambient surroundings.

smoke

“true colour” composite


•  The high temperatures result in very intense radiant energy emissions at IR wavelengths, particularly in the middle IR (3-5 µm) spectral region.

smoke

“true colour” composite


•  The high temperatures result in very intense radiant energy emissions at IR wavelengths, particularly in the middle IR (3-5 µm) spectral region.

smoke

“true colour” composite infrared composite

Zhukov et al. (2006)

Veg + 1% Fire

1%

Veg Only (300 K)

Sub-Pixel Fire Detection

x100

Veg + 1% Fire

1%

Veg Only (300 K)


Possible to detect active fires covering < 1000th of pixel!"

Wooster et al (2005) JGR

Spatial Resolutions GOES ( 2 km x 4 km)

MODIS (1 km x 1 km) BIRD (370 m x 370 m)


•  Assuming MODIS pixels = 1 km x 1 km pixel size •  MODIS pixel area = 1 km² = 1 x 106 m²

How Small a Fire Can we Detect?

Assuming fire size = 100 m long x 5 m wide

Fire area = 500 m²

Assume Fire temp = 850 K (background = 300 K)

•  Proportion of pixel as fire (p) = 500 / 1x106

•  p = 0.0005 or 0.05%

AVHRR Data of African Fires"

TIR – 10.8 µm TIR

11 µm MIR – 3.7 µm

AVHRR Data of African Fires"

MIR

11 µm MIR – 3.7 µm

Using MIR-TIR BT difference helps reduce influences due to ambient effects and highlights those due to fire"

MIR – TIR Brightness Temperature Difference

TIR MIR

Jan Dec

ATSR (World Fire Atlas) http://dup.esrin.esa.it/ionia/wfa/index.asp

MODIS Active Fires http://rapidfire.sci.gsfc.nasa.gov/

TRMM Global Fires ftp://ftp-tsdis.gsfc.nasa.gov/pub/yji/DAILY// http://eobglossary.gsfc.nasa.gov/ Observatory/Datasets/fires.trmm.html

Long-term (since ’95) but only night.

Every 6 hrs global since 2002.

~ Monthly diurnal sampling, but only tropics

Example LEO Fire Data

[email protected] (GSE/GEOG-741-S01)

Fire Location & Seasonality Global MODIS Active Fire Dataset

Intecomparison & Synergy: Active Fires & Burned Area Over Africa

MODIS Burned Area (Roy et al) Metetosat Active Fire (Roberts & Wooster)

Fire Radiative Power

The rate of thermal radiative energy release from an actively burning fire"

Fire Radiative Power vs. Rate of Fuel Combustion


Open points – grassy fuels Solid points – woody fuels

Fire Radiative Energy vs. Total Fuel Combustion

MSG SEVIRI

Fire Radiative Power

Large emissions variability

SEVIRI Fire Radiative Power (FRP) Product (http://landsaf.meteo.pt/)

Simulated “Global product” generated from FRP pixel derived for

different dates only (as a visual example; normally relatively few fires are burning in North and South Africa on the same date)

SEVIRI FRP Pixel Product

FRP Pixel product generated for four regions:

•  Euro (Europa): Red

•  NAfr (Northern Africa): Magenta

•  SAfr (Southern Africa): Blue

•  SAme (Southern America): Brown

Spatial Resolution : SEVIRI Pixel Temporal Resolution : 15 Minutes

Southern Africa FRP, 3-8 September 2003

Biomass Combusted

= 3.2 million tonnes (4.3-5.1 million tonnes adj. for cloud cover)

•  Integrate FRP [MW] over time..(can assume 15 mins [900 secs] x-axis interval) •  Biomass Burned [kg] = 0.368 x FRE [MJ]…and biomass is ~ 47% Carbon

Roberts et al (2005) JGR


Fire Seasonality and Location Temporal Emissions Variation

→ NH Africa 362 - 414 Tg → SH Africa 402 - 440 Tg [Very strong seasonal cycle]

Summary Active Fire Detections

•  Can be “Near Real Time” •  Provide good data on fire timing

and location •  Good for confirming or seeding

burned area mapping methods •  Can be used to give rough

estimate of burned area

but

•  Usefulness may depend on time of observation with respect to the fire diurnal cycle.

Fire Radiative Power •  All the points at left AND •  Provide direct information on

fuel consumption rate •  Can temporally integrate to

produce total C emissions •  Independent of burned area/

fuel load approaches.

but

•  Missing small fires & cloud cover mean these estimates are likely to be minimums if adjustments not made.

Greenhouse gas emissions from wildfires in Africa

•  Dr Bob Scholes, •  Sally Archibald. •  CSIR, •  Natural Resources •  and the Environment •  South Africa •  [email protected]

The basic wildfire emissions model

Emission = Area * Fuel * Completeness * Emission Factor*10-3

tons ha tons/ha % g kg-1

Can be applied to whole ecoregions, or on a pixel-by-pixel basis

To summarise:

•  Tier 1: Countries should stratify by IPCC vegetation categories and early-season or late-season burning. Default values are provided for combustion factors (Table 2.6 ), emission factors (Table 2.5), and above-ground biomass (Table 6.4).

•  Tier 2: Countries should develop their own stratification of vegetation and use country-specific combustion and emission factors.

•  Tier 3: Countries should develop algorithms to estimate the area burnt, validating the products obtained with data from field observation

Combustion completeness:

Early-season burn

Combustion completeness:

Hely et al (2003) J Arid Environments

late-season burn

Com

bust

ion

com

plet

enes

s Fu

el b

urne

d/Fu

el e

xpos

ed

IPCC

gui

delin

es T

able

2.6

0.72-0.92

Emission factors:

The ‘carbon neutral’ assumption

•  It is assumed that for vegetation that burns regularly and regrows to its original state after burning, the CO2 emissions during the fire are balance by CO2 uptake during recovery

•  This is only true if the fire frequency and fuel load are constant over time –  Not true if land is being cleared for agriculture –  If fires are becoming more frequent or intense, the carbon store on

land will decline, ie there are net CO2 emissions

•  It is not true for non CO2 emissions.

ANNUAL BURN

NO BURN IN 50 YEARS

Changing fire regimes to accumulate carbon:

Compound X g compound/

kg Dry Fuel burned

SD

Carbon dioxide CO2 1613 95

Methane CH4 2.3 0.9 Nitrous oxide N2O 0.21 0.1

Nitrogen oxide NOx* 0.31 0.24

CO* 65 20

* not a greenhouse gas, but a precursor to O3, which is. Estimation not required by non Annex-1 countries

Emission factors:

Emission factors:

(van der Werf et al.)

Total Carbon Emissions from Burned Areas (via Active Fire Data) – Global Fire Emissions Database

For more information: •  Andreae, MO 1997 Emissions of trace gases and aerosols from southern African savanna

fires. In: van Wilgen, BW, MO Andreae, JG Goldammer, JA Lindesay (eds) Fire in southern African savannas. Witwatersrand University Press, Johannesburg. Pp 161-183

•  Cachier, H., Liousse, C., Pertusiot, M., Gaudichet, A., Echalar, F. and Lacaux, J. (1996). African fire Particulate emissions and atmospheric influence, in Biomass Burning and Global Change: Volume 1. Remote Sensing, Modeling and Inventory Development, and Biomass Burning in Africa, J. Levine, Editor. MIT Press: Cambridge. p. 428-440.

•  Cachier, H., Ducret, J., Brémont, M. P., Gaudichet, A., Yoboue, V., Lacaux, J. P., and Baudet, J., 1991, Characterization of biomass burning aerosols in a savanna region of the Ivory Coast, in J. S. Levine (ed.),Global Biomass Burning: Atmospheric, Climatic and Biospheric Implications, MIT Press, Cambridge, MA, pp. 174–180.

•  Lacaux, J., Cachier, H. and Delmas, R. (1993). Biomass burning in Africa: an overview of its impact on atmospheric chemistry, in Fire in the Environment: The Ecological, Atmospheric, and Climatic Importance of Vegetation Fires, P. Crutzen and J. Goldammer, Editors. John Wiley & Sons: Chichester. p. 159-191.

•  Scholes, MC and MO Andreae 2000 Biogenic and pyrogenic emissions from Africa and their impact on the global atmosphere. Ambio 29, 23-29

•  Scholes, RJ, D Ward and CO Justice 1996 Emissions of trace gases and aerosol particles due to vegetation burning in southern-hemishere Africa. JGR 101, 23677-82

•  Ward, D. E., W. M. Hao, R. A. Susott, R. E. Babbitt, R. W. Shea, J. B. Kauffman, and C. O. Justice (1996), Effect of fuel composition on combustion efficiency and emission factors for African savanna ecosystems, J. Geophys. Res., 101(D19), 23,569–23,576.

•  Roy, D.P., Lewis, P.E. and Justice, C.O., 2002, Burned area mapping using multi-temporal moderate spatial resolution data—a bi-directional reflectance model-based expectation approach. Remote Sensing of Environment, 83, p. 263–286.

•  Roy, D.P., Jin, Y., Lewis, P.E. and Justice, C.O., 2005, Prototyping a global algorithm for systematic fire-affected area mapping using MODIS time series data. Remote Sensing of Environment, 97, pp. 137-162.

•  Giglio, L., Loboda, T., Roy, D.P., Quayle, B. and Justice, C.O., 2009, An active-fire based burned area mapping algorithm for the MODIS sensor. Remote Sensing of Environment, 113, pp. 408-420.

methods of detecting burnt area and estimating emissions...burned area (km2), number of scars and %...

Documents