Transcript Fire Heterogeneity - Southern African Fire Network

Estimating biomass fuel
loads for fire severity in
the KNP
Tobi Landmann
CSIR Environmentek
[email protected]
4th SAFNet meeting, Skukuza, August 2003
Fire severity – Ecosystem - Management
Aspects of fire research
Fire Heterogeneity
Parameters
Return interval
Biodiversity, Ecosystem
Management response
Heat yield
Climate
Pyrogenic
Emissions
Fuel
Global climate change research
Resource
management
objectives
Background
• Fire severity (FS) - a descriptor fire heterogeneity
– largely related to pre-burn fuel mass/condition
• There is a need for spatial explicit & dynamic
information on fire severity (FS) over phenological
diverse regions (GOFC, 2001)
• There are currently large uncertainties in models
that predict grass-tree ratios and derive litter mass
Fire severity in RS
Def: The magnitude/degree of change from
converting fuel biomass to charcoal and/or ash
RS can be used to infer parameters related to FS:
– Fuel consumption rates; pre-burn fuel moisture (e.g.
Ceccato et al 2001), biomass fuel mass [g/m2] and
composition (e.g. Michalek et al 2001)
– During-fire observations such as on heat yield (kJ/kg),
combustion efficiency (CE)
– Spectral characteristics of the burn scar, i.e. ash color
and combustion completeness (cc)
1. Approach…fuel modeling
2000
30m-Landsat ETM+
acquired coherently
with field
measurements
2001
2003
ETM+
acquired coherently
with field
measurements
ETM+
acquired coherently
with field
measurements
Calibration to
surface reflection
(ρ)
ETM+ diff. ratios,
indices, & single
band spectra vs. biotic
field data
Regional Southern
African fuel prediction
Model (Scholes et al.)
Cross validation/
parameterization
Create
models
2003 ETM+ baseline image
to extrapolate 2000,
2001, 2003 data
Field measured grass mass [gDM/m2]
correlates significantly with ETM+ Tasseled
Cap brightness [ρ]
y = 0.2659x-2.4707, R2= 0.66
760
Grass mass[g/m2]
660
560
460
360
260
0.27
0.29
0.31
0.33
0.35
0.37
ETM+ Tasseled Cap
0.39
Grass and leaf litter (g/m2)
predictions using 30-meter Landsat
(i) Pre-burn tree cover [%] and Landsat vegetation-index (R2=0.62)
(ii) Grass & leaf litter [g/m2] vs. tree cover [%] (R2=0.75)
R2=0.62
160
Grass & Leaf litter [g/m2]
Tree Cover [%]
80
60
40
20
0
0.17
0.19
0.21
0.23
0.25
0.27
0.29
Landsat Vegetation-Index
R2=0.75
140
120
80
60
40
0
20
40
60
Tree Cover [%]
80
2. Approach…spectral modeling
Blackash/whiteash = relative abundance of white ash endmember in ‘grey’ ash
Reflection (0-1)
0.62
100 white ash
0.52
0.42
40 percent white ash
0.32
‘brown’ vegetation
0.22
12 percent white ash
0.12
9 percent white ash
0.02
NIR
0.3
0.6
0.9
LMIR
SMIR
1.2
1.5
1.8
ETM+ wavebands (µm)
2.1
Spectral unmixing
Each fire-affected pixel signal (X) is the sum of the product
of several ‘pure’ endmember matix [M] of a component i
and its abundance [p]
Mpi + Mpi + Mpi + Mpi = X
The following ‘endmembers’ gave physically meaningful
results:
–
–
–
–
Black ash (with a 9% content of white ash)
‘Grey‘ ash‘ (with a 12% content of white ash)
non-photosynthetic “brown“ residual vegetation
photosynthetic “green“ vegetation
Abundancy_ grey ash
Results
R2= 0.49
0.7
’Grey’ ash is an
indicator of fire
effeciency; related to
the pre-fire fuel mass
0.5
0.3
0.1
-0.1
-0.3
600
700
800
900
1000
Abundancy_‘brown’
Grass + Shrub Biomass (g/m2)
2=
0.4 R
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0.4
CC is an (relative)
indicator of the fraction
of the fuel that is
consumed
0.73
Kambeni 5 Kambeni 7 Kambeni 11
0.5
0.6
0.7
Landsat CC (0<CC<1)
FS is the magnitude of change
induced by fire, thereby
FS = Mgap * CC
Mga p is the abundancy (0 <= p<= 1) of the grey
ash endmember
CC
is the combustion completeness (reflection change)
of the corresponding pixel or area
 Disadvantage: CC requires good scene to scene
geo-location and calibration; unmixing is complex
Fire severity amplitude
FS mapping using a categorical scale
0.48
0.43
0.38
0.33
0.28
0.23
0.18
0.13
0.08
0.03
-0.02
0.05
R 2 = 0.53
Kambeni 5
Kambeni 7
Kambeni 11
R2 = 1
0.10
0.15
0.20
Reflectance change (pre-burn - post-burn)
0.25
2
1
Constraints of using Landsat for fuel and FS modeling:
•
Short ‘window’:175 by 185km, 16-day revisit cycle, spectral limits
•
Models set up for KNP, Madikwe (semi-arid savanna) - spatially
aggregation to daily orbiting MODIS/WAMIS satellites
•
High data costs (600 USD per scene)
1 = Madikwe Game Reserve
2 = Kruger National Park (KNP)
MODIS DB/WAMIS
Hyperion ETM+
MAS
kJ/m/s
CC
Field fuel data
Litter DMP
MODIS
LC TC
LAI
AVHRR
NPP
DMP
Flux data: NPP vs. DMP
RS indices on dry matter production (DMP), composition & f. moisture
Fire characterization from RS
FS_fuel consumption rates
Fire prediction
model validation
(e.g. Rommel)
Conclusion
• RS information on FS is an effective and
powerful tool (e.g. in remote areas) to asses the
relationship between fire heterogeneity and
biodiversity (e.g.)
– erosion mitigation
– vegetation regeneration
• ETM+ fuel information - accurate to ‘really
measure’ the fuel mixtures/mass (in real time instead of modeling it)
Further needs
• Further research needs;
– relate fire intensity (kJ/s/m) to the
corresponding burn scar characteristics in
multi-temporal (pixel to pixel) fire observations
– spatially aggregate FS information to daily e.g.
MODIS observations
– Validate fire prediction models (fire danger index)
• Will FS or CE change as a result of Climate
Change - more woody fuel materials being
available as a result of increasing CO2 uptake or
less fire available fuels due to decreases in MAP?