Fire and the Fire Regime Framework - Fire in Mediterranean Ecosystems - page 36

derived from the Landsat thematic mapper (TM) sensor seem to hold the most
promise (but see Gitas et al. 2009 for a review of the ability of other sensors to
detect fire severity). One of the more widely utilized Landsat-derived indices is the
differenced Normalized Burn Ratio (dNBR), based on the difference between near
infrared (NIR) and middle-wave infrared (MIR), which is used to calculate the
Normalized Burn Ratio (NBR) before and after fire and is defined as

ð NIR MIR Þ

NBR ¼

ð NIR þ MIR Þ:

Indices of fire severity reflect the difference between prefire NBR and postfire NBR
and are termed the differenced Normalized Burn Ratio or dNBR, defined as:

dNBR ¼ prefire NBR postfire NBR :

This technique generates a spatially explicit estimate of fire severity based on the
detection of change in vegetation cover and biomass, the exposure of soil and
alteration of soil color. The images resulting from this analysis reveal that the
burned area within the perimeter of a fire comprises a mosaic of different fire
severities ( Fig. 2.3 ). There is some evidence that the calibration of the remote
signals to fire severity may differ between vegetation types (Hammill & Bradstock
2006 ; Roy et al. 2006 ). However, even across broad landscapes, satellite-based
indices of fire severity are strongly correlated with parameters assumed to be
related to fire severity. For example, Bradstock et al. ( 2010 ) found such measures
of fire severity to increase under severe fire weather and decrease with time since
last fire in Australian Eucalyptus -dominated landscapes. In southern California
chaparral there is a highly significant decline in this measure of severity with
number of times a site burned (Keeley et al. 2008 ).
Miller & Thode ( 2007 ) have proposed that there are advantages to use of the
relative dNBR, where the dNBR ratio is expressed relative to the prefire NBR (i.e.
a measure of prefire biomass cover),

RdNBR ¼ dNBR = prefire NBR :

It appears that the absolute dNBR gives an estimate of fire intensity as it
expresses a quantity related to the loss of total biomass, whereas the RdNBR
gives a measure more closely associated with the percentage mortality of above-
ground vegetation. As a result, RdNBR is widely used in forested areas subject to
a combination of surface and crown fires as it estimates tree mortality; however, in
non-forested crown fire regimes where complete top-kill is common, RdNBR may
not be very informative (Keeley et al. 2008 ).
Controls on fire intensity and fire severity include both endogenous and exogen-
ous ecosystem factors. Fuel loads and fuel moisture are key endogenous controls
and antecedent climate and fire weather as well as topographic location are
critically important exogenous factors (Odion et al. 2004 ; Keeley et al. 2008 ;
Bradstock et al. 2010 ). Past fire history as well leaves a legacy of standing and
downed fuels that affects severity of subsequent fires (Thompson & Spies 2010 ).