onmeters 
| fires (bright red) and their 
ea and Landsat path 92 and 
line grid). 
  
'e acquired 02 Feb 2009. 
  
Figure 3. Delburn classified soil burn severity image. Light 
blue = Low, Yellow = Moderate, and Red = High. 
By convention, NBR is normally scaled by 1000 to transform 
the data to integer format. For burn severity mapping purposes, 
the NBR is generally calculated for both a pre- and post-fire 
image and then used to derive a differenced NBR (dNBR) as 
follows: 
dNBR = NBR NBR 2) 
prefire postfire 
After being developed by Key and Benson (Key, 2006), NBR 
and dNBR have been widely used to map burned areas across 
the USA and internationally. Using dNBR and comparing a 
pre-fire image to a post-fire image captures the fire-related 
changes that interest BAER teams. For example, sites that were 
heavily forested prior to a fire and then experience complete 
vegetation-canopy loss are more likely to exhibit drastic 
increases in runoff during subsequent rainfall. In contrast, sites 
with little pre-fire biomass that experience complete canopy 
loss are less likely to exhibit drastic increases in runoff. Using 
pre-fire imagery in the mapping process also helps account for 
vegetation characteristics and changes unrelated to the fire, 
such as the current effects of historic fires, drought, and 
management activities. 
Landsat or other satellite imagery that possess SWIR and NIR 
spectral bands may not be available for some fires and therefore 
the dNBR index can not be generated. In these cases, EROS 
and RSAC use alternative satellite or airborne scanner image 
indices, most often the normalized difference vegetation index 
(NDVI). NDVI is an index that has been used to identify 
vegetated areas and estimate vegetation condition since the 
early 1970's (Rouse, 1973). It is also a standardized method of 
comparing vegetation greenness between satellite images over 
time. NDVI is computed using Landsat ETM/TM near-infrared 
(NIR) and visible red (RED) spectral bands (4 and 3) 
respectively. NDVI is calculated as follows: 
NDVI — (NIR — RED)/(NIR 4 RED) (3) 
When mapping burn severity with imagery lacking a SWIR 
band (a band used in the dNBR calculation) NDVI has been 
shown to be a suitable substitute for NBR and dNBR (Hudak, 
2007) (Zhu, 2006). For burn severity mapping or change 
detection purposes, the differenced NDVI (dNDVI) is 
calculated using two images acquired at different time intervals 
(i.e. pre-fire and post-fire) as shown in this example for 
mapping burn severity: 
NDVI (4) 
postfire 
dNDVI = NDVI 
prefire 7 
In the United States, BAER teams are tasked to identify these 
burn areas and to produce a full-coverage, four-class soil burn 
severity map. Within a fire perimeter, the maps generally 
possess unburned or very low severity, low severity, moderate 
severity and high severity map classes. RSAC and EROS assist 
in this process by providing BAER teams in the field with a 
number of near real-time remote-sensing based products. 
BAER teams rely heavily on the burned area reflectance 
classification (BARC) map. Figure 3 provides a BARC map 
developed for the Delbum fire. BARC maps are a 
generalization of the dNBR created for team members with 
varying geospatial skills. The BARC has two formats: BARC4 
and BARC256. The BARCA is a four-class (unburned to very 
low severity, low severity, moderate severity, and high 
severity) thematic map layer created by analysts with 
predefined, discrete severity classifications. The BARC256 is a 
continuous-value map layer with a 0—255 data-value range 
generated by rescaling dNBR values to less than full precision. 
If BAER teams analyze the BARC4 map and determine that 
certain elements are misclassified, they can easily re-classify 
the cell values in the BARC256 and recreate the BARCA to 
reflect conditions more consistent with ground sampling data 
and/or observations by local experts. 
In addition to BARC maps, BAER teams in the United States 
are provided with pre- and post-fire georeferenced and top-of- 
atmosphere (TOA) reflectance corrected satellite imagery 
(Huang, 2002). This allows the team to do manual image 
interpretation or additional digital image analyses if they 
possess the required technical expertise. The imagery also 
provides a synoptic view of the entire fire area that may be 
useful for briefings and public presentations. The majority of 
the imagery used to map wildfires in the United States is 
obtained from the Landsat series of Earth-observing satellites. 
The USGS provides this imagery at no cost to BAER teams and 
the public. On occasions when Landsat satellite imagery is not 
available, other domestic and international sources of imagery 
are used (Clark, 2011). 
3. DATA AND TRAINING AVAILABILITY 
3.1 USA BAER Online Data Archive 
Burn severity data layers as well as pre- and post-fire satellite 
imagery, for all Australian fires mapped in 2009, are available 
for download. These datasets include the BARC256 and 
continuous dNBR layers, as well as Landsat TM pre- and post- 
fire image subsets. One exception is the data available for the 
Wilson Promontory fire. Due to clouds, Landsat imagery was 
not available. The products for this fire were generated from an 
image acquired by the Advanced Spaceborne Thermal