protected areas), high (either existence of both protected areas
and protected valleys or existence of areas neighboring natu-
ral reserves), and very high (existence of natural reserves).
The environmental value index was mapped using informa-
tion about the occurrence of natural reserves and protected
areas within each cadastral unit in Lebanon.
Replacement Value Index
The replacement value index was calculated by estimating a
relative value for recovery of a forest ecosystem after fire. Giv-
en that the regeneration of treeless forest systems is relatively
fast, the actual replacement cost comes from the wooded
forest systems affected by the fire. To calculate replacement
value, the actual forest area within each cadastral unit form-
ing Lebanon has been taken into consideration. The intervals
that would determine the replacement value indexes were
defined through the analysis of the range of values obtained
for the whole country (maximum value - minimum value) and
through the division by the number of defined intervals. Ac-
cordingly, this index was mapped by extracting wooded forest
areas of cadastral units from the land-cover/land-use map of
Lebanon, and the intervals and categories were defined as per
Table 6.
Overall Vulnerability
The overall vulnerability represented the combined classifica-
tion of demographic and forest vulnerabilities at the super-
object level.
First, the demographic vulnerability was mapped by add-
ing the three indicators (occupation, boundary, and scatter) in
a pairwise classification analysis, using contextual informa-
tion of image objects, and following the previously discussed
cross-mapping approach. Image objects were classified into
the classes: low, moderate, and high demographic vulner-
ability. Second, the mapping of overall forest vulnerability
involved adding the replacement and environmental value
indices. The overall vulnerability map was the product of
the previously discussed demographic vulnerability and the
forest vulnerability maps. The classes of low, moderate, high,
and very high overall vulnerability were classified using
contextual features and following the same cross-mapping
classification approach as in the previous cases.
T
able
6. I
ntervals
and
C
ategories
of
the
R
eplacement
V
alue
I
ndex
Wooded forest surface (ha)
Replacement value index
0
0
]0-0.25]
1( low)
]0.25-0.5]
2 (moderate)
]0.5-0.75]
3 (high)
>0.75
4 (very high)
T
able
7. C
ombined
U
se
of
O
verall
V
ulnerability
and
W
ildfire
H
azard
for
M
apping
W
ildfire
R
isk
Overall vulnerability
\wildfire hazard
Low Moderate High Very
high
Low
Low Moderate Moderate High
Moderate
Moderate Moderate High High
High
Moderate High High Very
high
Very High
High High Very
high
Very
high
Mapping Wildfire Risk
Eventually, wildfire risk was mapped as product of wildfire
hazard and the overall vulnerability. Specifically, the com-
bined classification of wildfire hazard and overall vulnerabil-
ity resulted in the final wildfire risk map with the following
classes: no risk, low, moderate, high, and very high wildfire
risk (Table 7).
Results and Discussion
The mapping of current wildfire hazard in Lebanon involved
the use of (a) fire spread derived from fuel type, combustibil-
ity and the slope of the terrain, and (b) fire potential index
derived from geospatial climatic data.
Lebanon’s vegetation cover can be classified into a vari-
ety of fuel structure and density that exert strong control on
characteristics of fire spread. Monthly and annual wildfire
potential maps for current climatic conditions were generated
to estimate the fire potential index. All classifications were
documented in spatially explicit maps and the quality of the
calculated monthly
KBDI
values was evaluated by comparing
monthly
KBDI
averages for Lebanon and monthly
FWI
averages
for the Eastern Mediterranean as extracted from the European
Forest Fire Information System (
EFFIS
) for the past decade.
This comparison showed a similar trend of increasing wildfire
potential between March and September (Figure 3).
Furthermore, the
KBDI
results for each month of the year
comprised the following information for each of the previ-
ously listed wildfire potential classes: (a) the total area of the
class in each month, (b) the corresponding mean
DEM
(meters
above sea level (masl), and (c) the equivalent land-cover/land-
use area (i.e., forests, grasslands, among others). Overall, the
class “Very High” occurred in each month of the year with an
increasing trend starting the month of March and a decreas-
ing trend starting the month of December. The peak months
of “Very High” wildfire potential occurred in the months of
September, October, and November. It was also found that the
classes “High” and “Very High” gradually expanded to areas
at higher elevations starting April-May continuing to Sep-
tember. Comparing the distribution of forested and grassland
vegetation types in relation to wildfire potential, it was found
that these vegetation types experienced “Very High”
KBDI
val-
ues for the months of August throughout November.
The current wildfire hazard map provided an overview of
the fire hazard distribution across Lebanon (Figure 4). As a re-
sult, it was found that 203,109 ha (60 percent) of Lebanon (ex-
cluding agricultural and non-vegetated areas) were classified
as high hazard, while 123,832 ha (37 percent) were classified
as moderate fire hazard. Only 4,182 ha (1 percent) were clas-
sified as low hazard, and 7,096 ha (2 percent) were classified
as very high hazard. The determination of the final wildfire
hazard classes was affected by a combination of bio-physical
and climatic factors. This is expected to minimize the impact
of potential inherited errors in the originally employed data
(e.g., spatial climatic data). A more detailed investigation of
the classification results showed that slope steepness across
Lebanese territories and the presence of dense vegetation
cover were the main contributors to an increased fire spread,
and therefore, to an increased overall wildfire hazard. On one
side, 67 percent of Lebanon was classified as low combustibil-
ity fuel, 11 percent as moderate combustibility, 12 percent as
high combustibility, and 10 percent as very high combustibil-
ity. On the other side, 67 percent of Lebanon was classified
as moderate fire spread while, 18 percent as high fire spread,
and 3 percent as very high spread.
The extent of burned areas between 1999 and 2012 was
previously collected for evaluation of the performance of the
fire hazard classes. The sum of 7,260 ha of burned areas was
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
June 2015
503
