Content uploaded by Miltiadis Athanasiou
Author content
All content in this area was uploaded by Miltiadis Athanasiou on Mar 08, 2016
Content may be subject to copyright.
NATIONAL AND KAPODISTRIAN UNIVERSITY OF ATHENS
FACULTY OF SCIENCE
DEPARTMENT OF GEOLOGY AND GEOENVIRONMENT
Ph.D. Dissertation
Development of an optimal methodology
for forecasting forest fire behaviour in Greece
Miltiadis Athanasiou
Environmental Scientist
M.Sc. in Prevention and Management of Natural Disasters
ATHENS
DECEMBER 2015
Miltiadis Athanasiou
Development of an optimal methodology for forecasting forest fire behaviour in Greece xxii
Development of an optimal methodology
for forecasting forest fire behaviour in Greece
by Miltiadis Athanasiou
Dissertation Abstract
The objective of this dissertation was to improve wildfire behaviour prediction in
Greece for supporting fire management (prevention and suppression) and for
improving fire fighter safety.
A database of one hundred and ninety six fire behaviour data records was developed
from field observations and measurements made during the evolution of a large
number of wildfires in Greece, in eight fire seasons (2007-2014). The results of the
dissertation, were incorporated into a spreadsheet which can be used as a Decision
Support System (DSS).
Subsets of the database were used to test the agreement of predictions of the
BehavePlus surface fire behavior prediction system and the CFIS (Crown Fire
Initiation and Spread) with field observations. Furthermore, predictions of three
spatial wildfire spread simulation systems, namely FARSITE, G-FMIS and FLogA,
were compared to fire spread regarding the head Rate of Spread (ROS) and the
burned area of a part of a large carefully documented fire that took place in North
East Attica in August 2009.
The empirical capacity of firefighters, Forest Service employees and volunteers, with
varying firefighting involvement and experience, to assess expected fire behaviour
for a set of conditions was examined through a specially designed questionnaire. The
questionnaire consisted of illustrations, through photos and descriptions, of 15
wildfire situations. The respondents were asked to choose through multiple choice
questions of fire behaviour.
Furthermore, empirical equations were developed for the prediction of surface and
active crown wildfire ROS and for the estimation of the minimum surface ROS value
for active crowning. Plume dominated fire behaviour was analysed and compared
with wind driven active crown fire behavior. Presence or lack of spotting were also
examined for 106 fire cases, categorizing them in four empirical spotting classes
(absence of the phenomenon, rare, limited and massive spotting) and correlating
these classes with air Relative humidity (RH%) values. The results of all these efforts
were as follows.
It was found that for the Greek fuel models “Evergreen-schlerophyllous shrublands
(1.5 - 3 m), “Evergreen schlerophyllous shrublands (up to 1.5 m)”, “Phrygana II
Miltiadis Athanasiou
Development of an optimal methodology for forecasting forest fire behaviour in Greece xxiii
(Sarcopoterium spinosum)” and “Mediterranean grasslands”, BehavePlus can be a
useful tool for predictions of surface wildfire ROS in tall maquis, short maquis,
phryganic areas where the dominant species is Sarcopoterium spinosum and grass,
respectively. Four statistically significant linear regression equations describing
mathematically the relation of the predicted to the observed ROS were developed.
They can be used for adjusting BehavePlus predictions to match “real world” fire
behaviour and can also be incorporated in fire spread simulation systems used in
Greece. The analysis of the flame length (FL) for the same fuel types which were
described by the same fuel models, showed that BehavePlus predictions are not
reliable. The finding that FL is seriously under-predicted when using BehavePlus with
the Phrygana II fuel model to predict fire behaviour in Sarcopoterium spinosum
dominated phrygana fields is an important result that can be very useful for the
safety of firefighters. It should be seriously taken into consideration in operational
firefighting in the country as the underestimation takes place in a narrow band of FL
values that includes the FL threshold value of 1.2 m which is considered as the limit
for direct attack on the flames with hand tools.
CFIS failed to predict the crown fire type and the active crown fire ROS in Aleppo
pine forests with tall maquis understory. As a result, it is deemed as inappropriate for
operational use in Greece before further testing, but it may be useful as a training
tool for the estimation of crown initiation. The necessary input data can be estimated
using the DSS, which includes three different ways of Canopy Bulk Density (CBD)
estimation for
Pinus halepensis
forests.
FARSITE failed to produce reasonable fire behaviour simulations despite the use of
both the appropriate four custom Greek fuel models and realistic wind field data. G-
FMIS and FLogA produced better results and their performance improved further
when the Wind Adjustment Factor (WAF) was applied.
Surface headfire ROS in short maquis was modeled as a power function of midflame
wind velocity (Windmidflame, km/h). The empirical equation had an exponent equal to
1.036. The equation is valid for fine dead fuel moisture content (FDFMC) values from
4% up to 8% and Windmidflame values up to 16.5 km/h. Surface headfire ROS in
grasslands was also modeled successfully as a power function of Windmidflame with an
exponent equal to 1.199. The equation is valid for FDFMC values from 4% up to 9%
and Windmidflame values up to 25 km/h. The values of the exponents of both these
empirical equations are in the 1 to 2 range, which is in agreement with the results of
other similar published modeling efforts.
Miltiadis Athanasiou
Development of an optimal methodology for forecasting forest fire behaviour in Greece xxiv
The analysis of the subsets of passive crown and active crown fires led to the
generation of a ROS criterion for active crowning in Aleppo pine forests with tall
maquis understory. The equation that was developed calculates, for any specified
value of CBD, the critical surface ROS value (threshold) above which active crowning
can be sustained.
A preliminary result about active crown ROS in Aleppo pine forests is that it is 2.06
times greater than surface ROS in tall maquis understory. It was also found, using t-
test analysis, that the behaviour of the powerful, plume dominated fires of the
database was affected by the conditions that the plume generates rather than by the
characteristics of the fuels that the fire spreads in.
Analysis of the observations and measurements of spotting on maquis, small xeric
shrubs (phrygana) and grass, led to some preliminary conclusions. No spotting was
observed at air relative humidity (RH) values higher than 40.3%. Massive spotting
that triggered extreme fire behavior was documented for RH values lower than 17%.
In regard to the questionnaire, analysis of the empirical fire behaviour estimation by
firefighters and individuals with varying firefighting experience showed that fire
behaviour in fine fuels was seriously underestimated, a case of eruptive fire
behaviour associated with a box canyon was not recognized and there were
weaknesses on telling the difference between the behaviour of a heading and a
flanking fire. The results support the conclusion that fire behaviour training is needed
for filling the knowledge gaps which were revealed.
Based on the findings of the study, a table was assembled suggesting the wildfire
behavior prediction method of choice for each fuel and fire type. A second table was
also developed identifying the most appropriate prediction method for use by the
authorities for fire prevention, suppression and training purposes. Limitations,
weaknesses and strengths are also reported, and cases for which there is no
available wildfire behavior prediction method are identified.
Key Words: Wildfire, forest fire, fire behaviour prediction, fire behaviour modelling,
surface fire, crown fire, evergreen schlerophyllous shrublands, maquis, phrygana,
grassland, fuel models, Aleppo pine,
Pinus halepensis
, Canopy Bulk Density (CBD),
spotting, Decision Support System, Greece