FIRE BEHAVIOUR MODELLING IN A MARITIME PINE PORTUGUESE FOREST TO SUPPORT MANAGEMENT DECISIONS AT THE STAND AND LANDSCAPE LEVELS
[email protected]
Botequim, B1., Fernandes, P.M2. , Borges, J. G. 1 1
Forest Research Centre, Institute of Agronomy, Technical University of Lisbon, Portugal
2
Centre for the Research and Technology of Agro-Environmental and Biological Sciences, University of Trás –os- Montes e Alto Douro, Vila Real, Portugal
I. BACKGROUND
Wildfires have a substantial impact on forest landscape composition
Research purposes
and constrain the economic viability of maritime pine (Pinus pinaster Ait) in
Mata Nacional de Leiria (MNL)
Portuguese commercial forestry. Wildfire modelling is instrumental for designing effective fire suppression strategies, for reducing risk and to minimize damage. Specifically, to create guidelines to support forest managers, several modeling applications to detect significant fire-landscape interactions between stand-level features and fire behavior were fitted through logistic, nonlinear regression and classification tree analysis to classify maritime pine stands according to fire risk levels. Figure 1. Portugal map with the spatial distribution of Mata Nacional Leiria
II. DATA & METHODS
Mata Nacional de Leiria (MNL) was our case study (Fig. 1), a even-aged maritime pine public
Methodology: three steps
forest located in central Portugal (biometric variables from 539 plots, 10 881 ha).
Table 4. Meteorological scenarios and fuel moisture contents T (ºC)
H (%)
Dead Fuels (%) 1h 10h 100h
35,8
22,5
10
11
13
37,8
20,0
7
8
40,1
17,2
4
5
Meteorological Scenarios *1
STEP I. FlamMAP
Moderate (P75) Average (P90) Critical (P97)
Fuel moisture content *2
Fuel moisture content
(FFMC)
(applied)
91.9
8.8
10
10
94.2
6.5
7
7
95.9
4.8
4
*1
Moderate: 75th percentiles, i.e. higher values occur in 25% of the day from May to October 1998 -2008. The Live Fuels were 75% 120% for Shrub and Foliar, respectivelly. *2 The fuel moistures were calculated using “Fine Fuel Moisture Code” FFMC (Van Wagner and Pickket, 1985) “Canadian Forest Fire Weather Index System” (FWI).
Table 5. Fire behavior characteristics
Moderate Meteorological scenarios
Table 1. Geographical data DTM
Wind speeds (10, 20, 40 km/h)
Resolution 25x25 min max
Altitude(m)
4
140
Slope (º) Aspect
0
35
Min 152
Critical
Min 39.27
Slope, aspect, altitude
Topography
Canopy Characteristics
1 Surface
Vegetation Fuel Model
Canopy characteristics Stand height Crown Base Height (m) crown bulk density (Kg/m3) Canopy Cover (%)
Min/Max 7-29 2-23 0.37-0.315 19-97
+
References Inventory Data Torres ,2004 Faias ,2009 Torres, 2004
(years)
PPIN-05 F-PIN M-PIN V-MAb
+
+
2 Passive
3 Active
*3 Critical Scenario:
Output (intermediate results)
Age
Fuel Model
Max 152
Ocurrence of crown fire
Crown Fire
Table 3. Surface fuel model
Max 30876
Rate of spread (m/min)
Fire behavior characteristics
Nw
INPUT
Fireline intensity*3 (Kw/m)
Surface fire
Average
Table 2. Forest canopy characteristics
4% FMC x 40km/h wind speed
Description
References
20-40
Mature Pinus pinaster plantation
Cruz , 2007
> 60 40-60 < 20
P. pinaster litter P. pinaster litter and understorey Small (< 1 m) Erica, Ulex or Pteropartum tridentatum shrubland
Fernandes et al., 2009
+
+
+
+
7%
4%
10%
7%
4%
10%
7%
4%
10km/h
10%
40km/h
20km/h
+
Database
+
3.a)
3.b)
3.c)
Figure 3. Maps of specific elements of each scenario: (3.a) rate of spread (ROS), (3.b) fireline intensity (FLI), (3.c) and crown fire activity (CFA)
STEP III. JMP
STEP I. COMPUTE FIRE BEHAVIOUR A set of explanatory variables: geographical information (Table 1)and non-spatial as surface fuel model calibrated to Portugal (Table 2) and forest canopy characteristics (based on 539 inventory plots) estimated with specific models for Maritime pine (Table 3), were obtained and adapted as input parameters. Fire simulation was carried out with FlamMap 3.0.0 (Finney et al. 2003) for three typical weather scenarios derived from historical records gathered from May to October over 1998–2008 to represent moderate, average and critical fire weather conditions with 4%, 7% and 10% fuel moisture content respectively (Table 4), crossed with wind speeds of 10, 20 and 40 km/h. Fire behaviour characteristics and type of fire were obtained (Table 5).
STEP II. COMBINE LANDSCAPE LAYERS
Figure 2. Landscape files representing the required themes used to compute fire behaviour
OUTPUT
STEP II. ArcGIS
FFMC
Initial landscape data (Fig. 2), modeled fire behavior characteristics (Fig. 3) and biometric stand variables - tree density (Nº trees per ha), basal area (G, m2/ha), quadratic mean diameter in the stand (Dg, m), dominant height (Hdom, m) - were overlaid in ArcGIS 9.3 for each scenario. A total of twenty Landscape map layers were combined for each scenario combination. Furthermore, a database that stores landscape pixels that are homogeneous according to those attributes was established for each scenario to identify stand characteristics and spatial pattern metrics of fire prone areas.
STEP III. MODELLING PROCEDURE The database with the most critical combination (4% fuel moisture content with 40 km/h wind speed) was selected as input in JMP Statistical Software v. 8 for modelling purposes. All data were classified according to its acquisition costs so that logistic binary regression and nonlinear regression was applied to develop models suited to end users ranging from typical forest practitioners to researchers, providing: (1) two compatible modeling fire behavior equations to predict crown fire activity (Pfcrown) and (2) two compatible nonlinear mortality models for the percentage of basal area killed in maritime pine stands after a wildfire (Pm). i.e. Model I, based on simulator input data and Model II, using easily measurable stand characteristics suiting forest managers (inventory data). Furthermore, a classification tree approach was employed to model (1) the type of fire (surface, passive or active crown fire) (2) the difficulty of fire suppression (Alexander & Lanoville, 1989) and (3) to predict mortality from biometric data to support forest management.
III. RESULTS & DISCUSSION How alternative landscape management can potentially change fire spread?
FIRE BEHAVIOUR MODELLING MODEL 1. PfCrownI
1 1 e ( 53.705+ 1 Fmodel1+ 2 Fmodel2 0.347 ccover 6.502CBH 0.352Slope)
1 22.378 22.378 1
If
Fmodel1 0
If Fmodel1 1
2 22.584 22.584 2
If
Fmodel2 0
1 19.891 19.891 1
1 1 e ( 53.884 + 1 Fmodel+ 3.881 Hdom +1.206 G )
If
Fmodel 0(Litter) If Fmodel 1(Shrub)
A consistent set of models to ”build“ forest landscapes more resistant to fire, that replaces the need to use fire simulators was provided. The first model (PfCrownI) is based on simulator input data and can be used when fuel model, slope, crown base height and canopy cover are known. The second model (PfCrown II) just needs easily available inventory data, e.g. fuel model (litter or shrub dominated), dominant height and basal area.
PREDICT MORTALITY
CART :TYPE OF FIRE
If Fmodel2 1
MODEL 2. PfCrownII
The research efforts facilitated the understanding of the influence of both biometric and environmental variables and allowed to identify high-risk in maritime pine stands and, consequently, supports hazard-reduction silvicultural practices, through the development of management guidelines for fuel and stand structure modification in these fire-prone forest stands (Table 6). Furthermore, the results allow mortality prediction for a mosaic of Maritime pine forest and development stages across the landscape based solely on fuel model or relatively simple inventory data (Fig. 4). MODEL 1. PmI
Table 6. Illustration of fire behavior according the effect of changes in fuel characteristics with different stand structures. CBH (m)
BIOMETRIC VARIABLES CBD FModel 3) (kg/m