Project no. FP6-018505
Project Acronym FIRE PARADOX
Project Title FIRE PARADOX: An Innovative Approach of Integrated Wildland Fire Management Regulating the Wildfire Problem by the Wise Use of Fire: Solving the Fire Paradox
Instrument Integrated Project (IP)
Thematic Priority Sustainable development, global change and ecosystems
D3.2-2. Tree resistance to fire: first results
Due date of deliverable: Month 18
Actual submission date: 31
Start date of project: 1st March 2006Duration: 48months
Organisation name of lead contractor for this deliverable: XG-CIFAL (P10)
Revision (1000)
Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006)Dissemination Level
PU / Public
PP / Restricted to other programme participants (including the Commission Services)
RE / Restricted to a group specified by the consortium (including the Commission Services) / x
CO / Confidential, only for members of the consortium (including the Commission Services)
Authors: José A. Vega, Enrique Jiménez, José Ramón Pérez, (P10)
Paulo Fernandes, Manuel Fernandes, Carlos Loureiro, Hermínio Botelho (P19)
Marco Conedera, Boris Pezzatti, Lara Lucini (P13)
Contributing partners
P10:Centro de investigaciones Ambientales de Lourizán. Centro de Desarrollo Sostenible. Consellería de Medio Ambiente. Xunta de Galicia.
P13: WSL Swiss federal research Institute
P19:Universidade de Trás-os-Montes e Alto Douro
Corresponding author’s address:José A. Vega, CINAM-Lourizán PO Box: 127, CP: 36080, Pontevedra, SPAIN
Tel. +34 986 805011, Fax: +34 986 856420, E-mail:
Reference:
Vega, J.A., Fernandes, P., Conedera, M., Botelho, H., Fernandes, M., Jiménez, E., Loureiro, C., Lucini, L., Pérez, J.R., Pezzatti, B. 2008.Tree resistance to fire: first results. Deliverable D3.2-2 of the Integrated project “Fire Paradox”, Project no. FP6-018505, European Commission, 24 p.
Executive Summary
Knowledge on tree resistance to fire is critical in forest management. However, research in post-fire tree mortality in Europe is scarce. The objective within WP3.2 of the Fire Paradox project to study and quantify direct and indirect tree resistance to both prescribed fire and wildfire, including tree survival in relation to indirect measures of fire intensity, to enable tree damage prediction and deduce tree mortality with a physically based fire model, and to calibrate the 3D fire model outputs with field experiments data covering a range of fire severities. Preliminary results in tree resistance to firefor pine and broadleaved species are presented in this deliverable.
Table of contents
1. Introduction
2. Survival of Pinus pinaster after fire (UTAD and XG-CIFAL).
2.1 Stand structure and fire regime in Pinus pinaster remnants (UTAD)
2.2 Survival of Pinus pinaster trees after wildfire in Spain (XG-CIFAL)
3. Survival of young Pinus nigra trees after prescribed fire (UTAD)
3.1. Introduction
3.2. Study site and methods
3.3. Results
4. Prescribed fire injury in a mixed broadleaved-conifer stand (UTAD)
5. Tree response to fire for deciduous species affected by rapid spreading winter fires (WSL )
5.1 Methodological approach
Field surveys
Statistical analysis
5.2 Preliminary results
6. References
1. Introduction
Predictive models of tree survival after fire can assist in making post-fire management decisions related to hazard tree removal, salvage logging, reforestation, wildlife habitat and watershed quality (Brown et al., 2003). Also, knowledge on tree resistance to low- to moderate-intensity fire is critical to the sound use of fire in forest management. However, and in spite of its potential contribution to management decisions, research on fire resistance and post-burn mortality has been limited in Europe, presumably because of the prevalence of stand-replacement fire and the incipient development of prescribed burning (Fernandes et al., 2008). That knowledge is even more limited for broadleaved species; although these taxa are normally fire resilient, severe and frequent fires can degrade these stands.
This report presents the first results obtained in the frame of the FIRE PARADOX work package 3.2, related with tree mortality after fire in pine and broadleaved trees in Europe.
2. Survival of Pinus pinaster after fire (UTAD and XG-CIFAL).
Maritime pine (Pinus pinaster) stands are highly flammable and, consequently, are prone to stand-replacement fire, but can survive surface fire. Keeley and Zedler (1998) associated P. pinaster, as well as P. sylvestris, to a stand-replacement fire regime, but this perspective has been expanded and detailed by Fernandes and Rigolot (2007). P. pinaster stands historically subjected to surface fire (Tapias et al., 2004) with abundance of fire-scarred live trees (Vega, 2000) have been identified in Spain. Development and coexistence of fire-related traits, namely bark thickness, is highly variable between P. pinaster populations, which is quite obvious from Tapias et al. (2004) data. Nevertheless, bark depth can be generically classified as moderate to high and, regardless of fire intensity, mortality caused by bole injury is unlikely unless trees have DBH <20 cm or heat from extended smouldering girdles the stem base. Recently, the state of the art about our current knowledge on European pines survival after fire as well the main research gaps on this topic, has been compiled by Fernandes et al., (2008).
Two complementary studies are currently being carried out about tree resistance to fire on P. pinaster in theframe of the FIRE PARADOX project. One, developed by UTAD, has the objective of locating and characterizing groups of trees or remnant stands that have survived wildfire in Portugal. The other one, implemented by XG-CIFAL, has the goal of analysing the survival of Pinus pinaster after wildfire in different sites in Spain.
2.1 Stand structure and fire regime in Pinus pinaster remnants (UTAD)
Sampling is restricted to fire events that occurred three or more years before, such that secondary tree mortality induced by biotic factors is accounted for. The study sites are located by a combination of aerial photography and field inspection. The collected information includes physiography (aspect, slope, topographical position), basic fuel data (ground cover and height of shrubs, forbs and grasses; dominating species), tree morphology of live trees (height, DBH, live crown base height), burn severity indicators (depth of char, stem char height, crown combustion, preburn crown base height), DBH of standing or downed dead trees, existence of fire scars in live trees. The shape and dimensions of each patch are measured, characteristics of the surviving trees are averaged on each site, and additional stand structure descriptors are determined (basal area, distribution by size class). The number of fires experienced by each pine patch is determined in reference to the Forest Service fire atlas, which documents fire perimeters since 1975 and combines remote sensing and ground data.
Fig. 1. Preliminar data from Pinus pinaster remnant patches in northern Portugal.
Fig. 1 illustrates the currently available data. Most of the residual stands have burned twice or more (up to 6 times in the period subsequent to 1984), sometimes in consecutive years. An herb layer develops after fire in Cytisus spp. the prevailing shrubland type in most of the sites , which probably explains why the fire interval can be so short. In the following months the sampling programme will be finished. Trees will be cored in selected areas for ring analysis in relation to the fire regime.
2.2 Survival of Pinus pinaster trees after wildfire in Spain (XG-CIFAL)
The study sites are located in areas burned in different wildfires affecting Pinus pinaster stands in Galicia (NW Spain) and other inland regions (Central Spain). Table 1 summarizes the main characteristics of some of the areas. Burned areas have been surveyed and permanent transects including stands with different levels of damage have been installed. The selected trees were marked and different parameters were measured to characterize stand structure, tree size and the degree of damage caused by fire. A description of these parameters is presented in Table 2. Each transect has been periodically revisited to follow up the survival or death of each selected tree. Data presented in this deliverable refer to 15 or 16 months after wildfire occurrence, and consequently need to be considered as a first approach.
Table 1. Characteristics of some study sites
Wildfire / Date / Area burned (ha) / Province / Experimental plots / Number of treesEnfesta / 8-2005 / 3,100 / Ourense / 2 / 119
Tamicelas / 8-2005 / 1,675 / Ourense / 2 / 130
Guadalajara / 7-2005 / 13,000 / Guadalajara / 6 / 219
TOTAL / 468
Table 2. Measured and calculated descriptors of Pinus pinaster size and damage by fire.
Tree size variablesDBH / Diameter at breast height (1.3 m) / cm
TH / Tree height / m
D / Density / trees ha-1
BT / Bark thickness at 1.3 m, assumed from postburn measurements (mean of four measurements) / cm
Fire severity variables
Status / Live (1) or dead (0)
CVS / Crown volume scorched / %
CLS / Crown length fraction scorched / %
BF / Bark factor = (BTm)2
BCHL / Bark charring level
BLCme / Mean proportion of bole length charred
FS / Fire severity at soil level
The characteristics of sampled Pinus pinaster trees are presented in the Tables 3 and 4.
Table 3. Characteristics of sampled trees (for abbreviations see Table 2)
DBH / TH / D / BT / CVS / CLSMean / 20.2 / 10.8 / 786 / 2.2 / 44.5 / 0.4
Range / 4.5-45.6 / 7.0-16.2 / 57-2235 / 0.3-7.0 / 0.0-95.0 / 0.0-1.0
Table 4. Characteristics of sampled trees (for abbreviations see Table 2)
BF / BCHL / BLCme / FSMean / 0.8 / 0.24 / 0.43 / 2.8
Range / 0.0-1.0 / 0.00-0.75 / 0.10-0.97 / 2.1-3.3
Tree mortality data is binary (trees either live or die) which is well-suited to model mortality (or survival) as a probability by using logistic regression analysis. Receiver Operating Characteristic (ROC) analysis based on signal detection theory is often used to evaluate and compare models based on their predictive performance. The probability of survival was fitted following the next equation:
where P is the probability of survival after wildfire (0=tree alive; 1=tree dead), xn are the different explanatory variables, and bn are the coefficients to be fitted.
Table 5 provides alternative interim multivariate logistic models for the probability of Pinus pinaster survival to fire. The models employ either CLS or CVS and either BCHL or BF. The equations have similar performance, as measured by the concordance index (c-index).
Table 5. Multivariate logistic models for 15 or 16 months post-fire P. pinaster mortality
Variables / Coefficients / p / Concordance indexModel / x1 / x2 / x3 / b0 / b1 / b2 / b3
1 / CLS / BCHL / --- / -5.393 / 7.444 / 9.209 / --- / 0.000 / 0.903
2 / CVS / BCHL / FS / -7.145 / 0.046 / 1.060 / 0.000 / 0.887
3 / CVS / BF / --- / 4.006 / 0.049 / -7.897 / --- / 0.000 / 0.889
3. Survival of young Pinus nigra trees after prescribed fire (UTAD)
3.1. Introduction
Pinus nigra is a widespread montane pine of the Mediterranean Basin, but only a few studies containing relevant quantitative information on its postfire mortality have been conducted, respectively in Spain (Ordóñez et al. 2005; Fulé et al. 2008) and France (Pimont and Rigolot 2005). Based on such limited data and a modelling exercise, Fernandes et al. (2008) classify P. nigra in-between the more fire resistant European species (P. pinea, P. pinaster) and those more vulnerable to fire (P. halepensis, P. uncinata). The likelihood of cambium kill is similar between P. sylvestris and P. nigra but the former is comparatively more susceptible to crown kill.
Landscape-scale fuel management with prescribed burning has been implemented in 2005 by the Forest Service in the Marão mountain range in northern Portugal. A 3-ha area occupied by young P. nigralaricio (Corsican black pine) was included within a fuel-break and scheduled to be burned. Tree survival was a minor consideration, thus providing an opportunity to assess the resistance of small-sized P. nigra individuals to fire.
3.2. Study site and methods
The study site is located at 41o16’ 21’’ N, 7o54’ 53’’ W in an E-oriented steep slope (40%). Elevation ranges from 1145-1195 m. The P. nigra plantation had a standage and density of 16 years and 1400 ha-1, respectively, and trees were unprunned and aligned by rows. Site quality (Ribeiro 1994) ranged from low to moderate, depending on elevation. Understorey ground cover was almost total with a mean height of 1 m, and was dominated by the shrubs Pterospartium tridentatum, Erica umbellatta and E. australis. The prescribed fires proceeded downslope and against the wind under very mild fire weather (FWI = 1) in 12-13 December 2005. The existing heavy fuel load and vertical tree continuity were conducive to full crown scorch or even extensive tree torching, but these were reduced by the moderately windy (5-15 km h-1) and cold (5 – 8 oC) local weather conditions.
Fig. 2. Relationship between bark thickness and DBH. ○= control plot trees; ● = profile trees; * = trees from the burnt plot without stem scorch or char at 1.3-m height.
Pre-fire bark thickness BTm was estimated as BTm = 0.0373 DBH 1.2606 (see Fig. 2), an equation obtained by non-linear least squares and developed from 82 trees, of which 30 trees were from a nearby control plot, 32 trees were from the burned area (trees uncharred at 1.3 m), and 20 trees were trees sampled for bark thickness along the stem profile.
Table 6. Measured and calculated descriptors of P. nigra size and damage by fire.
Tree size variablesDBH / Diameter at breast height (1.3 m) / cm
HCB / Pre-fire height to live crown base: estimated from postburn observation / m
TH / Tree height / m
CL / Crown length = TH - HCB / m
BTb / Pre-fire bark thickness at 1.3 m, assumed from postburn measurements (mean of four measurements) / cm
BTm / Pre-fire bark thickness at 1.3 m, calculated from DBH (see text) / cm
BF / Bark factor = (BTm)2
Fire severity variables
Status / Live (1) or dead (1)
BTa / Mean postburn bark thickness at 1.3 m (n=4) / cm
CN / Mean char note, assessed on the 4 quadrants at 0.5 m height
CD / Char depth = (BTm - BTa) x 10 / mm
CDR / Char depth ratio = (BTm - BTa) / BTm
CPP / Charred perimeter percentage at the DBH level / %
CCH / Maximum height of combusted crown / m
CCRH / Relative height of combusted crown = CCH / TH
CCR / Combusted crown ratio = (CCH – HCB)/ CL
CKH / Crown kill height of live trees / m
CKRH / Relative height of crown kill of live trees = CKH / TH
CKR / Crown kill ratio of live trees = (CKH – HCB) / CL
RCL / Residual crown length of live trees / m
RCLRH / Relative residual crown length of live trees = RCL / TH
RCLR / Residual crown length ratio of live trees = RCL / CL
Field work to assess post-burn tree damage and survival was carried in January 2007, ca. 1 year after the fire. Table 6 indicates the measurements made and variables derived from measurements, which are separated in tree size descriptors and fire severity descriptors. We measured the maximum height of combusted crown as a flame length surrogate and have assumed it is a better indicator of potential tree survival than maximum stem char height, whose limits were not always easy to define due to the dark bark of this pine. Live crown descriptors obviously had no interest for modelling tree survival (because dead trees had no live foliage) but were retained to describe the characteristics of live trees. Live trees with poor vitality signs were disregarded. We have additionally sampled bark thickness at the heights of 10, 50, 90 and 130 cm (two measurements on opposite sides of the tree per height level) on 20 trees in the unburned control plot, in order to describe bark thickness variation along the stem profile.
3.3. Results
A total of 259 trees were measured in the study site, covering the entire observed range of tree dimensions and apparent injury. Table 7 displays the mean values for selected descriptors by tree condition (live or dead). Tree DBH and bark thickness were significantly correlated with the remaining dimensional descriptors but had no relationship with tree injury. Individual tree survival to fire was modelled with logistic regression, and started by examining the individual significance of independent variables (Table 8). Descriptors of the relative impact of combusted crown (CCR and CCRH) have clear advantage over the other variables.
Table 7. Mean (±standard error) and range of selected variables for the observations of Pinus nigra death (0, n=123) and survival (1, n=136) after fire.
Variable / 0 / 1 / Variable / 0 / 1DBH / 6.8 (±2.7) / 9.2 (±2.9) / CN / 1.9 (±0.3) / 1.7 (±0.5)
2.0-15.0 / 3.0-16.0 / 0.3-2.0 / 0.0-2.0
TH / 4.0 (±1.2) / 5.0 (±1.6) / CDR / 0.3 (±0.3) / 0. 2 (±0.2)
1.8-7.6 / 1.9-13.1 / 0.0-1.0 / 0.0-0.7
CL / 3.2 (±1.2) / 4.2 (±1.7) / CPP / 73.2 (±25.4) / 46.9 (±34.9)
1.0-6.8 / 1.1-12.3 / 0-100 / 0-100
BTb / 0.4 (±0.2) / 0.6 (±0.3) / CCH / 1.7 (±0.5) / 1.3 (±0.4)
0.1-1.1 / 0.1-1.3 / 0.9-4.0 / 0.8-2.3
BTm / 0.4 (±0.2) / 0.6 (±0.2) / CCR / 0.3 (±0.2) / 0.1 (±0.1)
0.1-1.1 / 0.1-1.2 / 0.0-0.6 / 0.0-0.4
BF / 52.8 (±38.8) / 93.2 (±51.8) / CCRH / 0.4 (±0.2) / 0.2 (±0.1)
4.0-225.0 / 9.0-256.0 / 0.0-0.7 / 0.0-0.6
Table 8. Univariate logistic regression models for the 1st-year probability of Pinus nigra survival after fire (n=259, d.f.=1): 2 values and variable significance
Variable / 2 / p-value / Variable / 2 / p-valueDBH / 45.7 / <0.0001 / CN / 8.1 / 0.0044
TH / 37.1 / <0.0001 / CD / 0.2 / n.s.
CL / 36.4 / <0.0001 / CDR / 17.0 / <0.0001
BTm / 46.2 / <0.0001 / CPP / 43.4 / <0.0001
BTb / 38.8 / <0.0001 / CCH / 32.6 / <0.0001
BF / 46.6 / <0.0001 / CCR / 83.6 / <0.0001
CCRH / 92.0 / <0.0001
Table 9 provides alternative multivariate logistic models for the probability of P. nigra survival to fire. The models employ either CCRH or CCR and either DBH or BF, but the percentage of charred perimeter (CPP) is common to all equations. The equations have similar performance, as measured by the ROC coefficient. Model predictive capacity is below what is found in most survival/mortality modelling studies, which is probably explained because crown scorch was not measured. Scorch-related variables are common to most post-fire tree mortality models.
The logistic regression approach can be complimented by recursive partitioning (CART) analysis, which is useful to reveal structure in data and derive practical discrimination rules. Fig. 3 indicates that trees survive if combustion affects approximately less than one quarter of crown length <0.25 and their DBH exceeds 10 cm, or if their DBH is lower than 10 cm and less than 50% of the perimeter is charred at DBH level. Most trees thinner than 6 cm have died, and all trees affected by 40% or more of crown length combustion did not survive.
The developed models have only descriptive value, but these study case results add to our understanding of fire-induced mortality of small pines and relatively thin-barked pines, especially because it provides insights on the relative roles of cambium and crown kill.
D3.2-2-18-1000Page1of23
Table 9. Logistic regression models for the 1st-year probability of Pinus nigra survival to fire. p<0.0001 for all models.
Model / 2 / ROC / Intercept / DBH / CPP / CCRH / CCR / BFS1 / 133.0 / 0.879 / 0.489 (±0.671) / -0.375 (±0.072)
<0.001 / 0.026 (±0.006)
<0.001 / 5.906 (±1.337)
<0.001 / - / -
S2 / 140.7 / 0.887 / 0.363 (±0.566) / -0.422 (±0.072)
<0.001 / 0.026 (±0.006)
<0.001 / - / 7.924 (±1.640)
<0.001 / -
S3 / 132.1 / 0.879 / -1.839 (±0.542) / - / 0.025 (±0.006)
<0.001 / 5.872 (±1.320)
<0.001 / -0.023 (±0.005)
<0.001
S4 / 139.6 / 0.886 / -1.105 (±0.434) / - / 0.026 (±0.006)
<0.001 / - / 7.762 (±1.607)
<0.001 / -0.026 (±0.005)
<0.001
D3.2-2-18-1000Page1of23