A comparative assessment of plant flammability through a functional approach: The case of woody species from Argentine Chaco region
Abstract
enRecent changes to fire regimes in many regions of the world have led to renewed interest in plant flammability experiments to understand and predict the consequences of such changes. These experiments require the development of practical and standardised flammability testing protocols. The research aims were (i) to compare plant flammability assessments carried out using two different approaches, namely functional trait analysis and testing with a shoot-level device; and (ii) to evaluate the effect of disturbances and seasonal variability on flammability. The study area was located in the Western Chaco region, Argentina, and 11 species were selected based on their representativeness in forests. We studied six functional traits related to flammability, growth habit and foliar persistence, in forests without disturbances over the three last decades as well as in disturbed forests. The seasonal variation of these functional traits was evaluated over two consecutive years. Functional trait flammability index (FI) and shoot-level measurements followed standard protocols. Sixty per cent of the species measured presented a high to very high FI. The results of both assessment methods were significantly correlated. Both methods identified the same species as having medium flammability, but differed in regards to the most flammable species. Senegalia gilliesii was identified as the most flammable species when using functional trait analysis, whereas shoot-level assessments found Larrea divaricata and Schinus johnstonii to be the most flammable. There were no disturbance effects on the FI but there was seasonal variation. Our results validate the use of functional traits as a predictive method of flammability testing and represent the first global effort comparing flammability obtained through functional trait analysis with empirical measurements. The significant correlation between both methods allows the selection of the one that is more appropriate for the size of the area to be evaluated and for the availability of technical resources.
Resumen
esCambios recientes en los regímenes de fuego a nivel mundial han aumentado el interés en los experimentos de inflamabilidad vegetal. Estos experimentos requieren el desarrollo de protocolos prácticos y estandarizados. El presente trabajo comparó dos metodologías para determinar la inflamabilidad: rasgos funcionales (RF) y pruebas con un dispositivo de baja tecnología (DBT); y evaluó el efecto de los disturbios y la estacionalidad en la inflamabilidad. Los experimentos se realizaron en la Región Chaqueña Occidental, Argentina, y once especies fueron seleccionadas según su representatividad. Se evaluaron seis RF relacionados con la inflamabilidad, el hábito de crecimiento y la persistencia foliar, en bosques sin disturbios (desde hace 30 años), así como en bosques disturbados. Se evaluó la variación estacional de estos RF durante dos años consecutivos. El índice de inflamabilidad (IF) obtenido a través de RF y las mediciones a través del dispositivo de baja tecnología siguieron protocolos estándarizados. El 60 % de las especies evaluadas presentó un IF alto a muy alto. Hubo una correlación significativa entre los dos métodos de determinación de inflamabilidad. Ambos métodos identificaron las mismas especies en el grupo de inflamabilidad media; sin embargo, hubo diferencias en cuanto a las especies más inflamables. A través del análisis de RF, Senegalia gilliesii fue la especie con mayor inflamabilidad, mientras que a través del DBT, Larrea divaricata y Schinus johnstonii tuvieron mayor inflamabilidad. No hubo efecto de los disturbios en el IF, pero sí hubo variación estacional. Nuestros resultados validan el uso de rasgos funcionales como un método predictivo de la inflamabilidad vegetal y representan el primer esfuerzo global para comparar la inflamabilidad obtenida a través del análisis de rasgos funcionales con mediciones empíricas. La correlación significativa entre ambos métodos permite seleccionar el método más apropiado según el tamaño del área a evaluar y la disponibilidad de recursos técnicos.
Introduction
Vast areas of the world are experiencing accelerated climatic and land-use changes. Natural and anthropogenic disturbances such as fires, pests, drought, livestock, logging and extensive agriculture can alter fire regimes through fuel availability (Batllori et al. 2017; Juárez-Orozco et al. 2017) and generate changes in environmental conditions within the affected areas (Gomes et al. 2018; Gómez-González et al. 2019). Globally, wildfires are increasing even in those ecosystems (e.g. rainforests) where they were not frequent (FAO 2015), and fire regimes are changing even in fire-prone ecosystems. Therefore, the monitoring of plant flammability under different conditions is highly necessary for institutions related to wildland fire prevention, and suppression activities (Yebra et al. 2018).
Flammability is defined by the plant's capacity to ignite and propagate fire and is determined by four components: (i) ignitability, defined as the time taken for a plant to reach the ignition; (ii) sustainability, related to the duration of combustion; (iii) combustibility, which refers to the heat released; and (iv) consumability, related to the proportion of fuel consumed (Anderson 1970; Martin et al. 1994; Hogenbirk & Serrazin-Delay 1995; White & Zipperer 2010). Flammability is an important functional trait in fire-prone ecosystems, such as those developed under Mediterranean climates (Pausas et al. 2004; Pausas & Moreira 2012).
Studies on fire behaviour according to different fuel models have been conducted in many areas with altered fire regimes (Boer et al. 2016; Parks et al. 2016). These models include plant traits related to flammability, such as moisture and dry-matter content in fractions of branches, twigs and leaves; fuel drying time; plant architecture; degree of ramification; flammable compounds (volatile oils, waxes, and resins contents); among others traits (Ormeño et al. 2008; Blackhall et al. 2012; Burger & Bond 2015; Wyse et al. 2016). It is expected that flammability will be greater with high dry-mass content leaves and low diameter twigs, which contribute to fast fire propagation under dry atmospheric conditions such as low air humidity and standing winds (Bradstock 2016).
Common methods of plant flammability assessment involve measurements of small parts of plants under laboratory conditions (Blackhall et al. 2012; Burger & Bond 2015; Wyse et al. 2016). However, assessments on plant flammability in the field could vary broadly. As a complementary alternative to these methods, a functional approach has been postulated by Cornelissen et al. (2003). The plant attributes or traits which are mostly related to flammability are leaf and twig moisture content and dry-matter content; plant canopy (degree of ramification); growth habit; and drying time. Plant flammability increases with higher dry-matter content in twigs and leaves, and with a greater plant architecture complexity, due to the increased availability of carbon and air for ignition, fire spread and combustion (Cornelissen et al. 2003; Pérez-Harguindeguy et al. 2013). Foliar persistence (evergreen/deciduous) could also modify the amount of fuel biomass during fires, especially when they occur during the dry season.
The functional approach of Cornelissen et al. (2003) proposes five plant flammability categories, each of them grouping together a range of functional trait measurements. The plant flammability value obtained by means of this method is a mean value of all the individual trait measurements falling within the category. Five flammability categories are then obtained as a mean value (rounded to one decimal) of each of the class scores. More recently, Pérez-Harguindeguy et al. (2013) recommended the use of functional attributes or traits and of a low-technology device, henceforth called a shoot-level device, for flammability assessments. The shoot-level device consists of a portable cylinder that is used for standardised flammability assessments in the field. Since the recommendation of these two approaches to estimate plant flammability (Pérez-Harguindeguy et al. 2013), there has yet to be any effort to see how the results of these approaches compare.
A comparison of flammability assessment methods could improve the predictive potential of each method applied separately, since functional traits determine how prone the vegetation is to ignite, and the shoot-level tests are based on how it responds to fire as fuel. A significant correlation between both methods makes it possible to choose the one that proves more appropriate for the size of the area to be evaluated and for the availability of technical resources at the time of assessment. In wide areas with scarce logistic availability, the functional trait analysis could provide a practical strategy to plant flammability estimations. In protected areas such as National Parks and nature reserves, the availability of vehicles and access roads allows fuel state monitoring using the shoot-level device.
Plant flammability can also be altered by climatic and land-use changes, since they influence the amount and desiccation of natural fuels (Van Bellen et al. 2010; Barros & Pereira 2014), as well as their seasonal variations (Ormeño et al. 2008; Kunst et al. 2012, 2014). Argañaraz et al. (2016) mentioned changes in the fire danger index in years with different water availability in the Chaco Serrano region. The Argentine Chaco region is strongly affected by land-use changes, which include overgrazing of native forests, wildfires, prescribed fires and mechanical treatments on native vegetation in silvopastoral systems (Boletta et al. 2006; Torrella et al. 2015). Studies of the seasonal variation of flammability and the effects of disturbances on it are scarce. The international organisations and institutions for natural disaster assistance highlight the importance of early detection of high fire danger for global communities, for which accessible and practical methods of plant flammability assessment are highly desirable. The increase in human population in rural–urban interfaces demands the enhancement of fire prevention capacity.
The objectives of this work were (i) to compare two different approaches for plant flammability assessments, namely one focused on functional traits, and a second one related to measurements using a shoot-level device; and (ii) to evaluate the effects of disturbances and seasonal variation on flammability in eleven native woody species of the Argentine Chaco region. We evaluated the following hypotheses: (i) plant functional traits can predict plant flammability in native woody species (Cornelissen et al. 2003; Mason & de Bello 2013; Pérez-Harguindeguy et al. 2013); (ii) plant flammability evaluated by means of a shoot-level device is correlated with the results of plant functional trait analysis (Jaureguiberry et al. 2011; Blackhall et al. 2012); and (iii) land-use history and seasonal variation of fuel conditions modify plant flammability (Ormeño et al. 2008; Kunst et al. 2012, 2014; Argañaraz et al. 2016). This work represents the first effort to improve plant flammability assessment comparing these two complementary methods. Hence, this work could validate the importance of functional traits as a predictive method for assessing flammability, beyond the study region where this research was conducted. A significant correlation between both methods will allow selecting the most appropriate according to the management objectives, size area and the staff training.
Materials and methods
Study area
The study area was located in the Argentine Western Chaco region. This region is characterised by a seasonal semiarid climate, with the most rainfall concentrated from October to March. The mean temperature of the coldest month (July) is 10.6°C, and the mean temperature of the hottest month (January) is 26.1ºC (Boletta et al. 2006). Soils are regosols (Lorenz 1995).
The mean annual precipitation is 574 mm (INTA climatic record, 1934–2000 annual series). The dry season extends from May to October, with a noticeable daily thermal variation, and deficit moisture increases over the dry season (Appendix S1). The Chaco region fire season coincides with the dry season during which vegetation experiments have demonstrated changes in plant moisture content and phenological state, which can alter flammability (Kunst & Bravo 2003; Bravo et al. 2014; Ledesma et al.2018).
The native vegetation of the Chaco region comprises a mosaic of forests, grasslands, savannas and shrublands, where fire has been a frequent ecological event, at least during the last two centuries (Morello & Adámoli 1974; Prado 1993; Bravo et al. 2001, 2010). The Chaco region is the second largest forested region in South America, and forests are considered highly desirable by society (Dinerstein et al. 1995). Chaco vegetation units are currently experiencing changes in land-use and disturbance regimes (Tálamo & Caziani 2003; Grau et al. 2005, 2015; Gasparri & Grau 2009; Bravo et al. 2010; Torrella et al. 2015). The ecotones between savannas and forests represent contact surfaces for different fuel loads and land uses in the current Chaco region landscapes. Since the 19th-century grasslands and savannas have been intensively burnt and pastured and currently occur in different successional states (Grau et al. 2015). Farmers initiate fires to promote the regrowth of grasses and enhance forage production. Elionurus muticus Spreng. Gramineae (aibe) is the dominant grass species in the open areas of the Chaco region, with values of plant coverage ranging from 30% to 90%. This species propagates fires, and its flammability has been attributed to a high rate of biomass production and citral terpenes in its foliage (Burkart 1969; Bravo et al. 2001; Kunst et al. 2012).
The sampling sites were located in INTA's ‘Francisco Cantos’ Experimental Station in Santiago del Estero, Argentina (28°03′S, 64°15′E). The study area extends for 8000 ha and includes a mosaic of forests, grasslands and savannas typical of the Chaco region (Appendix S2). Aspidosperma quebracho-blanco Schltdl. (quebracho-blanco) and Schinopsis lorentzii (Griseb.) Engl. (quebracho colorado) are the two dominant tree species of the upper forest stratum, reaching a height of over 20 m. The medium forest stratum can reach a height of 7–12 m and includes Prosopis nigra Griseb. (algarrobo negro), Sarcomphalus mistol (Griseb.) Hauenschild (mistol), Cercidium praecox (Ruiz & Pav. ex Hook.) Harms (brea) and Geoffroea decorticans Gillies ex Hook. & Arn, (chañar). In the understorey, the presence of several thorny Acacia, Senegalia, Celtis and Prosopis species, Atamisquea emarginata Miers ex Hook. & Arn (atamisqui), Moya spinosa Griseb. (abreboca) is common, accompanied by juvenile individuals of dominant species of the upper stratum (Brassiolo 2005).
Table 1 shows family growth habit and foliar persistence of eleven woody species selected for this study. The species selection was based on floristic inventories of the study area (Araujo et al. 2008) and represented the most abundant species, defined as those that, together, represent 70–80% of the community's standing biomass (Cornelissen et al. 2003). Names of species follow the nomenclature system devised by the Instituto de Botánica Darwinion, Universidad Nacional de La Plata. Growth habit and foliar persistence data were included, as these traits will be used in discussions about plant flammability. Foliar persistence was assigned following descriptions of species by Palacio and Roger (2016).
| Species | Botanical family | Growth habit | Foliar persistence |
|---|---|---|---|
| Aspidosperma quebracho-blanco Schltdl. | Apocynaceae | Tree | Evergreen |
| Cercidium praecox (Ruiz & Pav. ex Hook.) Harms. | Fabaceae | Tree | Deciduous |
| Sarcomphalus mistol (Griseb.) Hauenschild | Rhamnaceae | Tree | Deciduous |
| Schinopsis lorentzii (Griseb.) Engl. | Anacardiaceae | Tree | Deciduous |
| Atamisquea emarginata Miers Ex Hook. &Arn. | Capparaceae | Shrub | Evergreen |
| Celtis ehrenbergiana (Klotzsch) Liebm | Celtidaceae | Shrub | Deciduous |
| Condalia microphylla Cav. | Rhamnaceae | Shrub | Evergreen |
| Larrea divaricata Cav. | Zygophyllaceae | Shrub | Evergreen |
| Maytenus spinosa (Griseb.) Lourteig & O′Donell | Celastraceae | Shrub | Deciduous |
| Schinus johnstonii F.A. Barkley | Anacardiaceae | Shrub | Evergreen |
| Senegalia gilliesii (Steud.) Seigler & Ebinger | Fabaceae | Shrub | Deciduous |
Plant flammability assessments
Plant flammability assessments were performed using two methods: (i) functional trait analysis; and (ii) analysis using a shoot-level device, following Pérez-Harguindeguy et al. (2013). Table of abbreviation is also added after conclusion for more understanding.
Functional traits
In this work, the functional traits selected for plant flammability estimation were as follows: leaf dry-matter content (LDMC), twig dry-matter content (TDMC), foliar moisture content (FMC), twig moisture content (TMC), twig drying time (TDT) and degree of ramification (DR). The LDMC, FMC, TDMC and TDT assessments were carried out on two twigs with a diameter of <2.5 cm and a length of 20 cm extracted from each individual plant. One of the twigs was used for LDMC and FMC assessments, and the other for TDMC and TMC assessments, all using gravimetric methods following Pérez-Harguindeguy et al. (2013). The fresh weight of the samples was recorded using scales with an accuracy of 0.001 g. The material was dried in a controlled temperature oven at 40°C, until the constant weight was reached. The LDMC and TDMC (mg g−1) were assessed by calculating the ratio between the oven-dry mass (mg) and the water-saturated fresh mass (g). The same material was used for TDT assessments, and we measured the time (days) until they reached a constant weight. Foliar persistence and growth habit were considered as other explanatory variables in the statistical analysis. Evergreen species usually have coriaceous foliage with greater LDMC than deciduous species; therefore, we expected a higher combustion time for the former. On the other hand, a shrubby growth habit represents a plant architecture with multiple low diameter twigs, which facilitates oxygen flux and desiccation. The growth habit of trees impose other requirements to oxygen flux and desiccation that could hamper the ignition (Zanforlin-Martini et al. 2007; García-Guzmán et al. 2016).
The degree of ramification was used as a predictor of canopy architectural complexity. Following Pérez-Harguindeguy et al. (2013), we measured the number of nodes and/or branches along a 1-metre-long branch segment, and we categorised the values obtained according to a scale from 0 (no branches) to 5 (four or more orders of ramification). The values obtained from each variable considered were used to assign a flammability category to each species. Each flammability category was calculated as the mean class value (rounded to 1 decimal) of all the traits considered in this work, and five flammability categories (from 1 to 5) were identified. In this work, we considered the flammability index three (3), four (4) and five (5) proposed by Pérez-Harguindeguy et al. (2013) as categories named medium, high and very high flammability, respectively.
The shoot-level device
For shoot flammability assessments in the laboratory, we used a portable shoot-level device designed by Jaureguiberry et al. (2011). This device consisted of a half-cut metallic 60 × 85 cm barrel, placed horizontally and mounted on a removable metallic structure (Appendix S3). The inside of the barrel contained three parallel burners, a grill with an attached thermometer and a blowtorch. The burners and the blowtorch were connected to a propane-butane gas cylinder.
Samples of six individual plants (replicates) for each studied species were analysed in this work. Each sample consisted of a 70-cm-long branch with a diameter of less than 3 cm. This branch size usually starts ignition in the field during wildfires and prescribed fires since it can easily reach the desiccation threshold to initiate fire (Dickinson & Johnson 2001). The temperature of the grill was maintained at 150°C during the whole experiment. The samples were placed horizontally, preheated for 2 min and ignited for ten seconds using a blowtorch.
The variables registered were as follows: Burning rate (BR; cm s−1), which was determined by the ratio between the length of the twig burnt (BL; cm) and the burning time (BT; s); Flame height (FH; cm), as a measurement of fire intensity and combustibility; and burnt biomass (BB; %), which was estimated visually by two operators and then averaged to reduce the experimental error. These parameters were used by Stephens et al. (1994); Dimitrakopoulos and Panov (2001); Jaureguiberry et al. (2011); Burger and Bond (2015); Wyse et al. (2016, 2017); and Padulles Cubino et al. (2018), among others, for plant flammability assessments in herbaceous and woody species. The experiment was conducted in October 2016, coinciding with the flammability peak observed in the functional trait analysis and with the fire activity peak observed in the Chaco region (Kunst & Bravo 2003; Bravo et al. 2014; Ledesma et al. 2018).
Effects of disturbances and seasonality on plant flammability
To analyse plant flammability variations in forests under different historical disturbances, satellite images and a record of recent land-use history taken from INTA's Experimental Station were used for the selection of sampling sites (INTA 2015). Two sampling sites were located in forests with different disturbance histories: (i) disturbed forests (DF), which experience wildfires and mechanical treatments for the partial elimination of shrubby stratum; and (ii) control forests (CF) with no logging or any other forestry exploitation over, at least, the three last decades, which we considered in this work as the reference condition (Appendix S2). In disturbed forests, past fire events were detected by fire scars, charred barks and carbonised woods on the soil. Mechanical treatments such as roller chopping are used in native Chaco forests to control shrub encroachment and to improve the pasture growth for livestock (Kunst et al. 2012). A one-time roller chopping application was used throughout the understorey vegetation (low severity) 7 years before the sampling (INTA 2008). The CF and DF were separated by 4000 m, on the same type of soil, according to the soil map available from Experimental Station (INTA 2009). The study was performed between 2015 and 2016.
A 150 by 150 m plot was delimited in each sampling site. In the first analysis, we studied seasonal variations in plant flammability of the species selected for this study in CF and DF, at three sampling dates along the Chaco region during fire season: in June, August and October. On each date, six mature and healthy individual plants of each species were randomly sampled in each plot. In a second step, we studied the plant flammability annual variation among the species studied in DF, selecting March, May, October and December as the sampling dates, with the same sampling effort. These sampling dates allowed us to analyse seasonal changes in plant flammability over a wider temporal framework, which could be useful in the current climate change scenario. A functional trait approach was used in this section.
Statistical data analysis
For assessment of disturbance effects on flammability, data were analysed through a general lineal model (GLM) using the sampling site and the fire date as fixed effects. When the study area was delimitated to a disturbed forest (DF), a generalised linear mixed model (GLMM) was used to assess seasonal variations of plant flammability, using the Poisson distribution as well as the species and sampling date as fixed effects. A cluster analysis using Ward's method and Euclidean distance was performed, including growth habit and foliar persistence, to identify groups of species with different Flammability Index (FIs). These two last traits were considered relevant to explain the species’ FI since it determines the solid model and fuel particles distribution during combustion (Jaureguiberry et al. 2011; Blackhall et al. 2012). A correspondence analysis was performed to identify associations between categorical traits and flammability groups, and a principal component analysis (PCA) was carried out to determine the association between FI and LDMC, TDMC, FMC, TMC and TDT. To incorporate growth habit and foliar persistence in the PCA, a principal coordinates analysis (PCoA) was performed.
For assessments with the shoot-level device, we performed a PCA to determine flammability patterns considering BR, FH and BB in each species. At least 70% of the variation in the data was explained by the first axis of the PCA; therefore, that axis was used to obtain a flammability index (Fig. 2). This analysis identified variability in the flammability index (FI) and allowed us to categorise species into three flammability groups. The comparison between both methods was performed through a Spearman′s rank correlation using the first axis of each method′s PCA (Wyse et al. 2016). The statistical software used was Infostat/2017 (InfoStat Group, Universidad Nacional de Córdoba, Argentina) with an α = 0.05.
Results
Flammability assessments
Plant flammability in the species studied using functional traits varied from 3.4 in Condalia microphylla in DF to 3.9 in A. quebracho-blanco in CF (Table 2). Plant flammability in 2016 varied from 3.3 in C. praecox to 4.1 in Senegalia gilliesii, corresponding to the medium to very high categories put forward by Pérez-Harguindeguy et al. (2013).
| Species | Site | TDMC | TDT | LDMC | TMC | FMC | DR | FI | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M | SD | M | SD | M | SD | M | SD | M | SD | M | SD | M | SD | ||
| Atamisquea emarginata | CF | 603.61 | 36.82 | 3 | 0 | 524.07 | 31.97 | 39.64 | 3.68 | 47.59 | 3.20 | 4.33 | 1.15 | 3.6 | 0.1 |
| DF | 495.96 | 54.01 | 3 | 2 | 511.87 | 23.07 | 50.40 | 5.40 | 48.81 | 2.31 | 3.67 | 1.53 | 3.5 | 0.3 | |
| Aspidosperma quebracho-blanco | CF | 619.74 | 209.13 | 2 | 1 | 504.68 | 58.06 | 38.03 | 20.91 | 49.53 | 5.81 | 4.33 | 0.58 | 3.9 | 0.6 |
| DF | 469.66 | 48.90 | 2 | 2 | 557.64 | 24.17 | 53.03 | 4.89 | 44.24 | 2.42 | 4.33 | 0.58 | 3.8 | 0.3 | |
| Celtis ehrenbergiana | CF | 671.72 | 161.66 | 2 | 1 | 466.86 | 101.23 | 32.83 | 16.17 | 53.31 | 10.12 | 3.33 | 1.15 | 3.7 | 0.2 |
| DF | 550.99 | 166.25 | 1 | 1 | 484.37 | 106.92 | 44.90 | 16.62 | 51.56 | 10.69 | 3.33 | 1.15 | 3.6 | 0.5 | |
| Condalia microphylla | CF | 575.85 | 246.44 | 2 | 1 | 507.29 | 46.37 | 42.41 | 24.64 | 49.27 | 4.64 | 5.33 | 2.08 | 3.8 | 0.6 |
| DF | 585.86 | 15.45 | 3 | 1 | 485.49 | 19.22 | 41.41 | 1.55 | 51.45 | 1.92 | 4.67 | 1.53 | 3.4 | 0.3 | |
| Cercidium praecox | CF | 645.08 | 98.99 | 4 | 1 | 421.00 | 0.00 | 35.49 | 9.90 | 57.90 | 0.00 | 3.67 | 0.58 | 3.6 | 0.5 |
| DF | 568.71 | 43.28 | 3 | 1 | 301.66 | 0.00 | 43.13 | 4.33 | 69.83 | 0.00 | 4.00 | 0.00 | 3.6 | 0.4 | |
| Larrea divaricata | CF | 512.14 | 108.38 | 2 | 1 | 515.98 | 32.59 | 48.79 | 10.84 | 48.40 | 3.26 | 5.33 | 0.58 | 3.6 | 0.3 |
| DF | 415.52 | 79.33 | 3 | 1 | 508.58 | 67.96 | 58.45 | 7.93 | 49.14 | 6.80 | 4.67 | 1.15 | 3.5 | 0.3 | |
| Moya spinosa | CF | 593.58 | 37.73 | 3 | 2 | 546.10 | 22.50 | 40.64 | 3.77 | 45.39 | 2.25 | 3.33 | 1.15 | 3.6 | 0.3 |
| DF | 592.53 | 33.16 | 3 | 2 | 559.25 | 18.70 | 40.75 | 3.32 | 44.07 | 1.87 | 4.33 | 1.53 | 3.8 | 0.4 | |
| Senegalia gilliesii | CF | 567.57 | 68.84 | 2 | 1 | 516.99 | 71.28 | 43.24 | 6.88 | 48.30 | 7.13 | 5.67 | 0.58 | 3.8 | 0.2 |
| DF | 592.45 | 137.35 | 3 | 1 | 481.19 | 19.09 | 40.76 | 13.74 | 51.88 | 1.91 | 4.67 | 0.58 | 3.7 | 0.0 | |
| Schinus johnstonii | CF | 568.69 | 16.35 | 2 | 1 | 509.71 | 28.68 | 43.13 | 1.63 | 49.03 | 2.87 | 6.00 | 0.00 | 3.6 | 0.3 |
| DF | 518.95 | 78.59 | 2 | 1 | 519.93 | 30.76 | 48.11 | 7.86 | 48.01 | 3.08 | 5.00 | 1.73 | 3.7 | 0.2 | |
| Schinopsis lorentzii | CF | 576.44 | 92.75 | 3 | 0 | 467.40 | 20.27 | 42.36 | 9.27 | 53.26 | 2.03 | 4.00 | 1.00 | 3.5 | 0.4 |
| DF | 585.71 | 88.74 | 3 | 1 | 491.36 | 24.34 | 41.43 | 8.87 | 50.86 | 2.43 | 3.67 | 1.15 | 3.5 | 0.3 | |
| Sarcomphalus mistol | CF | 692.66 | 186.56 | 2 | 1 | 459.05 | 10.06 | 30.73 | 18.66 | 54.10 | 1.01 | 4.00 | 1.00 | 3.7 | 0.6 |
| DF | 563.01 | 126.95 | 2 | 1 | 470.66 | 52.31 | 43.70 | 12.70 | 52.93 | 5.23 | 4.00 | 1.73 | 3.6 | 0.4 | |
- CF, control forest, without disturbances over three last decades, considered as reference condition; DF, disturbed forest with fires and roller chopper applications. References: DR, number of ramifications orders; FI, (flammability index), 1–5 categories following Pérez-Harguindeguy et al. (2013); FMC, foliar moisture content (%); LDMC, leaf dry-matter content (mg g−1); TDMC, twig dry-matter content (mg g−1); TDT, twig drying time (days); TMC, twig moisture content (%).
The principal component analysis (PCA) of functional traits associated with flammability (LDMC, TDMC, TDT and DR), and the two-first coordinates of the principal coordinates analysis for foliar persistence and growth habit traits, explained 73.9% of the data for the first two axes and showed leaf dry-matter content and degree of ramification to be the most important functional traits (Fig. 1a). In the principal coordinates analysis, the PCO1 axis separated species by their growth habit, namely tree and shrub growth habit. The PCO2 separated evergreen species from deciduous species. FMC and TMC were not considered in this analysis to avoid redundancy in the PCA, owing to their strong association to LDMC and TDMC (Fig. 1a).
The first component (PC1) explained 41% of the variability and separated species, according to their growth habit, into the flammability groups established (Fig. 1a). PC1 separated a medium flammability group, integrated by tree species, from the high and very high flammability groups, integrated by shrub species. The second component (PC2) explained 32.9% of the variability and separated species according to their foliar persistence. Regardless of season, the general trend indicated that the more important functional traits related to flammability were LDMC and DR, followed by TDMC, growth habit, foliar persistence and TDT (Fig. 1a).
The first two axes of the PCA performed to assess shoot flammability using three flammability components (FH, BR and BB) which explained 97% of the data and allowed the identification of flammability groups. The first PCA axis (PC1) explained 70% of the variation in the data and was positively associated with the three flammability components (Fig. 1b). The second PCA axis (PC2) explained 27% of the variation and was positively associated with FH and negatively correlated with BR. This axis did not have correlation with burnt biomass (Fig. 1b).
A cluster analysis of DF using functional traits, foliar persistence and growth habit allowed us to identify three functional groups of species with different FIs (Fig. 2a). 63.64% of the species studied presented FI values included in the high to very high flammability categories. The medium flammability species group included tree growth habit, and herbaceous and evergreen leaves, but low dry-matter content in leaves and twigs. The high flammability species group was characterised by shrubby growth habit, and evergreen and coriaceous leaves, with high leaf dry-matter content. Senegalia gilliesii was the only species with a very high FI, with shrubby growth habit, a high degree of ramification, and high leaf and twig dry-matter content (Fig. 2a).
In the shoot-level flammability assessments, the first component (PC1) allowed us to identify the medium flammability group formed by C. praecox, Celtis ehrenbergiana, S. lorentzii and S. mistol as the least flammable group. This group had a negative association with the three flammability components evaluated in this work. A species group with high FI values included S. gilliesii, A. emarginata, C. microphylla and A. quebracho-blanco, Larrea divaricata and Schinus johnstonii showed the highest FI values (Fig. 2b).
The Spearman correlation coefficient between the first components of each method's PCA (functional traits/shoot-level device) showed a significant association between them (Spearman's P = 0.70, P-value = 0.0365; Fig. 3) and that 60% of the species evaluated matched between both methods. Species with a medium FI also matched between both plant flammability assessment methods. The highest FI species according to the assessment with the shoot-level device were L. divaricata and S. johnstonii, whereas functional trait analysis identified S. gilliesii as the species with the highest FI (Fig. 2).
Effects of disturbances and seasonality on plant flammability
Results of anova considering functional traits did not indicate significant differences in FI among sampling sites with different disturbance histories (Fig. 4; Table 2). Figure 4 showed significant differences in FI among sampling dates throughout the fire season, increasing from August to October. These preliminary results guided the selection of DF as the study area and encouraged the analysis of seasonal variations in FI within a wider temporal framework.
In DF, the GLMM analysis considering functional traits showed two flammability peaks during the year, one in March and another in October, with no significant differences between them. The lowest FI values corresponded to May and December, without significant differences between them (Fig. 5).
According to the seasonal variations of the functional traits considered in this study, the most flammable species in May were S. gilliesii, M. spinosa and C. ehrenbergiana (Fig. 6a). In October, the most flammable species were A. emarginata, M. spinosa, S. johnstonii and L. divaricata (Fig. 6b). All of these are shrubs with evergreen and coriaceous leaves, and both sampling dates corresponded to the Chaco region fire season. In December, the most flammable species were S. gilliesii, C. microphylla and S. johnstonii, and this period coincided with the beginning of the rainy season in the region. At this sampling date, the degree of ramification and leaf and twig dry-matter content were the functional traits with the highest influence on flammability (Fig. 6c). In March, S. gilliesii, C. microphylla, A. emarginata and L. divaricata were the most flammable species (Fig. 6d); and degree of ramification and leaf and twig dry-matter content were the most determining traits for plant flammability. At least 57% of the variations in the data were explained by the first two principal components.
Discussion
Comparative assessments of flammability
Flammability assessments by plant functional trait analysis showed a high correlation with the values obtained by a shoot-level device analysis. The medium flammability species group, including trees with low degree of ramification and leaf dry-matter content, matched between the two methods. Nevertheless, the most flammable categories showed differences between both methods, which can be attributed to the different variables considered by each of them. Plant functional trait analysis estimates a species flammability by relating morphological features with probable ignitability, spread rate and combustion time (Pérez-Harguindeguy et al. 2013), whereas assessments with a shoot-level device measures flammability under controlled conditions (Jaureguiberry et al. 2011; Wyse et al. 2016). The highest flammable species according to the functional trait analysis was S. gilliesii, but this species showed the lowest value within the high flammability species group in the shoot-level device test. On the other hand, the most flammable species as per assessments using this device were L. divaricata and S. johnstonii, which contain resins or volatile compounds (Murray et al. 2008; Ríos et al. 2008). Species of Anacardiaceae are recognised as aromatic species with high contents of volatile oils in leaves and fruits, especially Schinus terebinthifolius Raddi, Schinus polygamus (Cav.) Cabr., and Schinus molle L. (Montanari et al. 2012). Although S. terebinthifolius can initiate positive feedback with fire recurrence in some North American savanna environments (Stevens & Beckage 2009), Ledesma et al. (2018) showed that Schinus bumeloides had a higher combustion time than S. gilliesii and C. ehrenbergiana during experimental burns in Chaco environments.
Our results and the mentioned antecedents suggest that the shoot-level device assessments could be a better fit to analyse flammability in the field. Wyse et al. (2016) found a strong correlation using the shoot-level device with expert opinion on the flammability of tree and shrub species in New Zealand. However, the most important finding of our work is the significant correlation between both flammability methods, which makes possible the selection of the method most appropriate for the size of the area to be evaluated and for the technical resources available at the time of assessment. For studies covering wide areas, functional trait analysis represents a practical and predictive flammability assessment method, while a shoot-level device testing method requires more resource availability and logistic effort. We would like to stress that shoot-level device tests are necessary to make assessments more reliable, since the most flammable species according to this method were L. divaricata and S. johnstonii, which are characterised by having high volatile compounds.
Plant flammability assessment is a main issue for environmental management, especially in fire-prone ecosystems. White and Zipperer (2010) advocated standardising flammability assessment methods, and Jaureguiberry et al. (2011) standardised the methodology of the device used in our work. However, our tests were carried out under controlled conditions; therefore, caution must be applied when data is extrapolated to flammability in the field. Liodakis and Kakardakis (2008) also found three different categories of plant flammability in seven dominant Mediterranean plant species from wildland/urban interfaces in Athens, Greece, considering the ignition delay time, heat content, combustion duration and mass residues as relevant flammability variables. In our work, the PCA of FI obtained by the shoot-level device reaffirmed the positive association between flame height and burning rate with the FI in the species studied.
Flammability in native Chaco region species
The percentage of species with high or very high flammability (63.6% according to the functional trait analysis, and 60% according to the shoot-level device test) suggests that the Chaco forests could be considered as a fire-prone vegetation type, like Cerrado and Caatinga vegetation, which conforms to a recognised semiarid/arid diagonal in South America (Bucher 1982). Even though the functional traits considered in this work are directly related to plant flammability (Kunst et al. 2012; Pérez-Harguindeguy et al. 2013; Burger & Bond 2015), other multiple morphological and biochemical traits (Diaz et al. 2013) contribute to enhancing the plant ignitability, fire spread rate and combustion rate, modifying its features as fuel. The absence of significant differences in the FIs among the species studied using exclusively functional traits (LDMC, TDMC, FMC, TMC, TDT and DR) highlights the relevance of considering growth habit and foliar persistence as complementary traits for their assessment. Landi et al. (2017) considered that forest coverage is less flammable than grassland and shrublands in the Western Chaco region, attributing this to a greater fine fuel availability and fuel desiccation in these two last environments. The flammability categories obtained in our work have been confirmed through field observations during experimental burns in native forests from the same study area. In these experimental burns, differences in flammability, spread rate and intensity have been observed in some of the species studied and attributed to growth habit (Kunst et al. 2012; Bravo et al. 2014; Ledesma et al. 2018).
The high flammability of the shrub species studied could be related to the combination of traits such as low plant height, fine-branching, dense canopy and evergreen leaves with high dry-matter content (Gill & Allan 2008; Burger & Bond 2015). In our work, the medium FI species (S. lorentzii and S. mistol) are medium-stratum, canopy tree species of the Chaco forests, respectively (Araujo et al. 2008). The high FI species are shrub species with evergreen foliage (high LDMC), open canopy with high density, and low diameter twigs. The very high FI species, S. gilliesii, has deciduous foliage and branching architecture, promoting its ignitability. Shrublands from New Zealand, Australia and South Africa have a particular propensity to fire recurrence (Anderson et al. 2015).
Effects of disturbances and seasonality on plant flammability
The lack of significant differences in the FIs of the species studied among sites with different disturbance histories could be related to the high resilience of the native vegetation due to an evolution under these types of perturbations (logging, wildfires, and livestock). Therefore, null or slight changes in plant flammability traits could be expected, at least in this preliminary approach.
Even though Blackhall et al. (2012) and Carbone and Aguilar (2016) suggested changes in foliar flammability properties according to the time elapsed since the last fire or to different fire frequencies, our work did not observe changes in plant flammability among sites with different disturbance histories. Therefore, it is probable that the foliar traits analysed in our work were less sensitive to changes promoted by disturbances, or that these were low severity events. Rosenblum (2015) considered that plant communities and fire regimes make it difficult to discern the effects of fire suppression. The time elapsed since the last fire and low severity disturbances may not represent enough perturbations to promote changes in flammability traits in the studied species.
Bond and van Wilgen (1996) considered that water content and drying rates of plant parts are two of the most important properties affecting flammability at the level of both individual plants and the entire plant community. The seasonal variation of FI observed throughout the Chaco region fire season matched antecedents from experimental fires which study the effects of fuel moisture content (Kunst et al. 2014) and assess the fire severity effects on tree species within the same study area (Bravo et al. 2014). Fire severity increases towards fire season ending due to a decreasing fuel moisture content in fuels and the increasing mean monthly temperature (Kunst & Bravo 2003). The increasing plant flammability during fire season is common in Mediterranean climate areas in which dryness period coincides with high temperatures (Pausas & Moreira 2012). These results suggested that the variables considered in our work may be useful to detect seasonal variations in plant flammability, but could be less sensitive to changes promoted by disturbances.
The peaks of plant flammability observed in the second year of study in March and October could be attributed to different sets of functional traits influencing plant flammability. The peak of FI observed in October coincided with high fuel desiccation, high temperatures and the incidence of winds at a regional level (Kunst et al. 2014; Argañaraz et al. 2016). These environmental conditions have been reflected in the lowest values of twig drying time, foliar moisture content and twig moisture content observed. In the March peak, FI was affected mainly by high LDMC, TDMC and DR produced by leaves and woody growth modules fully expanded at the end of the growth season. In May, traits determining plant FI (leaf and twig dry-matter content) could be related to the percentage of evergreens among the species studied (45%) and the short time period without leaves in deciduous species.
The data discussed above suggest a high flammability for the shrubby stratum of the Chaco forests. These considerations are very important for fire prevention activities, which used to be always concentrated in the Chaco region fire season. Argañaraz et al. (2016) considered that leaf moisture content is a good indicator of fire danger for the Argentine Chaco Serrano. The species FI in March should be monitored during particularly dry years since vegetation characteristics could easily promote ignitions.
Conclusions
Our work represents the first global effort to compare plant flammability assessments by two different methods within the functional approach. The significant correlation obtained between both methods is the most relevant finding of our work due to it allowing the possibility of selecting one of those methods according to the size of the area to be evaluated and to technical resource availability. Functional traits related to flammability coupled with growth habit and foliar persistence enhance the predictive potential of this approach. Even though this study was carried out in a semiarid environment from Argentina, vegetation features are common in other tropical and subtropical environments where these results could be applied. More than 60% of the species studied showed high and very high FIs. There were no significant effects of disturbances on the flammability index of species. The functional trait approach allowed identifying seasonal variations in plant flammability over a year, beyond the fire season of the Western Chaco region. Our results represent a significant advance in fire ecology knowledge for the Argentine Chaco region. Other components of plant flammability as foliar area, specific foliar area and flammable biochemical compounds (terpenes, tannins and phenolic contents, among others) will be considered in future studies.
Table of abbreviations
| BB | Burnt Biomass |
| BL | Length of the twig Burnt |
| BR | Burning Rate |
| BT | Burning Time |
| CF | Control Forests |
| DF | Disturbed Forests |
| DR | Degree of Ramification |
| FH | Flame Height |
| FI | Flammability Index |
| FMC | Foliar Moisture Content |
| GLM | General Lineal Model |
| GLMM | Generalised Linear Mixed Model |
| LDMC | Leaf Dry-Matter Content |
| PCA | Principal Component Analysis |
| PCoA | Principal Coordinates Analysis |
| TDMC | Twig Dry-Matter Content |
| TDT | Twig Drying Time |
| TMC | Twig Moisture Content |
Acknowledgements
The authors are grateful to the authorities of Universidad Nacional de Santiago del Estero (UNSE); Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET); the Ministry of Science, Technology and Productive Innovation; and the Ministry of Environment and Sustainable Development of Argentina for funding our research and providing support for equipment. A special acknowledgement to Universidad Distrital Francisco José de Caldas, (UDFJC) Bogotá, Colombia, for supporting our work and providing us with scientific assistance of their reviewers. We would also like to thank the assistant staff and our field colleagues from Cátedra de Botánica General, FCF, UNSE for their help on fieldwork, with a special mention to Engineers Ana Belén Cisneros and Jaime Andrés Herrán Molano, and Dr Pedro Jaureguiberry for their advice on statistical data analysis and their support during this investigation. We would also like to acknowledge the Journal's editors and anonymous reviewers for their helpful comments on our manuscript.
