Introduction

Fire is an important factor controlling vegetation structure. This is especially true for fire-prone ecosystems, which cover approximately 40 % of the world’s land surface and owe their distribution and ecological properties to the fire regime (Bond et al. 2005). In recent decades, land use coupled with climate change led to changes in fire regimes making them more destructive and resulting in higher ecosystem vulnerability to wildfires (Benndorf et al. 2007; Pausas and Keeley 2009). Additionally, high urban pressure presents an increasing risk in terms of fire ignition caused by human activities and burned area threatening inhabited areas (Lampin-Maillet et al. 2010; Simeoni et al. 2012; White and Zipperer 2010). Thus, the need for management practices, which strive to minimize negative socio-economic impacts and maximize environmental benefits of fire, is widely recognized (Fernandes 2013; Loepfe et al. 2012).

Fire regime is the result of complex interactions between ignitions, weather, topography, and vegetation acting as fuel (Fernandes 2013; Moreira et al. 2011; Rothermel 1983). Even though the influence of vegetation varies regionally and sub-regionally (Pausas and Paula 2012), it remains the only component that can be directly managed in order to reduce negative consequences of wildland fires (Fogarty 2001; Moreira et al. 2011; Ortega et al. 2012). For this goal to be reached reliable information on vegetation flammability is required, making it one of the essential components of fire risk assessment and management planning (Dimitrakopoulos and Papaioannou 2001; White and Zipperer 2010).

One of the management options often suggested for the wildland–urban interface (WUI) is the recommendation and promotion of less flammable species (Behm et al. 2004; Ganteaume et al. 2013a, b; Liodakis and Kakardakis 2008; Weise et al. 2005; White and Zipperer 2010). Even though there are numerous lists of species based on their flammability (Barkley et al. 2010; Doran et al. 2010; Fitzgerald and Waldo 2001; Lippi and Kuypers 1998; Lorenson and Callahan 2010; Moore-Gough et al. 2010), the complexity of vegetation flammability makes such a ranking challenging and the resulting lists possibly misleading (Fogarty 2001; White and Zipperer 2010).

Flammability is comprised of: (i) ignitability—the fuel ignition delay once exposed to heat, (ii) sustainability—the measure of how well a fire will continue to burn with or without the heat source, (iii) combustibility—the reflection of the rapidity with which a fire burns (Anderson 1970), and (iv) consumability—the proportion of mass or volume consumed by fire (Martin et al. 1994). There is currently no validated method of integrating all these components into one single index of plant flammability (Behm et al. 2011) and recent work has shown that such single indices might be questionable (Kauf et al. 2014). Furthermore, vegetation fuels are highly variable; their flammability changes with genotype, age, season, location, and material tested (Fogarty 2001; Gill and Moor 1996).

Most of the laboratory research on flammability and its relationship to vegetation properties were made on selected types of vegetation fuel such as leaf litter, branch tips, fresh leaves, or fire brands (Bartoli et al. 2011; Curt et al. 2011; Delabraze and Valette 1974; Dibble et al. 2007; Dimitrakopoulos and Papaioannou 2001; Liodakis et al. 2008; Petriccione et al. 2006; Valette 1990). Some studies manipulated moisture content (Bernard and Nimour 1993; Sun et al. 2006) or fuel load (Mindykowski et al. 2011) in order to investigate the relationship between physical (Engstrom et al. 2004; Plucinski and Anderson 2008; Simeoni et al. 2012), chemical (Alessio et al. 2008a; De Lillis et al. 2009; Liodakis and Kakardakis 2008; Ormeño et al. 2009), and/or morphological characteristics (de Magalhães and Schwilk 2012; Doat and Valette 1980; Engber and Varner 2012; Gill and Moor 1996; Scarff and Westoby 2006) of the vegetation fuel and its flammability. Only early work, e.g., Trabaud (1976), and very recent work (Ganteaume et al. 2013a, b) simultaneously tested fresh leaves and corresponding litter, while the influence of season (Alessio et al. 2008b; McAllister et al. 2012; Pellizzaro et al. 2006; Weise et al. 2005) and location (Blackhall et al. 2012; Massari and Leopaldi 1998; Pausas et al. 2012) remained investigated only on fresh vegetation fuels.

The efforts made in laboratory vegetation flammability testing and in establishing vegetation flammability as a concept was recently challenged due to discrepancies between laboratory testing results, field testing, and modeling results (Fernandes and Cruz 2012). Simultaneously, studies that tried to connect flammability to plant traits were defended as an opportunity to better understand the influence of fire as a selective evolutionary force (Pausas and Moreira 2012). Those approaches were shown to be successful in differentiating between species based on their fire regime and fire life history (Engber and Varner 2012; Fonda et al. 1998; Kane et al. 2008). They are also considered relevant as potentially allowing upscaling from laboratory research to the field community level, thus providing practical information for environmental management (Schwilk and Caprio 2011). Even though laboratory experiments cannot be used for predicting or describing flammability of fuels under natural conditions, they can help to improve our knowledge of the effects of fresh and dead fuel properties on flammability and represent basic information for assessing fire risk (Ganteaume et al. 2013b).

The results of fire risk assessments govern management activities in fire prone areas. Thus, misleading or incorrect information can result in adverse actions and increased fire damage. The relationship between weather, dead fuel moisture, and fire behavior has been explored previously (e.g., Aguado et al. 2007; Anderson and Anderson 2009; Catchpole et al. 2001; Fernandes et al. 2008; Plucinski and Anderson 2008; Sow et al. 2013) and taken into consideration in various fire danger systems (Carlson and Burgan 2003; Cruz et al. 2013; Dimitrakopoulos et al. 2011; Vasilakos et al. 2007; Viegas et al. 2000). Nevertheless, characteristics of dead fuels are assumed to be constant even in the most sophisticated warning systems (Carlson and Burgan 2003). Furthermore, flammability lists are often transferred from one area to another (e.g., DeGomez et al. 2012), not taking into account potential spatial differences.

In this study, we investigated seasonal and local variability of morphology traits, moisture content, and flammability related parameters with the aim to check if current practices of: (i) relating plant traits to vegetation flammability, (ii) transferring flammability lists from one area to another, and (iii) presuming unchanged intrinsic characteristics of dead fine fuel could be improved. Here we present the results for leaf litter, the fuel in which most of wildland fires start (DeBano et al. 1998).

Materials and Methods

Study Area and Selection of Species

The present study was conducted in the Mediterranean biogeographic region of Croatia where forest fires present an important issue for environmental management (JRC 2013). Along the coastline, three main sampling sites were chosen, which represent a transect of the climatic conditions in the region (Fig. 1). From northwest to southeast, a distinct climatic gradient can be observed with an increase in: all temperature indices (average, max and min), yearly sum of sunshine hours, precipitation amounts outside of the drought period, duration of the frost free period, and severity and duration of drought (Fig. 2.).

Fig. 1
figure 1

Relief map of Croatia overlapped with map of biogeographic regions. Indicated are locations of: the largest cities, main and auxiliary sampling cites. Individual map elements were obtained from the Flora Croatica Database (http://hirc.botanic.hr/fcd)

Full size image
Fig. 2
figure 2

Walter–Lieth climate diagram for the meteorological stations Rovinj, Zadar-airport, and Orebić closest to the main sampling sites Rovinj, Kožino, and Trstenik, respectively. Data were obtained from the Meteorological and Hydrological Service, Croatia

Full size image

The northernmost main sampling site (45°6′N, 13°36′E) was located on the west coast of Istria, in the vicinity of the city of Rovinj, with an altitude between 1–25 m and S-SW aspect (Fig. 1). As all the pomegranate (Punica granatum L.) trees in the vicinity of Rovinj were intensively managed an auxiliary sampling site was established in Frančeskija (45°28′N, 13°31′E), where a cluster of three adult P. granatum individuals under minimal management was sampled. The sites Rovinj and Frančeskija had comparable climatic conditions and topography, thus they were regarded as one site during data analysis and are referred to as “Istria.”

The second main sampling site (44°10′N, 15°11′E) was located in Northern Dalmatia in the vicinity of Kožino, near the city of Zadar, with an altitude between 50–60 m and varying aspect. As only individual carob (Ceratonia siliqua L.) trees were present in the vicinity of Kožino, an auxiliary site for C. siliqua sampling was chosen in Marina (43°30′N, 16°6′E), where the northernmost mainland population of this species was found. Information gathered at these two sites was referred to as “Zadar” due to similar climatic conditions.

The southernmost sampling site, referred to as “Trstenik” (42°54′N, 17°23′E) was located in the Trstenik area, on the southwest coast of the Pelješac peninsula, with an altitude between 5 and 200 m, in a south oriented bay, with varying aspect (SW-S-SE). Here all six tree species of interest were present within the same location.

The six tested species were: Aleppo pine (Pinus halepensis Mill.), Holm oak (Quercus ilex L.), olive (Olea europaea L.), strawberry tree (Arbutus unedo L.), carob (C. siliqua L.), and pomegranate (P. granatum L.). These species were chosen as we wanted to include main forest (P. halepensis and Q. ilex) and agricultural species (O. europaea), as well as species which are neglected and underutilized in the study area, but have a potential to be used both in forestry and agriculture (A. unedo and C. siliqua). The last species (P. granatum) was included due to its deciduousness and unique leaf morphology (e.g., relatively small and remarkably thin leaves).

Sampling Schedule and Protocol

Leaf litter of all six species was sampled in May, July, and September/October 2010 on all sites, if possible (Table 1). P. granatum was not sampled in May at any of the sites as litter decomposition was too advanced for identifiable leaf litter fragments to be found. Furthermore, at Zadar, all P. granatum trees grew on stony terrain making retrieval of sufficient amounts of material unfeasible. C. siliqua samples were not gathered in Istria as winters are too cold for this species to establish; they were not gathered in May in Zadar (Marina) as there were problems identifying the sampling location.

Table 1 Number of flammability tests performed with regard to sampling location and sampling time
Full size table

May samples were mainly composed of leaf litter fallen during previous year’s abscission, exposed to the local environmental conditions since and partly decomposed, thus both whole leaves and bigger leaf litter fragments that could be identified as belonging to the species of interest were gathered. July samples were gathered as soon as possible after leaf abscission of evergreen species and were composed only of recently fallen leaves, which formed a recognizable, uniform top litter layer. In the case of P. granatum, leaf litter gathered in July was recently abscised due to environmental stress. P. granatum trees in the Trstenik area shed only a limited amount of litter during early summer, resulting in a reduced number of P. granatum flammability tests in July for this site (Table 1). Our intention was to gather the same material, i.e., leaf litter fallen in the year of study, during the September/October sampling in order to assess the influence of short-term summer exposure of leaf litter of evergreen species to environmental conditions on flammability. This was shown to be unfeasible as changes in the material were too large. Litter fallen in the year of study was no longer easily distinguished and it was naturally mixed with that of the previous year. As a result, September/October samples of evergreen species contained also some litter fallen in the previous year. The last sampling of all evergreen species was performed in September, whereas P. granatum samples were gathered in October, as soon as enough litter had fallen. The number of trees from which material originated varied between three and more than 30. It was governed by the presence of individuals, amount of leaf litter produced and possibility to attribute leaf litter to an individual tree. Three P. granatum individuals were sampled in Istria and five C. siliqua plants in Zadar (Marina). All the other samples were comprised of material originating from more than ten trees. Samples gathered after and during rain events were spread in a thin layer and air dried for 2 weeks before storing. Samples were stored in open paper bags at room temperature and humidity.

Fuel Moisture Content Determination and Morphological Measurements

During leaf litter flammability testing, one or two random subsamples per sample were taken for fuel moisture content (FMC) measurements. These samples were equal in size to the flammability test samples (3.0 ± 0.1 g) and their FMC was calculated on dry mass basis according to Eq. (1) and expressed as a percentage (%):

$${\text{FMC}}\,(\% ) = \frac{{{\text{FM}}\,({\text{g}}) - {\text{DM}}\,({\text{g}})}}{{{\text{DM}}\,({\text{g}})}} \times 100$$
(1)

FM presents the initial mass and DM the dry mass of the sample. DM was determined by reweighing the samples after drying for 24 h at 85 °C. The duration of 24 h was considered sufficient, as the samples were relatively small and a temperature of 85 °C was chosen to ensure fast drying without charring the sample (Turner 1981).

The same subsamples were used to determine leaf litter area. Fragile leaf litter was soaked in warm tap water and flattened with a cloth flatiron before its area was measured with a portable area meter. Specific leaf area (SLA) was calculated by dividing leaf litter area by the corresponding dry mass. Dry mass and leaf litter area of a sample were divided by their corresponding number of individual pieces in order to obtain the average dry mass (AM) and average area (AA) of a single particle, respectively. As C. siliqua leaves separate into petiole and leaflets during and after abscission, leaflet, bigger leaflet, or petiole fragments defined an individual particle; for the rest of the species a particle was an individual leaf or a bigger leaf fragment. All measured areas are expressed in cm2 and masses in grams (g).

Flammability Testing

Leaf litter flammability testing was performed using an epiradiator-based method. After consulting previously conducted research on epiradiator-based leaf litter flammability testing (Ormeño et al. 2009; Petriccione et al. 2006; Trabaud1976) and extensive pretesting, we chose to test 3.0 ± 0.1 g sample material at 400 ± 5 °C epiradiator surface temperature. This combination allowed us to achieve flaming combustion more often than Petriccione et al. (2006) and have longer ignition delay than Ormeño et al.(2009), while using comparable equipment (Kauf et al. 2014). In order to achieve the above specified temperature, a 500 W epiradiator with 10 cm diameter radiant disc and a nominal surface temperature of 420 °C (Alessio et al. 2008a; Delabraze and Valette 1982; Ganteaume et al. 2011a; Ormeño et al. 2009; Petriccione et al. 2006; Saura-Mas et al. 2010; Valette 1990) was connected to a variable voltage transformer allowing us to reduce the surface temperature, which was constantly monitored using a K-type temperature probe (GES 900, Greisinger, Germany). As measurements showed that the temperature across the epiradiator surface varied with a variability coefficient of approximately 10 % (data not shown), temperature was measured at a fixed point on the surface and the test would start only if the temperature reading was steady in the aforementioned range. The same temperature probe was used to monitor surface temperature and measure temperature related flammability parameters. All the temperature readings were in °C. Even though the used temperature probe has a relatively long response time, it was the same for all the samples allowing us to compare temperature results within our set of measurements.

Additionally, the chosen amount of material was sufficient to form a layer covering the whole epiradiator surface ensuring a similar heat exposure for all the samples and counterbalancing the potential influence of variable temperature across the epiradiator surface. The horizontal pilot flame was positioned 4.5 cm above the epiradiator surface. Four parameters were measured: (i) ignition delay (ID)—the time elapsed between placing a sample on the epiradiator surface and the appearance of a flame, (ii) flame extinguish time (FET)—the time elapsed between placing a sample on the epiradiator surface and the end of the flaming combustion, (iii) ignition temperature (IT)—the temperature measured at the moment of ignition, and (iv) max temperature (MT)—the maximal temperature reached during the test. Flame residence time (FRT) was calculated as the difference between FET and ID, representing the duration of flaming combustion. ID is considered to be a measure of ignitability and FRT a measure of the sustainability component of flammability (Blackhall et al. 2012; Ganteaume et al. 2011b; Ormeño et al. 2009). FET is taken as an additional sustainability measure. Flame intensity was not determined. Time variables were measured using a stopwatch and expressed in seconds (s). Our targeted minimum number of individual tests for different samples was five. An exception was made for P. granatum at the location Trstenik in July (Table 1).

Flammability tests were performed on three separate occasions: beginning of September (5–6.9.2010), beginning of October (4–5.10. 2010), and end of October (30.10.2010). May samples were tested on the first and second flammability testing occasion, July samples on the second and September/October samples on the third. Samples were tested in series with breaks between series. In every series, all samples tested on the given occasion were tested once in random order. There was an exception made for P. halepensis leaf litter gathered in May at Trstenik, which was tested twice in three separate series. This was because ID was obviously shorter than that of corresponding materials and we wanted to make sure that this was not due to a mistake in the testing procedure. Samples that were tested fewer times were randomly assigned to the series. Testing was performed in a closed room under a simple chamber, minimizing disturbance due to external air movement. As the samples were air dry before bagging, the bags were open and their position frequently changed and stored for a relatively long period under the same environmental conditions, we attributed differences in final moisture content to intrinsic characteristics of the material (Anderson 1990; Kreye et al. 2013) and did not additionally condition them before testing.

Statistical Analysis

IBM SPSS Statistics 21 software was used for statistical analysis of the results. The relationships between FMC, the measured morphological characteristics and flammability parameters were examined through correlation and regression analysis on the sample level. Correlation analysis was based on Pearson’s correlation coefficient and regression on ordinary least square method. Regression results are reported if they were significant and with a coefficient of determination (r 2) higher than 0.30. Exceptionally, FMC results are reported despite a lower r 2. As SLA of P. granatum is much higher than that of any other species included in this study, regressions examining the influence of SLA might be distorted and the significance level overestimated due to P. granatum values. In order to check for this possibility all regressions involving SLA were performed with and without P. granatum data. They were reported only if both analyses were significant and had similar coefficients. Pearson’s coefficient was used to assess correlations between parameters. For correlation analysis, the reciprocal of SLA was used as it was normally distributed.

The effect of species, location and sampling time, and their interactions on the determined parameters was explored by means of a generalized linear-mixed model (GLMM). Sampling time presented repeated measurement; location, species, and individual replicate number defined the data structure. Accounting for data structure was necessary since the subjects (location, species, and individual replicate) were not independent from each other. Rather, they were nested in each other, thus the samples were not completely random. In our study, species were nested within location, and individual replicates were nested within species, thereby creating a hierarchical structure in the data. We accounted for this hierarchy by indicating location as the highest, species as intermediate and individual replicate number as the lowest level of the hierarchy in the data. A separate model was created for each parameter, respectively. For all flammability parameter models species, location, sampling time, and all possible interactions were treated as fixed effects; with the final model being the one in which all effects were significant and further removal of any of them led to a substantial increase of the absolute value of the Akaike information criterion (AIC), indicating a reduced goodness of fit.

Additional analysis, using the same procedure while excluding repeated measurements and the influence of sampling time, was performed only on July samples in order to determine if different populations of the same species produced leaf litter with different flammability characteristics. This analysis was performed exclusively on July samples as they were gathered as soon as possible after leaf abscission of evergreen species, thus exposure to environmental conditions was minimal and it could be presumed that differences in leaf litter flammability, if present, were inherited from living vegetation.

When determining the influence of species, location, and sampling time on non-flammability related parameters, analysis was performed on sample means with exclusion of interactions. This was due to a lower number of measurements. Furthermore, all three main effects (species, location, sampling time) remained in the model even if their influence was not statistically significant, as their removal increased the absolute value of AIC to such an extent that retaining them was considered justified.

In all GLMM-Satterthwaite approximation was used to account for unbalanced data. Possible differences between levels of fixed effects left in the model were tested by pairwise contrast adjusted for multiple comparisons using the sequential Sidak method. Results were considered statistically significant at P < 0.05. Nevertheless, some of the results that failed to fulfill this criterion are discussed as well.

Results

Correlation and Regression Analysis

Due to the multi-collinearity of parameters (Table 2) and the relatively small number of variables, only simple regression analyses were performed. All reported regressions were significant at P < 0.001 (Table 3). When examining the relationship between morphology and flammability parameters, SLA was the morphology parameter exhibiting the highest r 2. An increase of SLA led to a decrease in FRT (r 2 = 0.446) and an increase in IT (r 2 = 0.432). Furthermore, our results revealed a relationship between particle size (AA and AM) and ignitability (ID and IT), since increasing particle size led to a prolonged ignition delay and a higher ignition temperature. Regressions involving FMC had lower r 2 in comparison to morphology parameters, with an increase in FMC leading to an increase of IT and a decrease of both FET and FRT.

Table 2 Pearson’s correlation coefficient for all measured parameters
Full size table
Table 3 Results of the linear regression analysis performed on the whole data set
Full size table

Influence of Species, Sampling Time and Location on Fuel Moisture Content, Measured Morphology, and Flammability Parameters

FMC values were relatively low for all samples and ranged from 6.3 to 13.7 %. Despite the identical storing conditions and low values, both species and sampling time had a significant effect on FMC (Table 4). At the species level, the average FMC ranged from 8.4 % for O. europaea to 11.6 % for C. siliqua, with three homogenous subsets identified by pairwise contrasts. During the sampling period, the average FMC decreased from May (11.4 %) to September/October (9.5 %). Pairwise contrasts separated May from the other two sampling times, with no sig nificant differences between July and September/October.

Table 4 Results of the generalized linear mixed models for all measured parameters
Full size table

Species had highly significant effects on morphology parameters. While SLA was significantly affected only by species, AA and AM were significantly affected by location as well. Both AA and AM increased along the climate gradient, from north to south, and had significant pairwise contrasts between Istria and Trstenik. Zadar did not differ from either location, based on AM, and built a group together with Trstenik based on AA. Even though the effect of sampling time on AA and AM was not significant, both of these parameters were highest in July.

ID and IT were the only flammability related parameters that were significantly affected by location (Table 4). For both of these parameters, the average value decreased from north to south, but pairwise contrasts were significant only for IT showing significant differences between Istria and Trstenik, whereas Zadar did not differ from the other sites. Species and sampling time had a significant effect on all flammability related parameters. Pairwise contrasts at the species level identified four groups based on ID. C. siliqua was identified as the least and P. halepensis as the most ignitable species, with both differing from all other species. P. halepensis was followed by Q. ilex and P. granatum. There were no significant differences in ID between P. granatum, O. europaea, and A. unedo. Nevertheless, these three species were significantly different when comparing their IT, FET, and FRT. P. granatum had the highest, A. unedo an intermediate and O. europaea the lowest IT. Their ordering based on sustainability was opposite to that based on IT. O. europaea was the most, A. unedo intermediate and P. granatum the least sustainable among them, regardless of the parameter used (FET or FRT). For the rest of the species FRT and FET resulted in different grouping.

Sampling in September/October yielded both the highest sustainability (FET and FRT) and ignitability (ID and IT). July samples had an intermediate (FET) to low (FRT) sustainability and low ignitability (ID), whereas May samples had an intermediate ID, the highest IT, and a low sustainability (FRT and FET).

Species × location was the only significant interaction (Table 4). It influenced all time related parameters revealing species-specific responses of these parameters to the local environmental conditions (Fig. 3). Out of six species included in our study, four (A. unedo, C. siliqua, O. europaea, and P. halepensis) showed a decrease in the ID value from north to south along the climate gradient, with pairwise contrasts being significant for A. unedo and P. halepensis (Fig. 3a). Q. ilex and P. grantum did not follow the same trend: Q. ilex reached the lowest ID value in Zadar and P. grantum in Istria. These two species (Q. ilex and P. granatum) together with A. unedo were the ones whose ranking changed the most when comparing species differences within one location. In contrast, P. halepensis, O. europaea, and C. siliqua had a stable ranking between locations as the species with the shortest, an intermediate and the longest ID, respectively.

Fig. 3
figure 3

Time-related flammability parameters for individual species at different locations expressed in seconds (s): a ignition delay (ID), b flame residence time (FRT), and c flame extinguish time (FET). Data represent means and standard errors, with all values in seconds (s). Lowercase letters indicate significant differences between species within one location, and capital letters indicate between locations within one species significant differences. Differences were considered significant at P < 0.05. In the case of flame residence time for olive (O. europaea) and Aleppo pine (P. halepensis), differences between locations at P < 0.10 are indicated with bold-italic capital letters

Full size image

Regardless of the sustainability parameter used C. siliqua, O. europaea, and P. halepensis showed an increase in sustainability from the northernmost to the southernmost location. Within species and between locations pairwise contrasts of both FRT and FET were significant for C. siliqua (Fig. 3b, c). FRT comparisons almost reached the significance threshold for O. europea (P = 0.057) and P. halepensis (P = 0.061). The same was true for the FET value of Q. ilex (P = 0.088), with the highest value reached in Istria, an intermediate in Trstenik and the lowest in Zadar.

None of the interactions involving sampling time had a significant effect in any of the models. Nevertheless, the influence of sampling time was not consistent across locations for all species, as can be seen when looking at ID (Fig. 4). The inconsistent effect of sampling time on different species can be best observed by ID results at the site Trstenik (Fig. 4c), where three species showed almost no change throughout the testing period, while the other three showed a very pronounced increase of ID values in July.

Fig. 4
figure 4

Mean ignition delay (ID) and standard errors of species at all locations across all sampling times. Sampling was performed in Croatia and graphs present following sampling locations: a Istria, b Zadar, and c Trstenik. Values are expressed in seconds (s)

Full size image

Influence of Location on Flammability Parameters of Different Species in July

The effect of location was not significant for any of the flammability parameters measured in July, while the effect of species was significant for all of them. Even though the species × location interaction was not significant (P = 0.148), two of the species, P. halepensis and Q. ilex, showed similar differences between locations (Fig. 5) compared to the overall results (Fig. 3).

Fig. 5
figure 5

Means and standard errors of ignition delay (ID) in July for individual species at different locations. Values are expressed in seconds (s)

Full size image

Discussion

Our study confirms the presence of relationships between moisture content, morphology parameters, and flammability. Nevertheless, our results, combined with those previously published, suggest complex interrelations between these characteristics. Furthermore, we showed that sampling location has an inconsistent and species-specific effect on flammability parameters, and that significant changes of leaf litter characteristics occurred within the fire season. One could argue that the relationships found and effects observed represent the result of an inappropriate method used for flammability testing [i.e., testing masses larger than 1 g (Ormeño et al. 2009) at low heat flux (Fernandes and Cruz 2012)]. Nevertheless, we consider the applied method appropriate for simulating the initial stage of wildland fires and testing ignitability of leaf litter with further justification being given in Kauf et al. (2014).

Fuel Moisture Content and Flammability

FMC is widely recognized as an important characteristic influencing flammability and fire behavior (Anderson and Anderson 2009; Chuvieco et al. 2004; Gill and Moor 1996; Pérez-Harguindeguy et al. 2013; White and Zipperer 2010). As adsorption and desorption of moisture has been shown to be species specific and related to litter characteristics (Anderson 1970; Kreye et al. 2013; Schunk et al. 2013), we chose to use air-dried litter without additional conditioning. The significant effect of sampling time, with the highest value reached for most decomposed May samples, is in accordance with findings of Anderson (1990) who showed that weathering shifts the equilibrium moisture content (EMC) to higher values and argued that this should be taken into consideration when estimating EMC throughout the year. As the air-drying period in our study was relatively long, we can presume that the measured FMC represents EMC. Unlike Kreye et al. (2013), we did not find a clear relationship between FMC and morphological characteristics of the samples, but we measured a lower number of morphology parameters, which covered a much smaller variability in parameters and dealt only with low FMC values.

Small differences and low values of FMC can explain the lack of a significant relationship between FMC and ID. An increase in FMC was shown to lead to an increase in ID (Alessio et al. 2008b; Bernard and Nimour 1993; Ganteaume et al. 2011a; Gill and Moor 1996; Saura-Mas et al. 2012; Trabaud 1976; Weise et al. 2005). Nevertheless, if moisture content is low its differences may have a moderate effect on ignition delay, but a significant influence on the mass loss at ignition (Consalvi et al. 2011). A negative relationship between FMC and FRT can be accounted for by a higher moisture content, which led to lower mass available for combustion at the moment of ignition (Bernard and Nimour 1993; Consalvi et al. 2011). Combustion does not start with the appearance of a flame (Trabaud 1976). Glowing combustion preceding flaming results in a pre-ignition reduction of mass and an increase in temperature of the material (Quintiere 2006). For piloted ignition of a solid to occur, pyrolysis products containing gaseous fuels need to be mixed with air in order to provide a flammable mixture at the igniter (Quintiere 2006; Rich et al. 2007). The presence of moisture in the fuel delays the thermal decomposition and water vapor dilutes the released gaseous fuels leading to both a prolonged ignition delay and a higher temperature of the surface exposed to the heat source at the moment of ignition (Atreya and Abu-Zaid 1991). The diluting effect of water vapor might explain the positive relationship between FMC and IT found in our study and previously reported by numerous authors (e.g., Alessio et al. 2008b; De Lillis et al. 2009; Dimitrakopoulos and Papaioannou 2001).

Specific Leaf Area and Flammability

In our study, an increase in SLA led to an increase in IT and a decrease in FRT. In previous studies, no such relationship was found (de Magalhães and Schwilk 2012; Scarff and Westoby 2006). This might be due to the fact that previous analyses were performed on SLA of fresh leaves originating from the same location (de Magalhães and Schwilk 2012). SLA of fresh material may differ from leaf litter SLA due to nutrient retrieval before abscission, decomposition and leaching (Gill and Moor 1996). In addition, a comparably larger data set distributed over a broader range of SLA values was presented here compared by Scarff and Westo (2006).

SLA can be defined as the reciprocal of the product between leaf thickness and density. These two parameters may vary independently in response to environmental conditions (Witkowski and Lamont 1991). Due to the independent variation of leaf thickness and density, we cannot exclude that the relationship found here is due to alterations in either of these parameters. Unlike SLA, these parameters were shown to influence flammability (Kane et al. 2008; Montgomery and Cheo 1971; Papió and Trabaud 1990). Elucidating this problem further would require examining the influence of all three morphology parameters (SLA, leaf thickness, leaf density) on fire behavior.

Particle Size and Flammability

Our study shows that an increase in AA and AM leads to an increase in ID and lower ignitability. In contrast, numerous studies indicate that an increase in size of leaf litter particles leads to higher flammability (de Magalhães and Schwilk 2012; Engber and Varner 2012; Kane et al. 2008; Scarff and Westoby 2006; Schwilk and Caprio 2011). Due to the complexity of vegetation flammability, it is of great importance to (i) determine all measured parameters and flammability components, (ii) know possible limitations of the method used, and (iii) reflect on the definition of flammability in a particular study, when comparing results of various flammability studies. In experimental studies, which reported a positive relationship between particle size and leaf litter flammability, material was exposed to a flaming ignition source which provided enough energy for the appearance of a sustained flame in the material (de Magalhães and Schwilk 2012; Engber and Varner 2012; Kane et al. 2008; Scarff and Westoby 2006). As the material ignited relatively fast and a high variability in the ignition data was present, this approach is considered inappropriate for measuring the ignitability component of flammability (de Magalhães and Schwilk 2012). Nevertheless, it can provide valuable information on other flammability components and an insight on how vegetation material will behave once sustained flaming is reached. Most of these studies focused on sustainability and combustibility, some of them on consumability, while ignitability parameters were not measured (Engber and Varner 2012; Kane et al. 2008; Scarff and Westoby 2006), or did not relate well to any other measured parameter (de Magalhães and Schwilk 2012). These studies imply that bigger particles result in better aeration of litter beds, leading to higher intensity and faster spread of fire, and thus higher flammability. Nevertheless, if sustainability is taken into account as well, the relationship between particle size and flammability becomes less straightforward as an increase in particle size leads to lower sustainability (de Magalhães and Schwilk 2012).

In our study, we focused on the initial stage of fire development. By applying low heat flux we were able to determine ignitability, but at the same time the prolonged ignition delay distorted information on sustainability (Kauf et al. 2014). Our ignitability results provide information on the tendency of litter materials to reach flaming combustion when they come in contact with a low heat flux ignition source, such as common negligence ignition sources. Better aeration of the litter beds might lead to a higher intensity and a faster fire spread in later stages of fire development. However, if ignition is induced by low heat flux, as in our study, a lower packing ration of bigger particles (Scarff and Westoby 2006), which is in turn leading to a higher heat transfer loss between them (Rothermel 1972), might account for the major part of the relationship between ID and particle size found. Our results, together with those previously published, indicate that particle size can have an opposite effect on different flammability components. An increase in particle size reduces ignitability and sustainability, but at the same time increases the combustability of leaf litter.

Effect of Species, Location, and Sampling Time on Flammability

The significant effect of species on flammability parameters is not surprising in view of the large differences in species morphology characteristics and the relationship between morphology and flammability.

The low overall effect of the location can be accounted for by significant species × location interactions, which show that location had a species-specific effect (Fig. 3). As July samples, which represented only freshly fallen leaf litter, did not show any significant within species and between locations differences, it can be presumed that most of these differences are induced by exposure of litter to different environmental conditions and different rates of leaching and decomposition between locations as well as between species. Nevertheless, two species, P. halepensis and Q. ilex, showed the same order of locations in July as compared to the whole data set. Presumably for these species the lack of a significant effect of location for July samples results from too few replications and low statistical power. The same might hold true for the lack of significant interactions involving sampling time. Though the lack of significant interactions could be interpreted as sampling time having a consistent effect on all the species across all the locations, our data revealed that this was not the case (Fig. 4). The observed overall effect of sampling time on flammability parameters (Table 4) and the results at the sample level (Fig. 4) were unexpected. We expected the highest flammability in July, at the peak of the drought season, and only a gradual change of flammability parameters through the research period. Nevertheless, our results show that flammability parameters of leaf litter change even within a relatively short-time frame. At the current state of knowledge and with the information gathered in this study, we cannot fully explain the mechanisms behind these findings.

July samples were gathered as soon as possible after leaf abscission; they were the least decomposed samples and had the largest AA and AM. On the other hand, decomposition and weathering led to fragmentation of leaf litter particles of May and September/October samples, thereby reducing their area and mass. Taking into account the relationship between particle size and flammability found in our (Table 3) and in previous studies (de Magalhães and Schwilk 2012; Engber and Varner 2012; Kane et al. 2008; Scarff and Westoby 2006; Schwilk and Caprio 2011), it can be presumed that these changes result in an increased ignitability and sustainability with a decreased combustibility and fire intensity. Nevertheless, decomposition also leads to chemical changes (e.g., Aponte et al. 2012; Fioretto et al. 2003; Murphy et al. 1998) and increases the ability of material to retain moisture (Anderson 1990), making the relationship between decomposition and flammability even more complex. Additionally, the fact that leaf litter of P. granatum was too far decomposed for us to find recognizable particles in May implies that decomposition influences flammability not only through changing chemical and physical properties on the particle level, but also by being one of the factors governing the fuel load.

Conclusions

Our work does not create a new management option for fire prone ecosystems, but it provides information for improvement of already existing ones and identifies gaps in knowledge relevant for further improvement and development of management options for these ecosystems. We demonstrated that change in a single leaf litter trait can have an opposite effect on different flammability components. This finding shows that, in order to avoid misinterpretation, the relationship between vegetation traits and fire behavior should address individual flammability components. Furthermore, interpreting trait—flammability relationships in terms of individual flammability components would provide more relevant information for fire managers, as different components have different importance in specific situations. For example, increased combustibility and consumability could be beneficial for belowground organisms as heating up of the soil would be reduced (Gagnon et al. 2010). Nevertheless, an increased combustibility in WUI could reduce the time available for evacuation of endangered areas and increase the total WUI area affected by the fire event, resulting in a higher probability of negative socio-economic consequences. Furthermore, the species-specific response at different sampling locations calls for caution when transferring flammability lists from one area to another. Additionally, the significant effects of sampling time on flammability parameters and the fuel moisture content indicate a change in intrinsic characteristics of the material within the fire season, in contrast to the presumptions made in fire danger systems. Further research investigating the relationship between the time of abscission, the amount of litter produced, the climatic conditions, the decomposition rate, and their influence on flammability related properties of litter beds, such as fuel load, bulk density, moisture retention, and drying, could provide important information not only for fire ecology and management, but also for a better understanding of carbon cycles in general.