Non-additive effects on plant mixtures flammability in a tropical mountain ecosystem

Study area

We conducted plant community sampling and collected plant material for trait measurements in the campos de altitude of Itatiaia National Park, Rio de Janeiro, Brazil (22°22′37″S, 44°42′28″W; 2400 m above sea level). This area features a mean annual temperature of 14.4°C, annual rainfall of 2200 mm and a distinct seasonal climate (Segadas-Vianna and Dau 1965; Safford 1999). It experiences wet summers (December to February) and dry winters (June to August), with winter precipitation under 50 mm per month and temperatures often below freezing. The region has ~200 fog events and 56 frost days annually (Segadas-Vianna and Dau 1965). Campos de altitude is a fire-influenced ecosystem (Hardesty et al. 2005; Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023), with predominantly anthropogenic fires, although lightning-induced fires can occur in the rainy season (Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023). Fires, which range in intensity but are mostly surface fires, occur every 2–15 years, mainly during the dry season (July to mid-October) (Aximoff and Rodrigues 2011; Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023). The vegetation is a mosaic of various life forms, including grasses, sedges, bamboos and small trees, with dominant species such as Cortaderia modesta, Machaerina ensifolia and Chusquea sp. (Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023). These species resprout quickly after fires, and the first two also exhibit fire-induced flowering (Brade 1956).

Field sampling and leaf morphological traits

We selected seven different species (Table 2), including the most frequent: Cortaderia modesta, Machaerina ensifolia and Chusquea pinifolia, as well as other species representing a variety of relative frequencies and life forms. The relative frequency of the species in the study area was measured as described in Matos et al. (2021). In short, we established 12 transects of 100 min at the field site, and for each transect, we laid out 1-m² quadrats at 3 m intervals (300 m² area sampled). Then, we counted all angiosperms rooted inside each plot, plus overhanging leaves or shoots. Relative frequency was then calculated as the number of sampling units where a certain species occurred divided by the total number of sampling units (N = 300). For simplicity, we refer to each species by its generic name (Table 3).

To measure SLA and LDMC, we collected branches (or whole individuals for herbaceous species) of 10 individuals per species and rehydrated them for 2 h (Matos et al. 2021). Next, we sampled five mature leaves (excluding the petiole) per individual to determine the water-saturated fresh mass with a precision scale (0.01 g). Leaf samples were then scanned (at 300 dpi resolution) to obtain leaf area (LA) using Image J software version 1.48. Finally, leaves were oven-dried at 60°C for 72 h to obtain leaf dry mass. We then used LA, fresh and dry mass to obtain SLA (=One-sidedleafareaLeafdrymass) and LDMC (=LeafdrymassLeafwater-saturatedfreshmass).

Flammability laboratory experiments

We conducted different tests to measure LFMCig, MCR and remaining biomass, using separate samples for each one, allowing us to meet the specific objectives of each test. To determine LFMCig, we conducted an ignition test following the methodology described by Cardoso et al. (2018). We could not test Hypochaeris owing to a lack of sufficient individuals in the study area. We collected leaf samples from 100 individuals of each species and rehydrated them for at least 2 h. We then selected subsets of 10 leaf samples from different individuals and weighed them daily. The ignition test involved holding a lighted match to the center of each leaf for 5 s until half of the samples (five) ignited. We manually held the leaves and conducted the ignition in a closed environment within the research station at Itatiaia National Park, ensuring no air currents interfered with the process. After ignition, we placed the samples in an oven to determine their dry mass, and calculated LFMCig (=(Leafmassatignitionpoint−Leafdrymass)Leafdrymass×100).

To determine the MCR, we followed the methodology of Simpson et al. (2016) and Cardoso et al. (2018). We collected branch samples from 10 individuals per species and standardized their initial mass to 9.95–10.5 g, as larger biomass facilitated the measurement of the mass loss rate over time. However, for smaller species (Eryngium, Hypochaeris, Leptostelma), standardization at 10 g was not feasible, resulting in masses ranging from 0.61 to 8.33 g. We dried all samples at 60°C for 3 days before testing. Each sample was then burned on a precision balance, ignited directly with a match and mass loss recorded via video. We conducted the experiment in a fume hood to ensure constant conditions. We recorded mass variation every 0.2 s using Free Video to JPG Converter. We plotted mass variation over time and fitted the resulting sigmoidal curve using the drm function in the drc package (Ritz et al. 2015). The inflection point of the curve indicates when mass loss occurred at the highest rate. Using this reference, we fitted a linear model to 3 s of data on either side of this point, with the slope representing the plant MCR.

We conducted a third test to determine the biomass remaining after combustion (proxy for consumability) of monospecific and mixed-species leaf mixtures. We conducted the burn in a JUNG 1000°C muffle furnace (model 0910). Conducting the experiment inside the furnace provided controlled and consistent conditions for biomass consumption. We dried all samples at 60°C for 3 days before testing. As consumability depends on the biomass ultimately burned, we used dry material to standardize conditions and minimize variability in fuel moisture content over the course of the experiment. To optimize temperature and burning time, we performed a pilot test (Supplementary Fig. S1) with 1 g of leaf material from two species at 300, 400 and 500°C for 30, 60, 120 and 180 min. Our goal was to find a time-temperature combination that distinguished species differences without excessive burning. One of the species used in the pilot study, Baccharis, was not included in the main experiment owing to limited material. The decision to include Baccharis in the pilot does not affect the validity of the results, as the focus of the pilot was on testing the experimental conditions rather than the specific species. We found that 400°C for 30 min best highlighted species differences, whereas 300°C also showed differences but was less realistic for surface fires. Thus, we selected 400 and 30 min as our experimental parameters (see Supplementary Material S1). We assessed consumability with five replicates per species, using 0.9995–1.0005 g of initial biomass. For three-species mixtures, we maintained total biomass within this range, allocating ~0.3332–0.3335 g per species. The decision to use 0.9995–1.0005 g allowed us to include rare species with limited biomass. To ensure consistency, we cut leaves into smaller pieces before burning, as graminoid leaves often exceed 1.0 g. This process was uniformly applied to both single-species and mixed-species samples to standardize biomass, minimize structural bias and ensure comparability between treatments. We placed each replicate in identical aluminium containers (Ø  9 cm, height  2.5 cm, volume  80 mL). We chose three-species combinations to include the three most frequent species (Cortaderia, Machaerina and Chusquea) identified in the Integrated Fire Management Plan of Itatiaia National Park (Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023). In total, we performed 35 three-species combinations with five replicates each.

We also calculated the average contribution for each species in a mixture, following the method of de Magalhães and Schwilk (2012). This calculation represents how closely the consumability of a mixture matches the consumability of each species. We first determined the difference in consumability between a species and all mixtures containing that species. We then calculated the average of these differences to obtain a representative value for each species. This value indicates the influence of each species on the overall consumability of the mixtures. The smaller the absolute value for a given species, the more that species dominates the consumability of the mixtures in which it occurs.

Data analysis

To assess variability in species consumability, we first conducted the Shapiro–Wilk and Levene tests to check for normality and homogeneity of the data. We then performed ANOVA to determine if there were significant differences among species. When significant differences were found, we used the Tukey test for multiple comparisons. For traits that did not meet the assumptions of normality and homogeneity even after Box-Cox transformation, we used the Kruskal–Wallis and Dunn’s tests for multiple comparisons to determine species differences in SLA, LDMC, LFMC, MCR, remaining biomass and Average Consumability Contribution.

To test for the presence of non-additive effects, we employed a paired t-test to compare the expected values for the remaining biomass of the mixtures (i.e. based on the remaining biomass values of their constituent species) with the observed values of the remaining biomass obtained from burning the mixtures (see Fig. 1).

Fig. 1.

Scheme of the consumability experiment for determining expected and observed values of remaining biomass.

We estimated functional diversity for each of the 35 mixtures using functional diversity indices. Specifically, we measured functional richness (FRic), which indicates the volume of trait space occupied and functional divergence (FDiv), which quantifies how trait values diverge within this space. They were calculated using the FRic and FDiv functions from the FD R package (Laliberté and Legendre 2010). To link functional traits with consumability, we ran linear regressions between each trait’s mean value per mixture and the remaining biomass. We also assessed the relationship between functional diversity indices (functional richness, functional divergence) and consumability using linear regressions between these indices and the remaining biomass for each mixture. All analyses were conducted in the R environment (R version 4.3.1, R Core Team 2023).

Leaf morphological and flammability traits

We observed significant interspecific differences in the leaf morphological (SLA, LDMC) and flammability (LFMCig, MCR) traits, remaining biomass and average consumability contribution across the seven studied species (Table 2).

Remaining biomass of monospecific leaf mixtures also varied across species (Table 2, Fig. 2a). For example, Leptostelma and Hypochaeris showed the highest values, whereas Machaerina had the lowest amount of remaining biomass.

Fig. 2.

Results of the consumability experiments and comparison between the observed remaining biomass. Results of the consumability experiments. (a) Remaining biomass (g) of monospecific leaf mixtures for each of the seven species evaluated: Chusquea pinifolia, Cortaderia modesta, Eryngium glaziovianum, Hypochaeris lutea, Leptostelma maximum, Machaerina ensifolia, Pleroma hospita. The green dots in the first boxplot represent the individual data points. (b) Comparison between the observed remaining biomass of the 35 three-species leaf mixtures and the expected values based on the individual constituent species remaining biomass.

Expected vs observed biomass and average species contribution

Overall, as hypothesized (H1), the observed remaining biomass of the three-species mixtures differed from the expected values calculated based on the individual remaining biomass species values (Fig. 2b). The results of the paired t-test indicated a mean difference of 0.0235 g (t = 8.0717, P = 2.70 × 10⁻⁹) between observed and expected values. Additionally, contrary to our hypothesis (H1), in 33 (out of 35) mixtures, the observed mixture remaining biomass was higher than the predictions based on the individual constituent species values and species with high consumability – M. ensifolia (0.0882) and Cortaderia (0.544) – had the lowest average contributions to mixtures’ consumability. The species that exhibited the largest contributions to mixtures’ consumability were those with low and medium consumability – Leptostelma (0.0208), Pleroma (0.0224) and Chusquea (0.0231).

Relationship between functional diversity and consumability

As hypothesized (H2), leaf morphological and flammability traits showed significant relationships with consumability (Fig. 3). Contrary to our expectations in H2a, SLA exhibited the strongest positive relationship with remaining biomass (R² = 0.48, P = 0.02; Fig. 3a), indicating a negative relationship with consumability. According to our hypotheses (H2b and H2d), both LDMC (R² = 0.19, P = 0.03; Fig. 3b) and MCR (R² = 0.33, P = 0.002; Fig. 3d) demonstrated negative relationships with remaining biomass, which translates to a positive relationship with consumability. Regarding functional diversity indices, only FRic displayed a significant negative relationship with remaining biomass, indicating a positive relationship with consumability (Fig. 3e), thus partially supporting our H3.

Fig. 3.

Relationships between mean leaf functional traits per mixture and functional diversity with remaining mixture biomass (g). The statistically significant linear regressions are in green: (a) Specific leaf area (SLA); (b) Leaf dry matter content (LDMC); (c) Live fuel moisture in the point of ignition (LFMCig); (d) Maximum combustion rate (MCR); (e) Functional Richness (FRic); and (f) Functional divergence (FDiv). See supplementary material for the complete statistical results (Supplementary Table S1: intercept, slope, residual s.e., degrees of freedom).

Non-additivity and the influence of low and high-consumability species

As hypothesized (H1), our results indicate non-additivity in consumability when species are mixed, showing that species interactions significantly affect consumability differently from individual species’ properties. Contrary to H1, high-consumability species contributed less to total mixture consumability, contradicting findings from previous studies (de Magalhães and Schwilk 2012; van Altena et al. 2012). Despite the local biodiversity managers’ belief that more abundant species with higher flammability drive fire behavior (Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023), their dominant effect on fire severity may be related to their higher relative abundance rather than non-additive effects. Instead, species with low to medium consumability, such as Leptostelma, Pleroma and Chusquea, played a significant role in the amount of biomass burned. Consequently, the observed consumability of the leaf mixtures was lower than that predicted by the sum of the consumability of each individual species. This result suggests that, although not dominant, these species may play a crucial role in modulating fire severity in the campos de altitude.

The role of traits and functional diversity in biomass consumability

Contrary to our hypothesis, SLA showed a negative relationship with consumability, whereas LDMC and MCR showed a positive relationship. The combination of low SLA and high LDMC means higher leaf densities, suggesting that leaf density is positively related to consumability in our study. This could have happened because leaf density is associated with the number of plant cells and the thickness of their cell walls (Roderick et al. 1999; Grootemaat et al. 2017), and denser tissues likely store a greater calorific content and can reach higher combustion temperatures (Kataki and Konwer 2001). Although several studies have observed a positive relationship between SLA and ignitability (Murray et al. 2013; Grootemaat et al. 2015), it appears that higher combustion temperatures have a stronger impact on consumability than ignition speed.

Another important aspect to consider when examining combinations of leaves mixtures is the formation of the litter layer. In the litter layer, fuel flammability is often influenced by bulk density (the mass of dry fuel per unit volume) or packing ratio (the proportion of fuel volume relative to air volume within the fuel layer; Grootemaat et al. 2017). At very high bulk densities or packing ratios, fire can become oxygen-limited. In these poorly ventilated litter beds, flammability can increase linearly with oxygen availability (Drysdale 1999). Conversely, in well-ventilated litter beds, other factors such as surface area available for combustion and leaf flammability become more critical (Drysdale 1999). As we mentioned earlier, leaf traits play a role in shaping the litter layer. In this way, the negative relationship between SLA and consumability may be explained by how SLA affects bulk density. Lower SLA values seem to increase bulk density (Burton et al. 2021), which then increases the fuel’s consumability. This likely occurs because oxygen was not a limiting factor in these litter beds. In this scenario, similarly to leaf density, higher bulk densities indicate higher caloric content per volume and higher temperatures reached during burning (Kataki and Konwer 2001). As a result, this enhances consumability.

As expected in H3, we found a positive relationship between FRic and consumability. FRic, which measures species functional diversity, reflects the complementarity of species niches in relation to resource use (Hooper et al. 2005; Mason et al. 2005). In the context of fire, higher FRic indicates complementarity in traits such as ignitability and combustibility. Traits like LFMCig and MCR, which influence ignition ease and combustion rate (Pausas et al. 2017), contribute to functional diversity, which promotes more complete biomass consumption. For instance, species that ignite quickly but burn slowly, such as Eryngium, and those that ignite more slowly but burn intensely, like Machaerina, complement each other when burned together, enhancing overall consumability. Species with fast ignition can cause the entire mixture to catch fire quickly, whereas species with intense burning can increase the combustibility of the mixture, leading to higher consumability. Therefore, greater FRic results in greater complementarity of species niches in relation to these traits, making it more likely that mixtures will contain species with different levels of ignitability and combustibility, ultimately increasing their consumability.

However, functional divergence, which measures the dissimilarity among species in trait space, did not show a significant relationship with consumability in our study. This suggests that although species within the community may differ in certain traits, the degree of dissimilarity among species within the trait space did not consistently contribute to variations in how biomass was consumed by fire. This result underscores that it is the complementarity rather than dissimilarity of traits among species that primarily influences consumability in this context.

Integrating species sensitivity to climate change with future fire severity

Campos de altitude are particularly vulnerable to climate change (Assis and Mattos 2016) and increasing numbers of anthropogenic fires, which tend to be more intense as they occur during the dry season (Aximoff and Rodrigues 2011). Therefore, understanding the environmental changes and plant functional traits that influence fire behavior is essential for the long-term conservation of these ecosystems. In a previous study, Matos et al. (2021) analyzed the response of campos de altitude species to drought, integrating the ecological strategies proposed by Grime (1977) (i.e. stress tolerance S, competitors C, and ruderals R) with the physiological strategies of Levitt (1972) (i.e. tolerance, escape, avoidance). They found that Leptostelma, one of the species with a major contribution to leaf mixtures’ consumability, follows a CR-escape/avoidance strategy, which may make it more vulnerable to future climate changes because it lacks mechanisms to tolerate drought. Consequently, as drought intensifies in mountainous areas, the loss of Leptostelma could lead to more intense and severe fires in the campos de altitude. The changes in fire severity due to a decline in Leptostelma abundance are likely to be non-linear and difficult to predict because of the non-additive effects its presence has on the community’s consumability. Other species with high contributions to reducing leaf consumability, such as Pleroma and Chusquea follow a more drought-resistant strategy (S-tolerance/avoidance). Although these three species showed a high contribution to consumability, Pleroma and Chusquea exhibited similar consumability values compared with C. modesta, a species with high consumability. Therefore, despite Leptostelma’s high sensitivity to drought, we recommend conserving this species as a management strategy to modulate fire severity in campos de altitude. This is because Leptostelma was the only species that both significantly contributed to the consumability of mixtures and showed lower consumability compared with high-consumability species like Cortaderia.

Contrary to our expectations (H1), the co-dominant species Cortaderia and Machaerina showed the lowest contribution to consumability. These species, classified as CS-escape/tolerance (Matos et al. 2021), have high consumability, low SLA, high LDMC and high MCR. They can benefit from increased fire frequency, as fire stimulates their flowering (Brade 1956), and their dense clumps may offer protection against high temperatures during fires (Safford 2001). As both drought and fire frequency intensifies, these species may become more abundant, increasing fire severity and reducing the presence of less frequent species, including Leptostelma.

Mechanisms of non-additive effects in ecosystems process

Non-additive effects have been widely studied in leaf litter decomposition experiments, where deviations from expected outcomes are attributed to interactions among species traits (Dias et al. 2013; Finerty et al. 2016). One possible explanation is the improved microenvironmental condition theory (Makkonen et al. 2013), which suggests that certain species, depending on their traits, create microclimatic conditions that change the decomposition of others. For example, plant litter traits that improve conditions for decomposers can promote the breakdown of co-occurring litter species (Makkonen et al. 2013). In fire ecology, non-additive effects may arise from species traits that alter the microenvironment within the litter layer. In our study, species with high SLA and low LDMC – traits associated with lower leaf density and possibly lower bulk density (Burton et al. 2021) – and low MCR decreased consumability of mixtures. This was likely because oxygen was not limiting, and the less dense litter reduced heat retention and calorific content per unit volume, creating unfavorable conditions for whole biomass consumption by fire. As a result, these interactions led to consumability values in mixtures lower than expected based on individual species values.

Our experimental set-up followed a trait-based approach, one focus being to assess how species’ traits influence ecological processes (Rosado et al. 2013). Controlled experiments are crucial for reliably comparing species while minimizing external variability (Pérez-Harguindeguy et al. 2013). In our study, heat exposure was controlled by burning leaf material in a muffle furnace, and small cut leaf samples were used to standardize species proportions, reflecting aspects of litter fuels. Although the role of leaf litter in campos de altitude fire dynamics remains unclear, its influence cannot be disregarded. However, this approach does not fully replicate natural fuel conditions, as the herbaceous layer, which often burns as whole individuals, is another key fuel component (Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) 2023). Leaf traits and flammability measurements have been criticized for poorly predicting field fire behavior, as they often neglect plant architecture, a critical factor influencing ecosystem flammability (Fernandes and Cruz 2012). Nonetheless, leaf traits like LDMC have been shown to exert stronger effects on shoot flammability than architectural traits (Alam et al. 2024). This underscores the relevance of trait-based approaches based on plant organs, where leaf traits can serve as robust proxies to predict fire behavior. By assessing non-additive effects, our study sheds light on species interactions and their role in fire severity, a critical but understudied aspect of campos de altitude.