A Global Perspective of the Functional Trait Responses of Graminoids to the Seasonality of Fire

Abstract

In fire-prone ecosystems, plant resilience to recurrent fires depends on certain fire-adaptive traits. However, how key functional and reproductive traits of graminoids respond to varying burning seasons is poorly understood. This meta-analysis, therefore, unpacks global perspectives on how resprouting, growth, reproductive, and productivity traits of graminoids (grasses and sedges) respond to different burning seasons. We recorded 569 observations from 80 experimental studies comparing graminoid plant trait responses in unburned vs. burned treatments over different seasons of burn. Weighted log response ratios and 95% confidence intervals (95%CI) were analyzed for each plant trait using random effects models and compared across burning seasons. Summer (0.35 (95%CI = 0.25 to 0.46)) and autumn burns (0.24 (95%CI = 0.16 to 0.31)) increased above-ground biomass m⁻², while biomass plant⁻¹ was increased only by spring burns (0.27 (95%CI = 0.22 to 0.32)). Bud production plant⁻¹ and tiller⁻¹ were reduced significantly by fire, especially spring, summer, and autumn burns. The shoot height (0.29 (95%CI = 0.17 to 0.41)), leaf length (0.15 (95%CI = 0.11 to 0.20)), and specific leaf area (0.08 (95%CI = 0.06 to 0.09)) increased only under summer burns, while flowering was enhanced by spring (0.19 (95%CI = 0.00 to 0.38)) and autumn burns [0.34 (95%CI = 0.02 to 0.66)]. However, seed production m⁻² was reduced by spring and summer burns and the opposite was true for seed production plant⁻¹. Overall, herbaceous plant trait responses to fire varied by the season of burn, disagreeing with the general principle that early spring burning is the best practice. We suggest that a decision on the season of burn should be informed by the objective of burning.

1. Introduction

Fire is an integral part of grasslands and woodlands and is key in preventing state shift to a bush encroached state and in increasing diversity and plant quality and productivity [1,2,3,4,5,6]. However, the effects of fire on vegetation are complex and vary both inter- and intra-specifically, depending on plant adaptive traits and the fire regime [7,8,9]. Certain ecosystems co-evolved with fire; hence, their structure, functioning, and services are sustained by positive feedback between vegetation and fire [6,10,11]. This feedback emanates from the fact that certain fire regimes select for certain plant adaptive traits [11,12,13].

Fynbos shrublands of South Africa, Chaparral shrublands and prairies of America, and Cerrado grasslands of Brazil are typical ecosystems in which vegetation employs specialized strategies to adapt to recurrent fires [14]. Plants respond to fire mainly by seed dormancy release [15], resprouting [7], and fire-induced synchronous flowering [16,17]. For instance, summer fires increased the flowering of grasses in European [18] and North American grasslands [17], whereas spring fires promoted bunchgrass resprouting in African [19] and Australian ecosystems [20]. Resprouting and recruitment from the seed, respectively, depend on bud and seed bank sizes [19], and these post-fire recovery strategies promote plant population persistence [11,19]. The ability to recover clonally following fires depends largely on whether plant buds are located below or above ground [19]. Generally, plant buds that escape fire, more so below-ground plant buds, remain dormant and resprout when conditions are suitable following fires [21]. However, in other ecosystems, non-prescribed fires may be detrimental to fire-sensitive species that lack certain fire-adaptive traits [22,23]. A worst-case scenario is the extinction of these species, followed by the ingress of invaders, which exacerbates biodiversity loss [24].

Graminoids, largely bunchgrasses, are amongst the most vulnerable growth forms to fires, due to the projection of their meristems to fire flame heights [11,21]. As a result, most bunch-type graminoids are obligate seeders, capitalizing largely on high flowering [16] and en masse recruitment from the seed following fires [2,25]. Seeders allocate more resources to rapid growth to reach the reproductive stage prior to the next fire event, whereas clonal graminoids invest more to non-structural carbohydrate reserves and protective structures [19]. In some instances, light fires may reduce residual senesced fuel on grass tussocks, which increases light availability, thereby triggering the recruitment of new buds [26].

Plant resilience to fire, however, is not exclusively dependent on plant phylogeny, but also on fire regime, including, amongst others, timing of burning [18]. The season of burn is crucial, as it affects the growth and reproductive cycles of graminoids, with consequential effects on plant persistence [27]. For example, burning when the meristematic tissues are elevated, and carbohydrate reserves are depleted during active growth hinders plant performance, survival, and persistence. Furthermore, burning during flowering has direct effects on plant fecundity, as seed production depends largely on the number of floral stalks [18]. These effects are modulated by seasonal variation in plant physiological and phenological changes and climatic conditions through influencing fuel load characteristics, which in turn influence the intensity of fire [27].

A global rule of thumb is that burning in spring is the best practice [28]. However, this fire management prescription appears to be informed by community-scale responses without detailed understanding of its implications at a plant trait level. This limits our understanding of the mechanisms and adaptive traits underlying plant recovery and persistence following fires. Plant trait responses to fire may provide insights into plant performance, productivity, and reproductive capacity and help derive plant conservation strategies [27]. Furthermore, the future climate, characterized by recurrent droughts, is projected to shift burning seasons and frequencies, with fires likely to occur within short frequencies over extended seasons of burn. Thus, understanding plant responses to different burning seasons at a global scale form the basis for understanding resilience strategies of herbaceous plant communities to future climate-driven shifts in fire regime. A comprehensive assessment of plant trait responses to different burning seasons is crucial to unpack how plants recover and persist following fires of varying seasons of burn. This global meta-analysis was conducted to assess how growth, productivity, and reproductive traits of graminoids respond to varying seasons of burn.

2. Materials and Methods

2.1. Data Collection

Using advanced search function, we searched peer-reviewed articles reporting effects of fire on functional traits of graminoids (grasses and sedges) from Web of Science, Scopus, Google Scholar, and BioOne Complete. The search was restricted only to studies published from 1985 to 2022. The following search terms and combinations were used: effect of “fire” or “burning” or “season of burn” on (1) grasses, (2) sedges, (3) graminoids, (4) functional traits, (5) tiller or bud production, (6) resprouting, (7) flowering or seed production or plant fecundity, (8) plant growth or height, and (9) biomass production.

For each study, we read the title to assess the relevance and thoroughly screened the abstracts to determine whether observations were conducted in burn vs. unburned scenario. Studies were scrutinized to meet the following criteria: (1) the study reports the effect of fire on graminoids (grasses and sedges), and (2) the following were provided: vegetation type (e.g., grassland, savanna, shrubland, etc.), season of burn, and sampling dates, and (3) the comparisons were conducted between burned and unburned scenario either in a form of Control-Impact (CIT), Before-After-Impact (BAIT) and or Before-After-Control-Impact (BACIT) experimental frameworks. The CIT framework refers to comparisons between burned and unburned treatments, whereas BAIT scenario refers to comparisons before and after burning [29]. The BACIT scenario refers to comparisons between burn treatment and control before and after fire application [30].

The response variables assessed were shoot and root biomass (on g m⁻² and g plant⁻¹ basis), bud production [on per area (m²), plant and tiller basis], plant height (cm), leaf length (cm), specific leaf area (g cm⁻¹) and seed production (on seeds m⁻², plant⁻¹, and stalk⁻¹ basis). In scenarios where fire was applied in combination with other treatments, e.g., grazing, shrub control, etc., we selected a control which included fire only. Moreover, the historic background of the burned and unburned treatments was scrutinized and the studies in which previous management histories of the two treatments differed were rejected. In the case of BAIT and BACIT experimental frameworks, we selected mainly the studies in which covariates, e.g., rainfall, temperature, etc., were included in the statistical models and reported to have had no additive or confounding effect. For the studies in which repeated measures were conducted, each observation was considered as an independent study. In each study, we recorded the reference, longitude and latitudes, species name, plant functional type (grass or sedge) and season of burn. For studies that did not report coordinates, we searched them by name of the study area and country using coordinate converter (https://coordinates-converter.com; Accessed on 20 July 2023). In cases where the specific name of a study area was not reported, for example when reported as “the site was located between area A and B”, we recorded the coordinates of the area mentioned first (A).

The statistical measures recorded were mean ($\overline{\mathrm{x}}$), standard deviation (SD) and the sample size (n) for both burned and unburned scenarios. The statistical measures were extracted from tables or through digitizing figures using WebPlot digitizer 4.5 (https://github.com/ankitrohatgi/WebPlotDigitizer; Accessed on 20 July 2020).

We obtained a total of 416 experimental studies, of which 95 lacked relevance; subsequently, they were disqualified. The remaining 321 studies were assessed for eligibility; 241 were disqualified for reasons of content and 80 studies reporting 569 observations qualified for this meta-analysis. Most studies (36%) were conducted in North America, whereas 20 and 23% of studies were conducted in South America and Africa, respectively (Figure 1). Of these studies, 50% were conducted in grasslands (Figure 1).

2.2. Meta-Analytical Procedures

Firstly, when the standard error of mean (SEM) was reported instead of SD, we converted it to SD using the following equation:

$$\mathrm{SD}=\mathrm{SEM}\times \sqrt{\mathrm{n} }$$

(1)

In cases where $\overline{\mathrm{x}}$ was reported with neither SD nor SEM, we computed the SD by dividing by 10. If 95% confidence interval (CI) was reported alongside with $\overline{\mathrm{x}}$ instead of SD or SEM, the following equation was employed to compute SD:

$$\mathrm{SD}=\mathrm{CI}\frac{\sqrt{\mathrm{n}}}{1.96}$$

(2)

The random effect models (REMs) were used to compute weighted log response ratios (logRR) and their 95% confidence intervals. REM was selected for its high parsimony relative to fixed effects models, especially when between-study variance is expected to be high, due probably to differences in plant functional types, vegetation types, fire characteristics and area-specific climatic conditions [27]. The logRRs were computed according to the following equation and displayed in a forest plot together with their weightings:

$$\mathrm{logRR}=\log \frac{{\overline{\mathrm{x}}}_{\mathrm{b}}}{{\overline{\mathrm{x}}}_{\mathrm{ub}}}$$

(3)

where ${\overline{\mathrm{x}}}_{\mathrm{b}}$ and ${\overline{\mathrm{x}}}_{\mathrm{ub}}$ are mean values of the response variables in burned and unburned treatments, respectively.

The significant effect of the season of burn on plant traits was noted if the CI did not overlap with zero, whereas the significant differences between seasons of burn were noted if the CI of the two seasons did not overlap [31]. We generated 1000 bootstrap permutations to determine the intercept (model estimate) of the REM and used a Z-test to assess the significance of the REM intercept at α = 5%. The model fit was assessed using a restricted maximum likelihood (REML), log likelihood specifically. Due to assumed variability in study variances, the logRRs were weighted using the inverse of the variance as follows:

$${\mathrm{V}}_{\mathrm{i}}\left(\mathrm{logRR}\right)=\frac{{\mathrm{SD}}_{\mathrm{b}}{}^2}{{\mathrm{n}}_{\mathrm{b}}{\overline{\mathrm{x}}}_{\mathrm{b}}{}^2}+\frac{{\mathrm{SD}}_{\mathrm{ub}}{}^2}{{\mathrm{n}}_{\mathrm{ub}}{\overline{\mathrm{x}}}_{\mathrm{ub}}{}^2}$$

(4)

where SD_b and SD_ub represent the standard deviation of burned and unburned treatments, respectively. The n_b and n_ub are sample sizes of burned and unburned treatments, respectively. All the analysis was conducted using the package MAJOR in Jamovi 2.3, which is a graphical interface of R.

2.3. Detection of Study Bias

A general assumption is that studies reporting significant differences between treated vs. untreated scenarios are more likely to be published than those reporting marginal differences. For this scenario, we conducted meta-regression between sample sizes and response ratios using tau as an estimator, with a significant tau value at 95% confidence level indicating bias. We further assessed between-study variance using Q-test and a Fail-Safe N value and complemented them with funnel plots of sample sizes vs. logRRs for visualization of data asymmetry. After detecting bias, studentized residuals and Cook’s distance measure were used to identify the potential outlier studies and to check that no study overly influenced the logRR. The sensitivity analysis using trim and fill method was conducted to ascertain what the effect sizes would be if missing studies were included in the meta-analysis.

3. Results

3.1. Above- and Below-Ground Biomass

The overall plant responses to fire in terms of above-ground biomass per area (Z = 0.89, p = 0.375) and per plant (Z = −1.33, p = 0.182) was not significant (Figure 2a–c). However, above-ground biomass (AGB) varied by the season of burn, with AGB per area exhibiting a significant (p < 0.05) increase for summer and autumn burns compared to spring and winter burns (Figure 2a). In the latter two seasons of burn, AGB was negatively affected by fire, declining significantly by effect sizes of −0.04 (CI = −0.13; 0.05) and −0.17 (CI = −0.39; 0.04), respectively (Figure 2a). Similarly, AGB per plant varied by the season of burn, responding positively to spring burns, while it responded negatively to other seasons of burn, more so to summer burns (−1.95 (CI = −2.26; −1.65); Figure 2b). The overall response of root biomass (RB) to fire was not significant (Z = 1.66, p = 0.096), but RB varied by the season of burn (Figure 2c). The spring burns increased RB four-fold higher than autumn and summer burns, whereas winter burns reduced RB by an effect size of −0.06 (CI = 0.40; 0.29; Figure 2c).

3.2. Bud Production

The overall bud production per area showed no significant (Z = 0.613, p = 0.540) responses to fire. However, bud production per plant (Z = −2.35, p = 0.019) and per tiller (Z = −2.44, p = 0.015) showed significant responses to fire, with bud production responses varying by the season of burn (Figure 3b,c). Bud production per area increased under spring burns and decreased under summer and autumn burns (Figure 3a). For both per plant and tiller, bud production was negatively affected by fire, with summer burns causing more lethal effects [−1.02 (CI = −1.23; −0.88)] than autumn [−0.55 (CI = −0.68; −0.42)] and winter burns (0.11 (−0.24; 0.46); Figure 3c). Bud production per tiller was reduced significantly by fire, with more significant declines being evident under autumn burns than in spring burns (Figure 3c).

3.3. Height, Leaf Length, and Specific Leaf Area

The overall response of shoot height (SH), leaf length (LL), and specific leaf area (SLA) to fire was insignificant (p > 0.05), but these parameters varied by the season of burn (Figure 4a–c). The SH (0.29 (CI = 0.17; 0.41)) and LL (0.15 (CI = 0.11; 0.20)) were enhanced significantly only by summer burns (Figure 4a,b). Specifically, spring burns had remarkable negative effects on SH and LL, but LL was reduced more by autumn burns (Figure 4a,b). The SLA was marginally increased by an effect size of 0.08 (CI = 0.06; 0.09) by summer burns, whereas autumn and winter burns reduced it by −0.05 (CI = −0.07; 0.02) and −0.28 (CI = −1.19; 0.63), respectively (Figure 4c).

3.4. Reproductive Traits

The fire had a significant effect on flowering (Z = 2.85, p = 0.004), with spring and autumn burns significantly increasing flowering relative to winter burns, which reduced flowering by a magnitude of −0.17 (CI = −0.93; 0.60; Figure 5). The overall response of seed production per area (Z = −0.782, p = 0.434) and per plant (Z = 1.30, p = 0.194) to fire was not significant (Figure 6a,b). However, seed production on a per area, plant, and stalk basis varied remarkably across seasons of burn (Figure 6a–c). The seed production per area declined significantly in summer and spring burns and the opposite was true for autumn and winter burns (Figure 6a). Conversely, spring and summer burns increased seed production per plant compared to autumn and winter burns (Figure 6b).

4. Discussion

4.1. Biomass and Growth of Graminoids Depends on Burning Season

The effect of the season of burn on above-ground biomass (AGB) depended largely on the scale of measurement, with spring burns reducing biomass per area and increasing biomass per plant. This suggests that spring burns are detrimental at a stand level, probably due to fire eliminating fire-sensitive species, thereby reducing their contribution to stand-level biomass. Generally, spring burns are applied after the first rains [28], a time coinciding with seedling recruitment and early plant establishment. Plant seedlings are more susceptible to disturbance at this stage and are less likely to survive fire [32]. Hence, burning in summer and autumn increased AGB per area because during these burning seasons, plants are well developed and the high moisture content in their phytomass insulates them against fire. Moreover, the nutrients from the ash produced during summer burn are readily available and accessible by plants because the root system is well-developed, and the soil moisture is adequate to solubilize these nutrients [4]. Thus, increased nutrient uptake from the ash and high soil moisture content in summer compensate the plants through early post-fire recovery [19]. Such a compensatory response could be justified by the increase in shoot height and leaf length observed under summer burns relative to spring burns (Figure 4a,b). The increase in SLA under summer burns, particularly, could be attributable to higher growth rates, as indicated by increased plant height relative to other seasons of burn (Figure 4a,c). Nonetheless, AGB per plant was increased by spring burns and reduced by summer and autumn burns, regardless of increased leaf length and stem growth under summer burns. This could be explained by the reduction in AGB per area under spring burns (Figure 2b), which probably reduced plant competition, in turn increasing resource availability to fire survivors [18]. The increase in AGB per plant for spring burns, however, appears to be caused neither by tillering nor by enhanced growth because bud production, plant growth and leaf length were reduced by spring burns (Figure 3 and Figure 4a). Thus, enhancement of AGB per plant by spring burns could be ascribed to other factors not considered in this study, e.g., leaf production.

The results further showed that root biomass (RB) was enhanced by all burning seasons except winter burns (Figure 2c). Plant dormancy, which increases fuel load together with low soil and fuel moisture facilitates intense winter fires that cause severe soil heating and combustion of plant crowns, leading to the death of plant roots [4]. The roots of graminoids are distributed shallower close to the soil surface; thus, they are more susceptible to intense fires [4,21]. This is more common where ground fires are prevalent as they burn below soil surface and cause severe soil heating [28]. Conversely, RB was enhanced by spring, summer, and autumn burns (Figure 2c), indicating that reduced AGB per plant in the latter two burning seasons is compensated for by below-ground biomass investment. Hence, one would expect higher post-fire plant recovery under these burning seasons relative to winter burns. High RB enhances water and nutrient uptake [4], of which both enhance plant recovery following fires.

4.2. Burning during Active Growth Reduces Bud Production of Graminoids

In agreement with previous experimental studies, we found that fire reduces bud production, especially spring, summer, and autumn burns. During these seasons, plant growth is still active, and fire becomes lethal to plant buds because the meristematic tissues are elevated to fire flames. Moreover, carbohydrates are also depleted during active growth, as they are channeled to photosynthesis, and this limits bud regeneration after fires [28]. Generally, the reduction in the most active growth points such as intercalary meristems due to burning fosters plant reliance on axillary buds for photosynthetic regeneration [33]. Since buds are highly sensitive to heat, cool fire from senesced annual plants is adequate to inhibit bud emergence [34], but perhaps not enough to remove apical dominance. Since the fuel load in spring, summer, and autumn is not as abundant as in winter, it is presumable that incomplete combustion due to high foliar moisture prevented the removal of apical dominance, thereby limiting the regeneration of axillary buds. This is explained by positive responses of bud production under winter burns relative to other burning seasons. The reduction in bud production by summer and autumn burns was also reported by Russell and Vermeire [35]. Reduction in bud production may exacerbate loss of plant vigor and population persistence, especially in the ecosystems dominated by perennials, which rely largely on clonal regeneration [36]. Our results suggest that herbaceous plant recovery depends largely on seed production rather than vegetative bud regeneration following spring, summer, and autumn fires.

4.3. Burning Enhances Flowering and Seed Production of Graminoids

Our results showed that burning facilitates flowering in graminoids, more so in spring, summer, and autumn burns (Figure 5). Nonetheless, when these different burning seasons were compared, there was no obvious difference on their effects, suggesting that fire stimulates flowering regardless of the burning season. This response is not surprising, as other studies, largely those conducted in grasslands, reported synchronized flowering in grasses after fires regardless of the timing of burn [16,37]. However, this effect appears not to translate to high seed production per area for spring and summer burns (Figure 6a), but rather increased seed production per plant in these seasons (Figure 6b). The increase in seed production per plant was expected given that the number of seeds correlates with the number of inflorescences produced [27]. Generally, there is a trade-off between seed production and resprouting [38]; hence, we recorded higher seed production per plant in spring and summer burns despite reduced bud production per plant. This is because plants, particularly seeders, invest more resources in fast growth and seed production instead of vegetative resprouting [19]. This meta-analysis suggests that, globally, seed production is the main persistence strategy employed by graminoids to recover from fires. Generally, this strategy maintains plant populations via two important ways, including forming large seed bank sizes that remain dormant if conditions are not suitable for recruitment [39]. Secondly, the seeds of these plants may be easily dispersed in safe sites for germination [39]. The effect of summer fires on flowering might depend on the time of fire ignition within the summer season, as the effect of fire before floral initiation may not be as severe as during floral initiation. It appears, however, that increase in flowering under summer burns is a common plant response triggered by chemical compounds of smoke and ash together with temperatures induced by fire [18,40]. Flowering is not only dependent on fire-induced cues, but also on soil nutrients and moisture [18,41,42,43], which are more abundant and easily accessible by plants in summer.

4.4. Limitations of the Study and Future Research Directions

This meta-analysis assessed plant trait responses to different seasons of burning. Generally, a fire regime includes various factors which could interact at large landscape scale with the season of burn, e.g., fire return intervals [44]. Thus, the assessment of functional trait responses to burning season alone may limit broad understanding of plant response dynamics under scenarios where a wide spectrum of fire regime factors interact. This knowledge is pivotal, given that future climate change is expected to cause changes not only in the season of burn, but also the frequency with which fires will occur [45]. This information is fundamental to creating awareness about climate-driven shifts in fire regime and to derive future coping mechanisms for these fire regime shifts for land users, particularly conservationists and livestock farmers. Thus, experimental research that incorporates burning season and fire return intervals could give a better perspective on recovery mechanisms and strategies utilized by plants to persist under recurrent fires. This could also consider the fact that plants with varying life histories and phylogenies respond interspecifically to varying fire regimes [19]. Thus, considering responses across plants of different photosynthetic pathways, growth habits, and lifespans may provide clear prospects on how ecosystems with varying plant functional traits will be shaped by recurrent fires of varying regimes.

5. Conclusions

The current meta-analysis indicated that biomass and bud production responses to fire depend on the season of burn. Biomass production varied considerably across seasons of burn, depending also on the scale of measurement, with spring burns reducing biomass per area, while it increased biomass per plant. Almost all fire seasons had negative effects on bud production, whereas they enhanced the flowering of graminoids, suggesting that sexual regeneration is a significant persistence strategy for recovery from fires. In a broader context, this meta-analysis revealed that global ecosystems dominated by graminoids respond differently to different burning seasons depending on the functional plant trait of concern. We, therefore, deduce that the general principle that spring is the best burning season is subjective and that the decision on burning season should be informed by the objective of burning. The results of this study are fundamental for understanding regeneration and persistence strategies of graminoids following fire for designing appropriate fire management prescriptions in respect to the season of burning.