Fire severity and prolonged drought do not interact to reduce plant regeneration capacity but alter community composition in a Mediterranean shrubland

Background Disturbance-regime shifts are often a manifestation related to climate change. In Mediterranean eco-systems, summer-drought lengthening and high fire-severity may be among the most detrimental processes for plant recovery capacity. However, although isolated effects have been usually assessed, the effects when both disturbances interact are less known. This paper examined the possible interactive impacts of increased fire severity and summer-drought lengthening on Mediterranean vegetation recovery. Our initial hypothesis maintained that both disturbances would interact and cause greater recovery damage than in an isolated way. For this reason, we performed an experimental fire in summer by creating two fire severity levels: control fire severity (CSev) and increased fire severity by adding dry biomass (IncrSev). Following fire, and using rainfall exclusions roofs, we extended summer drought conditions toward the first autumn after fire (AutExcl treatment) and toward the first post-fire spring (SprExcl). All the treatment-levels combinations were replicated in five 0.5 × 0.5 m plots. Results Emerged seedlings were not affected by treatments, but plant establishment was significantly impaired by extended droughts at the end of the first post-fire year, particularly for Cistaceae and subshrubs. Nevertheless, we found no effects of increased fire severity on either isolation or combination with drought. Notwithstanding, the combination of some treatment levels brought about changes in plant composition. These changes were driven mainly by the detrimental effects of treatments on perennial forbs. This functional group was affected by treatments, which suggests that they may be more sensitive to changes in fire severity and severe droughts. Conclusion Increased fire severity might not affect plant recovery either by itself or by interacting with drought because prolonged drought may mask increased fire severity impact on Mediterranean seeding species. However, fire-severity increases, together with sporadic drought events in the early stages of these communities, could imply long-lasting changes in community composition due to distinct functional-groups sensitivities. Neverthe-less, these impacts depend on the considered species or functional group. These findings provide information about the impacts that Mediterranean-shrublands ecosystems might face if the trends of fire and drought regimes continue shifting.


Introduction
The current state of many ecosystems worldwide is largely the result of different disturbance influences and their interactions (White and Jentsch 2001).Abiotic disturbances may play as an important role as biological interactions do (Sousa 1984) not only by mediating in community-assembly processes (Chase 2007) but by also exerting evolutionary pressure on functional traits (Canham and Marks 1985).In recent decades, intensified human activity has increased the frequency and magnitude of disturbances, which have led to the emergence of amplified individual and compound disturbances with unknown ecological impacts worldwide (Turner 2010;Buma 2015;Seidl et al. 2017).In terrestrial ecosystems, some disturbances like droughts or wildfires have been seen to alter forest productivity or to significantly amplify the ecological impacts of other disturbances (Seidl et al. 2017;Nolan et al. 2021;Yang et al. 2023).It is, therefore, most interesting to understand to what extent ecosystems can withstand current and future disturbance interactions (Johnstone et al. 2016).
Drought is a disturbance that has shaped plant traits and community assemblage in Mediterranean ecosystems along millennia (Nardini et al. 2014;Batllori et al. 2019;Peguero-Pina et al. 2020).Nevertheless, the present anthropogenic alteration of climate is changing drought regimes to longer dry seasons (Seneviratne et al. 2021;Williams et al. 2022) or higher drought intensities (Dai 2013).These drought regime changes are in accordance with already made climate projections (Vicente-Serrano et al. 2014;Giorgi and Lionello 2008;Hoerling et al. 2012), which also suggest higher drought intensities and more consecutive dry days (Vicente-Serrano et al. 2022;Satoh et al. 2022).These water shortages may impact at different ecosystem levels from plant metabolic pathways to seedling establishment success, community composition, or biogeochemical cycles (Peñuelas et al. 2018).Moreover, drought after disturbances could
severely undermine seedling establishment and subsequently cause long-term effects on ecosystems (Karavani et al. 2018).For example, successional trajectories and ecosystem functioning in the long-term could be affected because they depend on drought-regime changes (Kreyling et al. 2011;Ingrisch et al 2022).Therefore, disentangling specific plant responses to water scarcity is crucial seeing that higher drought intensities might lead to abrupt disruptions in ecosystem resilience (Ingrisch et al. 2022).
Similarly to drought, fire has also shaped Mediterranean ecosystems as a disturbance force by selecting adaptive traits in most Mediterranean plants to persist (Keeley et al. 2011;Rundel et al. 2018).This persistence can, however, be threatened if fire regimes shift largely beyond common ranges (Baeza et al. 2002;Benayas et al. 2007;Enright et al. 2015;Karavani et al. 2018).For example, variations in fire severity, a fire-regime component, is often positively related to the magnitude of ecosystem change (Keeley 2009;Mitsopoulos et al. 2019).When this change is sharp (i.e., very high severity fires), the regeneration of affected communities can be compromised (García-Llamas et al. 2019;Nolan et al. 2021).One of the most important drivers of fire-severity increase is fuel load (Alvarez et al. 2012;Lecina-Díaz et al. 2014;García-Llamas et al. 2019), which has considerably increased in many fire-prone ecosystems as a consequence of fire-suppression policies or land abandonment (Benayas et al. 2007;Jones et al. 2022).Moreover, harsh prefire drought conditions may also influence fire characteristics because they can affect the likelihood of occurrence, extension, and severity throughout different processes.This can happen through an increase in litter load as a consequence of crown defoliation (Keane 2008), through a drought-induced forest dieback (Buma 2015;Holden et al. 2018;Ruthrof et al. 2016;Cardil et al. 2019) or through decreasing fuel moisture towards critical flammability levels (Boer et al. 2017).This could account for not only an increase in fires that burn with high severity (Parks and Abatzoglou 2020) but also postfire recovery because severe fires can impact nutrients pools or seed banks (i.e., material legacies; Johnstone et al. 2016).Specifically, longer exposure times to heat released by fire have been shown to undermine the viability of the seeds stored in both canopy and soil banks as well as species composition (Moreno and Oechel 1991;Fernández-García et al. 2019;Shi et al. 2022;Aguayo-Villalba et al. 2021).Similarly, high values of fire-release heat have been seen to damage meristematic tissues by impairing the recovery of resprouting species (Lloret and López-Soria 1993).Therefore, high fire severities could bring about changes in the spatial distribution and density of seedlings after fire (González-De Vega et al. 2016;Rodríguez-García et al. 2022).In the long term, although Francos et al. (2016) did not obtain effects in species dominance with different fire severities, other authors have found that high fire severities can compromise the resilience of shrubland ecosystems (Huerta et al. 2022).Thus, as no signs of fire-severity drawdowns are expected in present and future scenarios, the comprehension of their effects on ecosystems should be considered not only during fire but also in the recovery stage.
Different disturbance types do not, however, occur and impact the ecosystem in an isolated way because they can happen at the same time or shortly after another one.For instance, drought can exert a large amplifying effect on other disturbances (Seidl et al. 2017).Hence, its effects on Mediterranean ecosystems can be exacerbated if paired with a fire event (Batllori et al. 2019;Dewees et al. 2022).For example, when summer drought extends toward the next wet seasons after fire, it can significantly impact emergence, survival, and establishment by amplifying its impacts on species with seed-based regeneration (i.e., seeders; Moreno et al. 2011;Salesa et al. 2022).The direct and indirect effects of fire (Santana et al. 2013), together with the rainfall season after summer, are crucial constraints on which the establishment of these species depends.For this reason, the longer summer seasons trends of recent decades (Seneviratne et al. 2021) can impair the recovery process during the first autumn post-fire, a fundamental cohort in the recovery process (De Luis et al. 2008;Moreno et al. 2011;Karavani et al. 2018;Salesa et al. 2022).However, not only seeders are impaired by drought lengthening.For instance, drought events in the months preceding and following fire might impact the regeneration capacity of species that resprout after disturbances (Pratt et al. 2014;Nolan et al. 2021).Combinations of different disturbances and intensities could trigger complex overlapping effects on plant communities' response (Bowman et al. 2017;Nolan et al. 2021).Therefore, in the light of the expected increases in fire severity and longer droughts, which may exacerbate impacts and disrupt Mediterranean resilience feedbacks (Karavani et al. 2018;Batllori et al. 2019), further efforts are needed to disentangle the possible compound effects of drought and fire regime shifts on the regeneration process.
With this work, we intended to understand the impacts caused by the interaction between different fire severity and summer-drought lengthening levels during the postfire recovery of Mediterranean shrublands.To assess this effect, we conducted a prolonged summer drought experiment with three different rainfall exclusion treatments after an experimental fire.We specifically and experimentally created a soil water shortage in (i) the first autumn and (ii) the spring after fire as well as a (iii) third treatment without drought-induced treatment (control).Apart from control severity, we generated an increased fire severity level by adding dry biomass.These drought and severity treatments were crossed within a complete factorial design.We hypothesized that an increase in fire severity and dry conditions would impact negatively and separately vegetation regeneration in plant population abundance and community diversity terms.We also expected the combination of both factors to interact and enhance negative effects on plant response's capacity, which would lead to significant community regeneration failure.

Study area and experimental design
The study area is located in Ayora (39° 07′ N, 0° 57′ W), in the eastern Iberian Peninsula (Valencia region).This area is 1050 m.a.s.l. and is characterized by a Csa climate (Kottek et al. 2006).The average annual temperature and annual average rainfall in this area is 15.8 °C and 537 mm, respectively, and autumn is the wettest season.Here, vegetation is a shrubland dominated by Cistus albidus L., Ulex parviflorus Pourr., and Salvia rosmarinus Schleid.The grass Brachypodium retusum P. Beauv. is the main species found in the herbaceous layer.The study area was the regenerated community after an experimental fire was carried out 9 years before.
In summer 2016, a new experimental fire was performed in a 1400 m 2 plot (35 × 40 m).This plot had a fuel load of 1385 ± 337 g m −2 (mean ± SD) considering both vegetation and litter.The rate of spread of experimental fire was 0.23 ± 0.20 m s −1 with a fire line intensity of 4548 kW m −1 .In average, there was a fuel consumption of 1014 ± 204 g m −2 .See Table S1 for more details of biomass content, soil and biomass moisture, fire behavior, and meteorological conditions during our experimental fire.Before this fire, thirty 0.25-m 2 subplots were randomly distributed in the plot.Two different treatments were applied to these subplots within a two-factor factorial design with all the possible levels combinations.These factors were (i) drought with three levels (Control; Autumn Exclusion, AutExcl; and Spring Exclusion, SprExcl) and (ii) fire severity with two levels (Increased Severity, IncrSev; and Control Severity, CSev).Every combination involving the two factors was replicated in five subplots.Briefly, the experimental design was composed of 5 subplots × 3 drought levels × 2 severity levels.Drought treatments consisted in the application of rainfall exclusion structures after fire to simulate two different extended-drought conditions together with a third Control treatment without rain exclusion.The AutExcl treatment consisted in the rainfall exclusion throughout the subsequent autumn after the summer fire, which started on September 15 and ended on December 15, 2016.The SprExcl treatment excluded rainfall during the entire spring season after burning (started on March 30 and ended on June 20, 2017).A previous work in these systems showed that applied rainfall exclusion treatments led the system to the 1 st percentile of drier years in the long-term series (Maturano 2022), which is considered an extreme drought event (Knapp et al. 2015).Rainfall exclusion structures consisted in 13 V-shaped colorless polycarbonate (Polycasa ® ) bands (2 mm thick) without UV filters, arranged in two levels and covering 2.25 m 2 at 0.8 m above soil (see Salesa et al. 2022 for further details).In the subplots with the IncrSev treatment, 2000 g m −2 of dry biomass were added, mainly composed of grass species (specifically Brachypodium retusum and, in a lower proportion, B. phoenicoides (L.) Roem.& Schult and Helictotrichon filifolium (Lag.)Henrard) to increase fire intensity.Control fire severity was determined by the soil temperatures reached from the previously existing vegetation without fuel-load manipulation.In this way, we were able to test the effects of summer-drought lengthening on postfire vegetation recovery and the effect of increased fire severity as well as their interaction.Additionally, five extra rainfall-exclusion structures, with the V-shaped bands inverted, were included to assess their possible effects on environmental conditions.For more details of the rainfall exclusion treatments and the experimental fire, see Salesa et al. (2022).

Monitoring fire temperatures and posterior soil water content and soil temperature while applying rain exclusion treatments
During experimental burning, soil temperature was recorded in the 30 subplots (15 to IncrSev and 15 to CSev) using chrome-alumel thermocouples (EasyLog EL, K-type; Lascar electronics, Wiltshire, UK) inserted at a 1-cm depth in the center of the subplot.Each thermocouple was connected to a data logger and recorded soil temperatures every 2 s (EL-USB-TC; Lascar electronics, Wiltshire, UK).This allowed us to test if the addition of dry biomass was reflected as an increase in a fire-induced temperature increase at the soil level.The increased fire severity treatment resulted in a peak soil temperature of 314 ± 39 °C, which is 1.9-fold higher than CSev (165 ± 41 °C) (Welch two sample t-test; df = 8206.8,p < 0.001).In addition, the time duration of soil-temperatures in all the temperature intervals were longer in Incr-Sev (Fig. 1).
To assess the effect of the rainfall exclusion treatments on soil-water availability, soil volumetric water content (VWC; m 3 m −3 ) was continuously monitored.This assessment began in the first autumn after the fire and continued until the end of the study period.
VWC was measured by EC-5 sensors (Decagon Services Inc., Pullman, WA, USA) connected to multiple 5-channel Em5b data loggers (Decagon Devices Inc.).Three soil moisture sensors per treatment were installed at a 5-cm depth.The rainfall exclusion treatments led to a drop in soil water content of between 30-44% and 20-32% for AutExcl and SprExcl, respectively, compared to the Control (Kruskal-Wallis test: AutExcl: chi-squared = 41.91,p < 0.001; SprExcl: chisquared = 119.71,p < 0.001).In the same subplots, soil temperature was also monitored with two data loggers (iButtons Thermochron ® ) buried at a 1-cm depth (one at the top and one at the bottom of the subplot).The rainfall exclusion structures increased soil temperatures by between 0.79 and 1.24 °C for the AutExcl (ANOVA, F = 14.91; p < 0.001) and by between 1.63 and 2.10 °C for SprExcl (ANOVA, F = 14.81, p < 0.001).For both soil moisture and temperature, data were collected every 2 h.Measurements began being taken on November 17, 2016 (Fig. 1 and Fig. S1), 2 months after the rainfall exclusion shelters were installed (15-09-2016).During these 2 months, 106.8 mm were recorded b Average volumetric water content in soils for the three rainfall exclusion treatments during the study period.Soil water content is obtained as the average of each treatment (n = 3).Shade areas show the standard deviation for each treatment, and bars show the daily-precipitation amount (mm).Vertical lines indicate the period when the treatments were applied: FA, First Autumn when AutExcl (autumn exclusion) treatment was applied between 15 th September and 15 th December, and Spring when SprExcl (spring exclusion) treatment was applied between 30 th March and 20 th June.Soil water content measurements began 17 th November distributed in 21 different days.See Salesa et al. (2022) for more details of the rainfall exclusion treatments.

Seedling monitoring
In order to determine the rainfall exclusion treatment effects on plant regeneration, we evaluated the seedling emergence and survival of all the present species.Seedling fate was evaluated at the end of each season, with monitoring beginning in the first postfire autumn and ending in the second postfire autumn.Therefore, we assessed plant recovery throughout the first year after fire grouping plant emergence and survival in five different seasonal cohorts: First Autumn, Winter, Spring, Summer, and Second Autumn.For each cohort, we used coloredcoded rings to indicate each plant's germination, which simultaneously allowed us to identify which cohort each dead individual belonged to.We also visually assessed plant cover in percentage per species at the end of the study.

Statistical analyses
All the statistical data analyses were carried out in the R software environment (v.2022.07.0;R Core Team 2021).We employed two-way analyses of variances (ANOVA) to determine if extended drought and fire severity, as well as their combined effect, had an impact on plant recovery.We specifically assessed the impact in the total emerged and total established seedlings at the end of the study period.These analyses were performed for each species separately and grouped by families and functional groups.The identified species were sorted into four different functional groups according to their life form following Tavşanoğlu and Pausas (2018).These groups were: annual forbs, perennial forbs, subshrubs, and shrubs.Liana was removed from the analyses because it was composed of one species (Rubia peregrina L.) and only two emerged individuals.The resprouting plants present before the experimental fire were not included in the data analysis because they were not significantly affected by treatments.This cover was dominated by perennial grasses, mainly Brachypodium retusum, whose coverage 1 year after treatments was 10.7 ± 5.9%, 10.7 ± 9.9%, and 13.3 ± 8.3% for the Control, AutExcl and SprExcl, respectively (ANOVA: F = 0.301, p = 0.742).By also considering other previously established individuals from facultative species (seeding species with resprouting capacity), this cover increased to 12.4 ± 6.1, 13.8 ± 4.6 and 15.8 ± 3.0 for the Control, AutExcl, and SprExcl, respectively, but there were no significant differences (ANOVA; 0.432, p = 0.652).Therefore, we focused our study on seeding species and their postfire germination process from soil seedbanks, which are supposed to be more affected by fire severity and droughts during early post-fire periods (Salesa et al. 2022).We determined the fulfillment of ANOVA assumptions using the Kolmogorov-Smirnov test for normality and the Bartlett test of homogeneity of variances for test homoscedasticity.When assumptions were not met, data were square root-transformed.For model construction purposes, we used emerged seedlings as the response variable and drought and fire severity treatments as the fixed factors as well as their interaction.When treatments, fire severity, or their interaction significantly influenced plant emergence, multiple pairwise comparisons were made by the Compute Tukey Honest Significant Differences test.All these functions are included in the stats-package (R Core Team 2021).
We also assessed the impact of treatments at the plantcommunity level.To do this, first we used a two-way ANOVA to explore the differences in plant richness at the end of the study period.Secondly, the impacts of treatments on community composition were assessed with a permutational multivariate analysis of variance (PMAV).For this analysis, we used the adonis2 function in the vegan package (Oksanen et al. 2013).For this purpose, we constructed an abundance-species matrix in which the weight of abundant species was lowered employing an abundance fraction of five species by the downweight function.This was done to increase analysis accuracy because only one species accounted for about half of all the seedlings that had emerged (Table S2).Bray-Curtis distances were calculated with 999 permutations, and the spatial homogeneity of variances was checked using the betadisper function and data met all requisites.PMAV was performed with all the treatment combinations as level within a single factor (Control, Control + IncrSev, SprExcl, SprExcl + IncrSev, AutExcl, and AutExcl + Incr-Sev).If PMAV was significant, pairwise PERMANOVAs were made to assess differences among treatments following the Bonferroni's p-adjust method.This procedure was done separately for fire severity treatments, drought treatments, and all their possible combinations.Finally, nonmetric multidimensional scaling (NMDS) was used to observe the compositional differences produced by treatments in the ordination space.For the NMDS implementation, we applied the MetaMDS function with Bray Curtis distances as the dissimilarity index.The envfit (vegan package) function was performed to obtain ordination scores to know the species that more influenced spatial ordination.To visualize each treatment ordination, we used the ordiellipse function (vegan package).

Emergence
By families, emerged seedlings showed wide response variability.All the recorded individuals belonged to 14 different families (Table S3).Of them, Cistaceae represented 62% of all the emerged seedlings, followed by Leguminosae (28%).Drought and severity treatments (nor their interaction) did not impact the total emerged seedlings throughout the study period (Fig. 2; Table S4; two-way ANOVA, p > 0.05).By functional groups, shrubs were the dominant group and accounted for 59% of the total emerged seedlings (Table S5).They were followed by subshrubs and perennial forbs, which respectively  S4 and S6 for expanded results represented 34% and 4% of seedlings.None of the treatments influenced functional-groups' seedling emergence (Table S6, two-way ANOVA, p > 0.05), but subshrubs marginally reduced when the AutExcl treatment was applied (Fig. 2e; ANOVA, F = 2.976; p = 0.07).

Established seedlings
At the end of the study period 30% of the emerged seedlings remained alive.Of them, Cistaceae was the most abundant family with 48% of the established seedlings, followed by Leguminosae (33%) (Table S3).Drought treatments had significant effects on the final established seedlings for Cistaceae and for the whole set of emerged individuals (Fig. 3; Table S4; two-way ANOVA p < 0.001 and p = 0.007, respectively).For Cistaceae, both the rainfall exclusion treatments reduced established seedlings in relation to the Control, and the same effect appeared for the total seedlings established when considering all the families together.There was no effect of increased severity nor any interaction with drought.Similar results for subshrubs were found when grouping established seedlings by functional groups (Table S6; two-way ANOVA p = 0.011).In comparison to the Control, both treatments AutExcl and SprExcl significantly reduced the number of established seedlings.
Regarding species, drought treatments had significantly reduced the established seedlings for the two most abundant species by the end of the study period: C. albidus (ANOVA; F = 3.687, p = 0.040) and H. cinereum (ANOVA; F = 4.567, p = 0.021) (Table S7).Drought impact was driven by AutExcl differences on the Control (post hoc Tukey test; p = 0.040 and p = 0.017 for C. albidus and H. cinereum respectively, but not on SprExcl.

Plant richness and community composition
We have found no differences among treatments in either total plant richness or sorted by families by the end of the study period (Table S4; two-way ANOVA p > 0.05; mean ± SD species per 0.25 m 2 plot: 5 ± 2).Nevertheless, by functional groups our results showed that the IncrSev treatment marginally impaired annual forbs' richness (Table S6; two-way ANOVA p = 0.070) because no annual forbs were found in the IncrSev plots.Despite the weak impact of treatments on species richness, significant differences in community composition were observed when considering the combination of all the possible treatments (Fig. 4; Table S8; PMAV, F = 1.766, p = 0.038).The obtained differences were driven mainly by rainfall exclusion treatments because all the observed differences were between the various drought levels (Table S7).Specifically, the post hoc PERMANOVA revealed differences between AutExcl and Control + IncrSev (pairwise PERMANOVA; F = 4.379; p = 0.003) and also between AutExcl + IncrSev and Control + IncrSev (pairwise PER-MANOVA; F = 3.519; p = 0.027).Marginal significant results were also obtained between AutExcl and Control (pairwise PERMANOVA; F = 4.379; p = 0.059) and between Control + IncrSev and SprExcl + IncrSev (pairwise PERMANOVA; F = 2.626; p = 0.084).We observed that the species which contributed the most to spatial ordination were perennial forbs Hieracium pilosella L. (p = 0.001), Sanguisorba minor Scop.(p = 0.023), Cirsium acaule (L.) All.(p = 0.039), and Reseda sp.(p = 0.039).The shrub Cistus albidus also significantly contributed to spatial ordination (p = 0.011), which was also the most abundant species.C. albidus was found mainly in treatments with IncrSev (AutExcl + IncrSev, IncrSev + SprExcl and IncrSev + Control) and the Control, but its lowest density appeared in the treatments with rainfall exclusion (AutExcl and SprExcl).H. pilosella, Reseda sp., and C. acaule were the most abundant in the Incr-Sev + SprExcl treatment, while the Control was the only treatment where H. pilosella was not found.S. minor was found mainly in the Control subplots but was inexistent in AutExcl (neither without nor with IncrSev) and in the Control + IncrSev.

Discussion
The postfire summer-drought lengthening season (in either the subsequent autumn or spring) impacted plant recovery in the first postfire year by reducing seeding species' seedling establishment.Therefore, as the population of these species was established mainly in this first postfire year, we suggest a permanent legacy effect on vegetation composition (Salesa et al. 2022).However, contrary to our hypothesis, an increase in fire severity did not notably impact the seedling emergence and establishment of the main species, families, and functional groups neither by itself alone nor by interacting with prolonged summer drought.We only found a combined effect of fire severity and prolonged summer drought on plant community composition 1 year after fire, driven mainly by changes in the presence of perennial forbs in some of our treatments.
Water availability is crucial for plant success in the first postfire stages (Moreno et al. 2011).This implies that extended summer droughts can have significant impacts on plant recovery, as we demonstrate here.The obtained results reinforce some previous works, which suggest community composition shifts as a consequence of drought-regime alterations (Moreno et al. 2011;Enright et al. 2014;Karavani et al. 2018).This is because our studied community very much depends on disturbances for regeneration, and no notable changes in community  S4 and S6 for expanded results composition should occur in the absence of disturbances (Ooi et al. 2014).This is specially the case of the dominant species, C. albidus (Santana et al. 2012).In fact, although we obtained lesser establishment by drought in the whole assessed community, our results were largely influenced by Cistaceae, the main family recorded in the first year after fire.This family (and by extension, shrubs) was the most affected in survival terms, with stronger detrimental effects on AutExcl (93 ± 3% of shrub-seedlings lost, while this figure lowered to 40 ± 12% for subshrubs).Prolonged drought immediately after a fire may not only indicate more harm to a functional group but can also indirectly favor annual forbs.While AutExcl was the most detrimental treatment for shrubs, it was the treatment with the most emerged annual forbs (60% of the total emerged seedlings) and only 33% mortality.This functional group is considered a pioneer species, and it uses available resources quickly to establish itself.This implies a tradeoff in its competitive capacity (Pierce et al. 2016), although this capacity can vary with minor rainfall changes (Van Dyke et al. 2022).For this reason, we think that annual forbs could have been favored in a less competitive environment.However, the effect was weak due to the few annual forbs and we expect this effect to vanish as succession advances.
Some climate variables are expected to keep intensifying fire severity in the next decades as part of ongoing global change (Parks and Abatzoglou 2020;Grünig et al. 2023).However, the expected impact of this severity increase is not clear.While some works have found that an increase fire severity causes enhanced impacts on ecosystem properties (Díaz-Delgado et al. 2003;Shi et al. 2022;Aguayo-Villalba et al. 2021;García-Llamas et al. 2019;Huerta et al. 2022), other authors have not found clear detrimental effects on Mediterranean-type ecosystems (Francos et al. 2016;Smith-Ramírez et al. 2022).We did not find as many impacts in plant recovery as we expected.This could be related to the firereleased heat reached in the IncrSev treatments.Our IncrSev treatment achieved a significantly fire-released heat increase compared to the CSev treatment, and values came close to other fire-intensity values obtained both with and without fuel addition (Kennard et al. 2005;Knapp et al. 2018;Gonzalez et al. 2022), although the results also depend on measurement depths.However, in some experimental fires in which detrimental effects of increased fire severity were seen, dry biomass addition was up to 4-fold more (4 and 8 kg/m 2 ; Moreno and Oechel 1991) than in our case (2 kg m -2 ).Although it is a notable difference, the characteristics of the added biomass could influence heat release and thus treatment effects.In this experiment, we added dry biomass of herbs in a shrub-dominated community, which not only modified fuel load, but also it might have influenced fire behavior (e.g., accelerating fire spread and shortening temperature residence time; Dimitrakopoulos 2002) thus counterbalancing the increase in fuel load.In the IncrSev treatments, we unnaturally converted a shrub/ grass system into a grass system, which may have led to the fire-released heat reached in the IncrSev treatments not being high enough to harm seeds, or at least to cause detrimental effects on emergence.Most of the detrimental effects of fire severity reported in Mediterranean ecosystems have been assessed in non-experimental fires characterized by their size or severity (e.g., García-Llamas et al. 2019).Therefore, the absence of IncrSev effects can be linked with the ability of obligate seeder seeds to cope with high temperatures.Obligate seeders have been suggested to be more resistant to high fire severity than other functional groups (Moreno and Oechel 1991;Pausas and Keeley 2014).This could be due to the presence of a hard coat, which enables seeds to persist after fire by protecting embryo or seed tissues and, thus, reduces the effects of IncrSev.In our study, around 90% of the emerged and 80% of the established individuals belonged to Cistaceae and Leguminosae, which are families with this hard coat that facilitates tolerance to the IncrSev treatment.In fact, we observed an increasing, albeit nonsignificant, trend in Cistaceae germination and establishment, promoted by IncrSev, which could be a direct positive effect of fire-induced soil temperature increase on the breakage of seed physical dormancy and germination stimulation rather than a deleterious effect.The smaller amount of emerged and established seedlings from other functional groups in our studied community, such as annual and perennial forbs, could have hampered the observation of an increased severity effect on vegetation regeneration.These groups mainly lack this hard coat and could have experienced a stronger impact than that previously reported (Abedi et al. 2022).As a matter of fact, we only obtained emerged and established annual forbs in CSev.
Contrary to our expectations, we found no interaction effect between IncrSev and drought treatments.Previous works that have assessed plant recovery have found different fire-severity effects depending on climate conditions, with more detrimental impacts for high severity under drier conditions (Huerta et al. 2021).However, in our work, the fact that there were no interactions between fire severity and drought could be due to the strong influence of water availability, and any possible effect of IncrSev due to plant competition for water vanished.This could be one of the reasons to explain why we only found effects on plant establishment rather than on emergence.Postfire water availability has been seen to have a major effect not only on postfire germination (Luna and Chamorro 2016;Salesa et al. 2022) but also on recovery success (Prieto et al. 2009;Moreno et al. 2011;Batllori et al. 2019).This is especially true if water shortage occurs during critical plant establishment periods for obligate seeders or exceeds certain intensity thresholds (Karavani et al. 2018;Ingrisch et al. 2022).Our results show that drought in the first autumn and in the first spring after fire reduced establishment preferably in Cistaceae and subshrubs, both of which are composed mostly of obligate seeders.This functional group reestablishes itself properly in post-disturbance environments (Santana et al. 2012), and some families have features that favor its persistence in dry and open habitats throughout the Mediterranean Basin (Thanos et al. 1992).Nevertheless, establishment success in obligate seeders is closely linked with environment conditions and more specifically with water availability (Quintana et al. 2004;Moreno et al. 2011;Karavani et al. 2018;Chamorro and Moreno 2019).These environmental conditions in the first months after fire for obligate seeders could be even more relevant for mortality than those of the summer season (Quintana et al. 2004), and this could explain why treatments AutExcl and SprExcl significantly impaired plant establishment (Fig. 3).Although we know that obligate seeders plants have been linked with high resilience to fire and drought (Rundel et al. 2018;del Cacho and Lloret 2012), knowledge about how plants subjected to drought and high severity fire interactions could face future disturbances remains scarce (Nolan et al. 2021).For this reason, it would be helpful to ascertain whether the combination of treatments can deplete soil-seed banks by lowering these species' capacity to withstand with new disturbances (Santana et al. 2020).Furthermore, shedding light on the understanding of potential legacy effects of those plants affected by drought in the early life stages, for example biological fitness, seems crucial for understanding the emerging impacts of new disturbance regimes.
Fire effects have been suggested to be contextdependent, for example through local fire regime, recovery strategies, time since last fire, and even the assessment scale (McLauchlan et al. 2020;Miller and Safford 2020;Smith-Ramírez et al. 2022).However, the absence of detrimental effects found by treatments on species richness is in accordance with previous research.For instance, Abella and Fornwalt (2015) obtained no differences in species richness with different fire severity classes.Similarly, González-De Vega et al. (2016) reported no differences in richness among three severity classes in natural fires, independently of the time since fire.Obtaining no effects might be partially explained by Mediterranean plants' high resilience to fire and drought, as previously discussed (Rundel et al. 2018;del Cacho and Lloret 2012).Another reason for no effects on richness could be because of increases in species evenness, which might allow larger suitable niches for more species and could, thus, counteract negative treatment effects (Kardol et al. 2010;Alon and Sternberg 2019).
Finally, there might exist different types of relationships between fire severity and plant-diversity effects which could depend on historical disturbance regimes and modern departures from those regimes (Miller and Safford 2020).Therefore, a wider approach considering disturbance regimes might help to understand fire severity effects in future experiments.Despite lack of influence of treatments, some fire severity and drought combinations have revealed differences in community composition.This effect is expected because changes in vegetation type have been previously related to fire severity (González-De Vega et al. 2016), which might, in turn, be related to seeds' different heat sensitivity (Moreno and Oechel 1991).In our case, we observed that species richness in annual forbs was marginally affected by IncrSev (Table S6) because we only recorded individuals in no-drought and control fire-severity treatments.However, the number of established individuals was small (n = 3).Moreover, plant composition shifts have been driven mainly by perennial forbs seeing that four of the five most influential species that we obtained in our NMDS ordination belonged to his functional group (Fig. 4).Perennial forbs were not widely represented, but our results suggested that their presence responded sensitively to treatments.Mediterranean-forbs richness can benefit from protracted autumn droughts as a consequence of a drawdown of dominance in some species (Nogueira et al. 2017).They may also be positively affected in terms of cover and richness by fire but to a lesser extent than those affected by cutting (Tárrega et al. 2001).This suggests that this functional group adequately performed in disturbed environments, but exceeding certain fire-released heat thresholds might surpass the positive effects (Elliott et al. 2011).This statement matches the results obtained by Abella and Fornwalt (2015) and Huerta et al. (2022), who showed that forbs were affected more by increased fire severity classes.These results suggest that forbs are unequally sensitive to fire regime shifts.So, in addition to impairing plant establishment, the joint effect of increased fire severity, together with protracted summer droughts (i.e., altered disturbance regimes), could generate tradeoffs and changes in plant community in such ecosystems (González-De Vega et al. 2016;Karavani et al. 2018).In addition, the results herein obtained could have important implications in recently burned areas.Forbs are pioneer species in these environments and they increase soil roughness, which has been proven to play an important role in sediment retention (Kervroëdan et al. 2019).Therefore, stronger impacts on this functional group caused by increases in fire severity could imply further impacts on burned areas.

Conclusions
This work addresses the compound effects of drought and the experimental increase in heat released by fire on the regeneration process in a Mediterranean shrubland.Increased heat release by fire might not affect plant recovery either by itself or by interacting with drought because prolonged drought may mask any impacts of increased severity fire on Mediterranean seeding species.We conclude that water limitations during critical periods, such as the first postfire autumn, are crucial to Mediterranean communities, especially for Cistaceae and subshrubs, because they diminish seedling establishment during the first year postfire.Additionally, despite the absence of detrimental effects on plant emergence and establishment caused by increased fire severity, a distinct sensitivity of functional groups to altered disturbance regimes may cause shifts in community composition.These findings have important implications for understanding how increased fire severity and prolonged summer-drought seasons may lead to tradeoffs and changes in plant communities in Mediterranean ecosystems, particularly for pioneer species like forbs.

Fig. 1
Fig.1Treatment effects at soil level.a Average soil temperature recorded in the subplots from the beginning of the experimental fire to 4 h after fire.Shaded area shows the standard error for each treatment (orange: Increased fire severity treatment, IncrSev, n = 15; green: Control severity treatment, CSev, n = 15).In the upper right corner, the average residence time at the different temperature intervals for both treatments appears.b Average volumetric water content in soils for the three rainfall exclusion treatments during the study period.Soil water content is obtained as the average of each treatment (n = 3).Shade areas show the standard deviation for each treatment, and bars show the daily-precipitation amount (mm).Vertical lines indicate the period when the treatments were applied: FA, First Autumn when AutExcl (autumn exclusion) treatment was applied between 15 th September and 15 th December, and Spring when SprExcl (spring exclusion) treatment was applied between 30 th March and 20 th June.Soil water content measurements began 17 th November

Fig. 2
Fig. 2 The total emerged seedlings at the end of the study period.In a, c, e, and g, the values belonging to the functional groups are shown.The results in b and d show the emerged seedlings for the two most abundant families and f the seedlings of other remaining families.The total emerged seedlings without groupings appear in h.The distinct drought treatments are depicted in different colors; in each one, the line-filled bars indicate the increased fire severity treatment.Bar errors denote standard error (n = 5).The two-way ANOVA results are shown in each plot to display the drought (D) and increased fire severity (IncrSev) effects as well as their interaction (D*IncrSev).See TablesS4 and S6for expanded results

Fig. 3
Fig. 3 Established seedlings at the end of the study period.In a, c, e, and g, the values belonging to the functional groups are shown.The results in b and d reveal emerged seedlings for the two most abundant families and f for the seedlings of the remaining families.The total emerged seedlings without grouping appear in h.The distinct drought treatments are depicted in different colors and, in each one, the line-filled bars indicate the increased fire severity treatment.Bar errors denote standard error (n = 5).The two-way ANOVA results are shown in each plot displaying the drought (D) and increased fire severity (IncrSev) effects as well as their interaction (D*IncrSev) Different capital letters denote significant differences between treatments.See Table S4 and S6 for expanded results

Fig. 4
Fig. 4 Nonmetric multidimensional scaling (NMDS) ordination for species abundance at the end of the first postfire year.Species ordination is illustrated with each point representing distinct species.Only the species with the strongest influence on ordination (p < 0.1) are labeled.Ellipses are drawn using standard deviation and encompass all the drought and fire severity levels (n = 5 per treatment).The displayed p-value corresponds to the performed PERMANOVA analysis.The species abbreviations in a decreasing ordination contribution are as follows: Hie_pil, Hieracium pilosella; C_alb, Cistus albidus; San_min, Sanguisorba minor; Cirs_aca, Cirsium acaule; Res, Reseda sp.Atr_hum, Atractyllis humilis; Kna_purp, Knautia purpurea