Effects of Drought and Fire Severity Interaction on Short-Term Post-Fire Recovery of the Mediterranean Forest of South America

Abstract

Wildfires and drought stressors can significantly limit forest recovery in Mediterranean-type ecosystems. Since 2010, the region of central Chile has experienced a prolonged Mega Drought, which intensified into a Hyper Drought in 2019, characterized by record-low precipitation and high temperatures, further constraining forest recovery. This study evaluates short-term (5-year) post-fire vegetation recovery across drought gradients in two types of evergreen sclerophyllous forests and a thorny forest and shrubland, analyzing Landsat time series (1987–2022) from 42 wildfires. Using the LandTrendr algorithm, we assessed post-fire forest recovery based on NDVI changes between pre-fire values and subsequent years. The results reveal significant differences in recovery across drought gradients during the Hyper Drought period, among the three forest types studied. The xeric forest, dominated by Quillaja saponaria and Lithrea caustica, showed significant interaction effects between levels of drought and fire severity, while the thorny forest and shrubland displayed no significant interaction effects. The mesic forest, dominated by Cryptocarya alba and Peumus boldus, exhibited additional significant differences in recovery between the Hyper Drought and Mega Drought periods, along with significant interaction effects. These findings underscore the critical role of prolonged, severe drought in shaping forest recovery dynamics and highlight the need to understand these patterns to improve future forest resilience under increasingly arid conditions.

1. Introduction

Natural disturbances, such as wildfires and droughts, play a crucial role in the ecological dynamics of forest ecosystems by shaping their structure, composition, and function [1,2]. These disturbances leave long-lasting imprints on forest trajectories, influencing their future states [3,4]. However, anthropogenic climate change and land-use practices have significantly altered these disturbances’ frequency, intensity, and interactions [5,6,7]. As climate-related disturbances increase in both duration and intensity, they may transition from isolated incidents to persistent stressors, exemplified by extended drought conditions [8]. These persistent stressors can severely limit vegetation recovery and elevate the risk of surpassing ecological thresholds, potentially leading to irreversible changes in forest structure and function [6,9]. This vulnerability is particularly pronounced in Mediterranean regions, where the interaction between wildfires and droughts can push ecosystems toward critical thresholds [10], potentially reducing forests’ capacity to provide essential services such as carbon sequestration and hydrological regulation [3,11]. Comprehensive assessments of post-fire vegetation recovery under these stressors are essential for understanding how these ecosystems could change. Such evaluations also help to anticipate the broader impacts of environmental changes, ensuring the continued provision of ecosystem functions, including resource supply, biodiversity conservation, and ecosystem service maintenance.

Among Mediterranean ecosystems, central Chile faces intensified environmental stressors [12,13]. Located along the southern Pacific coast of South America, central Chile presents a unique opportunity to study the effects of shifting disturbance regimes and the interactions between wildfires and drought. The region is recognized for its high level of endemism, making it one of the world’s biodiversity hotspots [14]. Central Chile has been undergoing a significant dry period since 2010, characterized by rising temperatures and a long-term decline in rainfall [15,16,17], with a persistent precipitation deficit ranging from 25% to 45% for over a decade, referred to as the Mega Drought [12,16,18]. The ‘Mega Drought’ intensified in 2019 and 2021, with rainfall dropping to less than 30% of climatological records and temperatures elevating during the spring and summer seasons [19]. This phase, termed a Hyper Drought, marked the peak of an extended drought that has persisted for over a decade. During this time, the region showed an abrupt decline in vegetation productivity compared to the 2000–2010 baseline period [20].

These severe drought conditions have coincided with an increase in wildfire frequency and magnitude, exacerbated by climatic conditions favorable to the spread of fires, posing substantial risks to native vegetation [13,21,22,23,24]. Similar to other Mediterranean ecosystems, the historical transformation of land use in central Chile has significantly reduced the extent of original forests, leaving remaining forests mostly confined to slopes and limited access to water. These forests, mainly found in mesic, south-facing areas, consist of three forests with different species compositions (forest subtypes): degraded thorny shrublands dominated by Vachellia caven & Maytenus boaria located in xeric areas; coastal Mediterranean forests primarily covered by Cryptocarya alba & Peumus boldus located in mesic conditions; and Andean Mediterranean forests dominated by Quillaja Saponaria & Lithrea caustica established in drier, xeric locations [25,26,27]. However, dense canopy coverage areas are mostly restricted to topographically water-accumulating sites or microclimates that favor their persistence. Evergreen sclerophyllous forests in central Chile are adapted to withstand low soil water potential and fire [28], and natural ignitions are rare in central Chile due to infrequent lightning-producing storms without rain [29]. Many tree and shrub species in central Chile’s Mediterranean forests exhibit resilience by regenerating primarily through resprouting, while some species maintain seeds even after low-intensity fires [30,31,32].

Post-fire and drought recovery in these forests frequently rely on stored resources within survival structures and the varying ability of basal buds, such as lignotubers, roots, rhizomes, and nodes, to sprout [32]. While resprouting from underground structures is possible if sufficient carbon reserves are available, it may fail due to hydraulic limitations, making forest recovery uncertain if drought conditions persist and water balance declines [33]. Moreover, seedling recruitment after disturbances in Chile’s sclerophyllous forests is limited by the rapid establishment of exotic species and the high abundance of herbivores [34]. It remains unclear how different forests, and their dominant species, respond to post-fire recovery under drought-driven declines in productivity.

Remote sensing is a widely used tool for monitoring post-disturbance vegetation recovery [35,36]. Analysis of time-series vegetation indices derived from satellite imagery, such as the Normalized Difference Vegetation Index (NDVI), helps track vegetation recovery and photosynthetic capacity after events like wildfires [37,38,39]. However, the effects of the interaction of long-term drought, Hyper Drought, and wildfire have been poorly documented because they rarely occur in extensive areas, high intensity, and decadal long periods. The Landsat program has been instrumental in offering high-resolution imagery, supporting advanced research using techniques like LandTrendr [40], which enable in-depth analysis of vegetation changes following disturbances [41,42,43]. The NDVI has been widely used as a reliable proxy for assessing post-fire vegetation productivity recovery [35,44,45]. NDVI is closely associated with photosynthetic capacity, as its greenness primarily indicates chlorophyll content [46,47]. By integrating advanced remote sensing techniques with analytical models [48,49,50], we enhanced our ability to characterize, assess, and monitor forest responses to fire, providing robust insights to inform conservation and management strategies effectively.

Given the expected increase in aridity for this region and the higher frequency of extreme events [51,52], comprehensive assessments at both regional and local scales were crucial for enhancing our understanding of vegetation recovery. While recent studies have begun to explore the complex relationships between forests, drought, and wildfires in central Chile [13,21,53,54], to our knowledge, no studies have directly addressed the specific effects of post-fire recovery in these ecosystems. Mediterranean forests in the Southern Hemisphere, particularly in central Chile, remain underrepresented in global analyses despite their ecological importance [47]. As a result, there is a need for targeted research to better assess the dynamics of vegetation recovery in response to increasing climate stressors and fire severity.

Given the context described above, this research aims to characterize and compare short-term (5-year) forest recovery following wildfire, as measured by NDVI. We focused on recovery variability across different levels of drought and fire severity between forests of different species compositions (forest subtypes). Specifically, we sought to answer the following questions: Are there significant differences in forest recovery across levels of drought and fire severity, based on forest composition? Are there interaction effects between drought and fire severity that shape forest recovery? By addressing these questions, our research provides initial insights into how drought and fire severity shape recovery patterns within central Chile’s Mediterranean ecosystems, establishing a foundation for further investigation into these dynamics.

2. Materials and Methods

2.1. Study Area

The study area is situated in central Chile, spanning from the Pacific coast to the Andean foothills, between 71.5° W and 70.0° W, and from 32.0° S to 36.0° S (Figure 1). This region encompasses Chile’s Mediterranean sclerophyllous shrublands and forests, stretching from coastal zones to the pre-Andean mountains. The area is a Mediterranean ecosystem, characterized by winter precipitation and a prolonged dry summer season, heavily influenced by the El Niño-Southern Oscillation (ENSO) [16]. Annual precipitation ranges from approximately 220 mm in the central valley to around 700 mm along the coast, while mean annual temperatures range between 12 °C and 15.1 °C (data obtained from http://explorador.cr2.cl/ (accessed on 17 Jun 2024)). Although this ecosystem typically endures one to two years of drought, since 2010 it has been experiencing an unprecedented, prolonged drought that continues to the present (Figure 2), leading to a marked decline in vegetation productivity [20].

2.2. Selection of Burned Forest Areas

We used fire perimeter and severity data from a previous study in the area [55]. Forest spatial distribution and type were obtained from the official vegetation map provided by the National Forest Service [56,57,58]. From this dataset, which covers the four administrative regions of the study area and includes more than 5533 wildfires, we applied a multi-criteria filter for vegetation attributes considering natural forest, the main subtypes of Mediterranean sclerophyllous forest and shrublands (Q. saponaria & L. caustica, V. caven & M. boaria, and C. alba & P. boldus), and dense canopy cover. The selection of dense canopy cover prevented the retrieval of a combined spectral response. We avoided areas with multiple fires and ensured that there were no changes in land cover, such as urban development, plantations, or agricultural expansion, by evaluating five years before and after the wildfires. This was performed using the LandTrendr change detection algorithm [59] and visually confirmed with Google Earth imagery. Consequently, 44,108 pixels from 42 wildfires, spanning 1992 to 2017, met the vegetation, wildfire, and land-use criteria (Figure 1 and Figure 2, and Table S1). The total number of burned pixels selected varied annually, with the largest number being in 2017 due to the mega-fire season (Figure 2) [60].

Figure 2. Temporal distribution of burned areas analyzed by forest subtype. Above the bars are indicated the number of selected wildfires by year. The blue line represents the annual precipitation from TerraClimate data [61], describing the three distinct drought periods analyzed in this study.

Figure 2. Temporal distribution of burned areas analyzed by forest subtype. Above the bars are indicated the number of selected wildfires by year. The blue line represents the annual precipitation from TerraClimate data [61], describing the three distinct drought periods analyzed in this study.

2.3. Landsat Time Series for Post-Fire Recovery

We implemented all our input data and processing in Google Earth Engine (GEE), an open cloud-computing platform designed for geospatial analysis. GEE provides access to a comprehensive public catalog comprising satellite imagery, topographical data, and climate and environmental datasets [62]. Prolonged periods of consistent satellite data were essential to assess interannual changes. We used NDVI to characterize the trajectory and spectral response post-fire. NDVI time series was derived using the LandTrendr spectral-temporal segmentation algorithm [59], which was proposed by Kennedy et al. [40]. This algorithm adopts a singular perspective based on a pixel’s spectral history. It undergoes a systematic procedure to discern breakpoints that delineate intervals of sustained alterations or constancy in spectral trajectories. The identified breakpoints, characterized by the year of occurrence and spectral index value, represent a pixel’s spectral history as a sequence of vertices bounding line segments. The outcome is an idealized trajectory-based time series devoid of noise, where each observation is situated within the framework of a spectral-temporal trend. In this study, LandTrendr contributes to two methodological steps. The first is detecting land use changes for selecting burned areas, and the second is collecting noise-reduced NDVI time series to evaluate the post-fire forest response. We opted for this adjusted data format instead of using unmodified surface reflectance to minimize the impact of minor fluctuations in time series data attributed to instabilities in climate, atmospheric conditions, phenology, solar angle, and limitations in available images, on the determination of post-fire NDVI recovery.

The dataset included images that had already undergone atmospheric and geometrical corrections. We utilized all available TM/ETM+/OLI/OLI2 Level-2, Collection 2, Surface Reflectance Tier 1 images from Landsat 5, 7, 8, and 9 (for details on pre-processing images, see Supplementary Materials), spanning the period from 1987 to 2022. Our focus was on the local summer season (21 December–31 March). In instances where cloud-free pixels were limited, the season was extended to ensure a representative summer composition, extending up to one month beyond the typical summer season if necessary (see cases in Table S2). This summer period was selected to minimize the presence of herbaceous vegetation from the wet season, thereby reducing potential confusion in post-wildfire forest recovery and concentrating on the regeneration of trees and shrubs. The result was a single cloud-free NDVI-band image per year, representing the median NDVI values for the summer season.

2.4. Post-Fire Recovery Percentage

To evaluate forest recovery, we independently calculated the relative change in NDVI for each year following the wildfire, up to year 5, between pre-fire values for each fire (Equation (1), Figure S1). This calculation was repeated annually to create a 5-year time series, using the recovery index proposed by Lloret et al. [63]. This approach, focusing on recovery variability, provides a metric that captures the degree of recovery concerning the extent of damage experienced during a disturbance event and pre-disturbance levels. We explain this metric using our spectral index converted to a percentage, calculated as:

$$\%{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{{\mathrm{r}\mathrm{e}\mathrm{c}}_{\mathrm{i}}}={\displaystyle \frac{{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{y}\mathrm{i}}-{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{y}1}}{{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{p}\mathrm{r}\mathrm{e}}-{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{y}1}}}\times 100$$

(1)

where NDVI_yi is the value of NDVI in the year of post-fire observation, NDVI_y1 is the value next to the disturbance at year 1 (we expect a lower value than pre-disturbance), and NDVI_pre is the weighted NDVI value for the 5 years before the wildfire, with greater weight given to the pre-fire years. For NDVI_rec, values closer to 1 show greater recovery (1 = full recovery), values close to 0 show less recovery, and negative values show a decrease in recovery. This value allows comparison of recovery between forest subtypes and wildfires. We selected NDVI_y1 as a surrogate for post-fire vegetation conditions; this could represent both open-growing canopy recovery and the presence of a remnant or recovery of overstory trees [64,65].

2.5. Statistical Analysis of Recovery Differences in Drought Levels and Fire Severity

We structured our analysis by categorizing drought conditions into three distinct levels: No Drought (1992–2009) as the climate reference [21], the Mega Drought (2010–2016) [15,16], and the Hyper Drought (2017–2022), which represents the most severe stage of the prolonged drought, including 2019 and 2021, the driest and warmest years on record during this period [17]. Additionally, we classified fire severity using Relative Delta Normalized Burn Ratio (RdNBR) [66] (Equation (2)), and classified it into two categories: medium severity (316–640 RdNBR) and high severity (≥641 RdNBR). We excluded low-severity (69–315 RdNBR) pixels because there were fewer than 200 observations, which were concentrated in Q. saponaria & L. caustica. We ran a subsample selection to avoid slight differences being considered statistically significant due to the large sample sizes (Figure 2). For resampling, pixels were split into groups by combining drought level and fire severity. Within each group, pixels were drawn using stratified random sampling, ensuring samples were proportional to the strata (Table S3). All statistical analyses were performed in R software version 4.4.1 [67].

$$\mathrm{R}\mathrm{d}\mathrm{N}\mathrm{B}\mathrm{R}={\displaystyle \frac{\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{N}\mathrm{B}\mathrm{R}-\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{N}\mathrm{B}\mathrm{R}}{\sqrt{\mathrm{A}\mathrm{B}\mathrm{S}(}\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{N}\mathrm{B}\mathrm{R})}}$$

(2)

We examined the effects of drought and fire severity on the recovery of the three forest subtypes following fire disturbances, as well as potential interactions between these factors. We conducted a two-way analysis of variance (ANOVA) to determine significant differences in post-fire vegetation recovery among drought levels and fire severity categories for each forest subtype, and to test for factor interactions. Factor interaction refers to the case where each factor not only exerts an effect on the response variable (described as main effects), but also influences the other factor, resulting in additional joint effects on the response variable (known as interaction effects). We utilized type III and II approaches for calculating sums of squares in two-way ANOVA, depending on the presence or absence of interactions between factors and the imbalance of the data [68]. When a significant difference was detected in the ANOVA, we conducted a post hoc analysis using Tukey’s Honestly Significant Difference (HSD) test to identify specific differences between groups. If the interaction was significant, we explored which factor was significant at each level of the other factor. In cases where the interaction was not significant, we focused on examining the main effects. Before conducting the ANOVA, we assessed the assumptions of normality and homoscedasticity by inspecting the residuals of each model. To evaluate normality, we applied the Kolmogorov–Smirnov test with the Lilliefors correction, while homoscedasticity was evaluated using the Levene test.

To describe drought periods, we used the annual Palmer Drought Severity Index (PDSI) from TerraClimate data [61], a widely recognized metric for assessing drought intensity that integrates temperature and precipitation data to estimate relative soil moisture levels [69]. Negative PDSI values indicate drier conditions, while positive values suggest wetter conditions, offering insights into the severity and duration of short-term and prolonged drought periods.

3. Results

The NDVI recovery five years post-fire exhibited differences across forest subtypes. In Q. saponaria & L. caustica, significant effects were identified for drought level (ANOVA: F = 203.518, df = 2, p < 2.2 × 10⁻¹⁶), fire severity (ANOVA: F = 30.436, df = 1, p < 2.2 × 10⁻¹⁶), and the interaction between these factors on NDVI recovery at year five (ANOVA: F = 3.263, df = 2, p = 0.038). We observed the highest recovery under conditions of No Drought and High fire severity, while the lowest recovery occurred during Hyper Drought across both fire severity levels (Tukey’s multiple comparison test, p < 0.01) (Figure 3 and Tables S5 and S6). Given the interaction between factors (Figure 3 and Figure S2), the significance of the interaction of fire severity for each drought level group was evaluated, showing significance only for No Drought and Mega Drought conditions (p < 0.01). For the V. caven & M. boaria, we noted significant effects for drought level (ANOVA: F = 173.88, df = 2, p < 2.2 × 10⁻¹⁶), while fire severity did not show significant effects (ANOVA: F = 8.829, df = 1, p = 0.003), nor did the interaction between these factors on NDVI recovery at year five (ANOVA: F = 0.276, df = 2, p = 0.759). The highest recovery rates were recorded under both No Drought and Mega Drought conditions, whereas the lowest recovery occurred under Hyper Drought at high severity (p < 0.01) (Figure 3 and Table S4). In the case of C. alba & P. boldus, significant effects were found for both drought level (ANOVA: F = 186.892, df = 2, p < 2.2 × 10⁻¹⁶) and fire severity (ANOVA: F = 73.491, df = 1, p < 2.2 × 10⁻¹⁶), along with their interaction (ANOVA: F = 15.102, df = 2, p = 3.3 × 10⁻⁰⁷). The highest recovery was observed under No Drought conditions across all fire severity levels, with the lowest recovery noted during Hyper Drought (p < 0.01). The interactions between fire severity were found to be significant for each drought level, affecting NDVI recovery (p < 0.01) (Figure 3 and Table S5).

A descriptive analysis of the overall means indicating marked differences among forest subtypes showed the highest recovery in high severity than medium severity (Figure 3 and Figure S3, and Table S6). For high severity, C. alba & P. boldus achieved the highest recovery during No Drought levels, averaging approximately 100%. This was followed by Q. saponaria & L. caustica at 89.8%, and V. caven & M. boaria at 63.0%. Under Mega Drought, recovery was higher for Q. saponaria & L. caustica (80.2%), and V. caven & M. boaria (65.6%), whereas C. alba & P. boldus declined to 55.8%. As drought intensity increased to Hyper Drought, recovery converged to 39%, 30.5%, and 27.9% for C. alba & P. boldus, Q. saponaria & L. caustica, and V. caven & M. boaria, respectively. For medium fire severity, variability among the three forest subtypes was reduced under No Drought and Mega Drought conditions (69.2% to 45.3%), with the lowest recovery rates observed under Hyper Drought (21.1% to 30.3%). The confidence interval is shown in Table S6.

When considering the No Drought period as a reference for post-fire vegetation recovery, the results reveal notably lower recovery percentages across all forest subtypes during the Hyper Drought level, regardless of fire severity. Hyper Drought, both in high and medium fire severity, shows a significant difference relative to No drought (p < 0.01) (Figure 4). Additionally, C. alba & P. boldus show significant differences for Mega Drought (p < 0.01) and lower relative recovery percentages compared to Q. saponaria & L. caustica and V. caven and M. boaria.

An analysis of the recovery trajectory from year one to year five revealed distinct differences based on drought level and fire severity. Wildfires occurring under the No Drought condition showed continuous recovery each year. In contrast, Mega Drought and Hyper Drought conditions exhibited interrupted recovery phases, particularly for Q. saponaria & L. caustica and V. caven & M. boaria. The Hyper Drought condition was notably associated with more pronounced disruptions in recovery, corresponding to the increased severity of drought and productivity decline observed in year 3 (Figure 5 and Figure 6).

4. Discussion

Understanding ecosystem responses to environmental variables is essential for effective conservation management and planning [70,71]. In the case of Mediterranean vegetation ecosystems, evidence suggests their evolution is closely tied to a climatic history of increasing aridity throughout the Cenozoic [28]. This trend has been accompanied by persistent fire disturbances, resulting in a range of fire-adapted tolerance and reproduction mechanisms associated with fire ecology [72]. However, Chile represents an exception, where localized regional climatic changes linked to the Andean uplift during the Miocene led to a suppression of fires compared to other Mediterranean climate regions [28,72]. Therefore, examining fire tolerance responses in the flora of central Chile provides a unique opportunity to understand the adaptive capacity of an ecosystem that evolved under the stress of climate-induced drought but now faces increasing wildfire pressure [73]. This analysis can help predict potential ecological trajectories and alternative states for this ecosystem [74].

Our analysis revealed statistically significant differences in short-term post-fire recovery based on drought levels, fire severity, and interaction effects (Figure 3). We observed disparity in recovery across forest subtypes, underscoring their distinct resilience capacities. Specifically, C. alba & P. boldus, which exhibited higher average recovery percentages under No Drought conditions, demonstrated increased vulnerability to increasing drought levels. In contrast, Q. saponaria & L. caustica, along with V. caven & M. boaria, showed greater recovery. However, under extreme and prolonged drought conditions, all forest subtypes experienced a significant decline in recovery. These findings suggest that, regardless of forest subtype or fire severity (which appeared to have a lesser effect than drought), a critical threshold of drought level substantially compromised forest recovery.

In other Mediterranean regions, such as Southern California and the Mediterranean Basin, severity and extended drought periods, especially in the early state post-fire, can significantly reduce soil moisture content, hindering seed germination and seedling establishment in vegetation [75,76,77]. This diminished water availability limits the capacity of Mediterranean plant species to initiate the regrowth process after a fire event, thereby impeding post-fire recovery [78]. Blanco-Rodríguez et al. [76] report that the effect of drought duration on post-fire vegetation recovery varied depending on the aridity level, intensifying at extremes of aridity gradient (for semi-arid and humid areas) in the western Mediterranean Basins. In arid environments, the adverse effect of drought on recovery aligns with contemporary findings suggesting that adaptations to drought (for example, embolism resistance) may act antagonistically to the post-fire recovery process [79].

The higher recovery rates observed in Q. saponaria & L. caustica, as well as V. caven & M. boaria, during the Mega Drought likely stemmed from their greater adaptation to limited water availability compared to C. alba & P. boldus, which are typically associated with more humid environments [17]. Moreover, Q. saponaria & L. caustica, particularly those on northern slopes, frequently coexisted with xeric species well adapted to drought conditions. However, during the Hyper Drought, Q. saponaria & L. caustica exhibited lower recovery rates than C. alba and P. boldus. This discrepancy may relate to the browning phenomenon reported in sclerophyllous forests of central Chile in 2020 [20], an event not directly analyzed here but potentially insightful for interpreting our results. Prior research noted that the xeric forest subtype, particularly in the Andean range and the northern extent of its distribution, demonstrated a more pronounced decline in productivity relative to other forest types, aligning with our observations [17]. It is possible that species in xeric forests are nearing a threshold of adaptive capacity under such extreme conditions, where they increasingly reduce foliage in summer or activate alternative drought-adaptive strategies.

In thorny forest and shrublands, we observed relatively lower recovery rates of 27.9% and 21.1% for areas with high and medium fire severity, respectively. This outcome can be attributed to the traits (from plant to ecosystem) exhibited by Vachellia caven which are involved in its recovery dynamics. This species can regenerate rapidly through vegetative means, such as resprouting from stumps and roots, which is especially vigorous after disturbances such as grazing [80]. Its presence is associated with anthropogenic activities (e.g., overgrazing) [81], rather than solely with the degradation of forest ecosystems under desertification, as previously suggested [82]. However, even with its drought tolerance [83], V. caven shows susceptibility to present hydraulic failure under extreme seasonal drought [84]. Thus, under hyper-drought conditions, extreme water deficits further restrict soil moisture availability, impeding both vegetative resprouting and seed germination. Seed germination, in particular, may depend on conditions such as the presence of grazing animals, which facilitate seed dispersal to suitable sites. Despite its prolific seed production, fire can destroy seed banks or severely impact germination success [85]. Furthermore, the degraded state of these shrublands exacerbates soil erosion, altering soil properties and potentially limiting recovery potential under severe drought conditions [86]. While V. caven demonstrates considerable ecological resilience, its slower recovery during hyper-drought likely reflects a complex interplay between adaptive strategies and intense environmental stressors. Future studies should explore longer-term recovery dynamics to capture more nuanced recovery trends.

The analysis of post-fire NDVI recovery trajectories over the five years revealed significant differences based on drought levels and fire severity (Figure 5). Forests experiencing No Drought conditions demonstrated consistent and positive recovery across all forest subtypes, indicating strong resilience. Supporting field observations [73] suggested that initial increases in NDVI were primarily driven by resprouting rather than new seedling growth, particularly under No Drought conditions. However, recovery was notably disrupted during the Mega and Hyper Drought periods, with setbacks observed across every forest subtype. Consistent with the recovery patterns observed in year five, the NDVIrec trajectories during the Mega Drought showed that C. alba & P. boldus and Q. saponaria & L. caustica experienced a considerable decline in recovery by years three and five. This decline was particularly abrupt for C. alba & P. boldus, whose recovery percentages returned to similar levels in year one. These findings support that, for these forest subtypes, more severe drought conditions—evidenced by the continuity of drought—significantly impacted post-fire recovery. In Hyper Drought conditions, these setbacks became even more pronounced for Q. Saponaria & L. caustica and V. caven & M. boaria, with recovery values reaching negative levels by year three (−20%). This decline was related to the persistence of drought conditions and an increase in drought severity, as indicated by more negative precipitation anomalies across all drought levels [13,19].

As an alternative interpretation, the observed drop in NDVI recovery during the third year could be indicative of an ecological succession event. This trend may indicate a shift in community structure, where the decline in photosynthetic productivity by year three could be linked more closely to the senescence or replacement of herbaceous vegetation than to any substantial loss in resprouting shrubs or trees. Consequently, this pattern may signal a transition from herbaceous-dominated to shrub or tree-dominated cover types, potentially driven by climatic events (e.g., dry episodes) that favor such successional dynamics. Similar patterns have been documented in other Mediterranean ecosystems, such as those in California [87,88]. Following this period of decline, a marked recovery was observed, which again fell by year five. This aligns with the NDVI trajectories observed, where rapid vegetative responses likely drove initial recovery under No Drought conditions. In contrast, drought conditions severely hampered these processes, resulting in less recovery (Figure 5).

Our observations revealed that only C. alba & P. boldus achieved an average recovery rate of 100% under No Drought conditions and high fire severity. In contrast, V. caven and & M. boaria displayed the lowest recovery rate at 63.0% under the same conditions. The slower recovery observed in V. caven & M. boaria suggests that these species face greater challenges in regenerating and restoring pre-fire conditions in the short-term than other forest subtypes. However, rather than indicating a lower recovery of V. caven & M. boaria, these patterns may reflect a post-fire shift in community structure, where species dominance changes toward V. caven and R. trinervia rather than evergreen species such as C. alba or P. boldus, particularly in areas affected by repeated fires [50]. This shift, along with slower biomass recovery, coupled with a less complex structural response, indicates that after a fire, V. caven & M. boaria may take longer to re-establish, potentially allowing for the invasion of exotic species or loss of vegetation cover, which could negatively affect the overall recovery [89,90]. Despite being a species more adapted to extreme drought conditions, V. caven shows limited post-fire recovery capacity in short terms, hindering its regeneration.

Previous studies in the Mediterranean region of Central Chile have reported long-term post-fire coverage recovery of dense vegetation by remote sensing, spanning approximately 10 years [54]. However, these studies often included mixed observations of forest, shrubland, and forest plantation, and were based on vegetation coverage rather than photosynthetic productivity. In contrast, our study has demonstrated significant differences in short-term NDVI recovery between forest subtypes. Similar or slower recovery rates have been reported in other Mediterranean regions through studies utilizing remote sensing-based vegetation indices [76,91,92,93,94,95] (Table S7). These findings highlight the importance of considering forest subtype-specific responses and using photosynthetic productivity metrics in assessing post-fire recovery dynamics.

We observed variation in forest recovery by fire severity (Figure 3, Figure 4 and Figure 5). This result can be attributed to the fire-adapted characteristics of the Mediterranean forest type. Mechanisms such as lignotubers, which are located underground, and epicormic buds, which are latent buds found on the trunk or main branches of the plant, can activate to sprout when the plant has lost its foliage due to fire [32,33]. Field observations within the study area further supported these results, demonstrating rapid vegetation regrowth following fire, even in the face of high fire intensity and extensive damage [96]. These findings highlight the resilience of Mediterranean forests to wildfires and underscore the importance of understanding the adaptive mechanisms that drive recovery in these ecosystems. Recovery was especially notable for Q. saponaria & L. caustica and C. alba & P. boldus under No Drought conditions, where significant differences in recovery rates were observed based on fire severity. However, in these Mediterranean ecosystems, once drought severity exceeds certain thresholds, fire severity becomes a less influential factor. Overall, the results emphasize the intricate interplay between fire severity, drought intensity, and forest type in shaping post-fire recovery dynamics.

Our analysis did not determine whether post-fire changes occurred in the structure and composition of the forests. We assumed that tree and shrub regeneration in the short term primarily occurred through resprouting [50,96] (Figure 7). Generally, studies measuring photosynthetic activity and post-disturbance forest productivity often lack field validations that explore the composition and structure of the forest. This omission may result from the high costs associated with analyzing large areas [16,17,97,98]. This limitation can lead to confusion in interpreting results, particularly given the differences in plant community composition [99].

In our calculation of the recovery index, we chose to use the NDVI from year one instead of the immediate post-fire NDVI from year zero. This decision was based on the potential for NDVI values to continue declining throughout the first year after a fire disturbance. Even when vegetation survives the initial burn, it may experience gradual declines during the post-fire period due to stress factors and delayed mortality [78,100,101]. Furthermore, significant shifts in species composition during this period can lead to fluctuations in NDVI values [93,102]. Additionally, changes in surface reflectance caused by the leaching of burned residues can further influence NDVI measurements after a fire [95,103].

Remote sensing tools, particularly satellite-derived spectral indices, have increasingly proven essential for assessing post-fire vegetation recovery [47,104] (Table S7). In Mediterranean ecosystems like central Chile, on-the-ground monitoring is limited [17,50]. By selecting NDVI as our primary metric due to its proven applicability in Mediterranean regions [45,93] and its effectiveness in capturing disturbances like drought [17,20], we provided valuable insights into vegetation conditions based on changes in cover density and photosynthetic activity following wildfire. However, we acknowledged that NDVI cannot detail vegetation composition and structure, highlighting the challenge of integrating field data with remote sensing for a comprehensive understanding of complex ecosystems [76,95]. Our focus on NDVI aimed to establish a foundation for future investigations that could utilize advanced remote-sensing technologies to explore specific vegetation characteristics, such as LiDAR or airborne laser scanning [36,105,106]. Recovery metrics based on remote sensing are particularly relevant for planning and managing conservation areas, as they inform adaptive management practices by identifying regions where post-disturbance recovery is slow or incomplete [107,108].

This study encountered unique challenges compared to other Mediterranean regions, concerning the fragmentation of natural forests in central Chile. The presence of forest remnants limited the spatial scale of our analysis [17,109,110]. This fragmentation resulted in a reduced number of pixels in areas with dense vegetation coverage, potentially affecting the accuracy of our analysis. Moreover, the precision of our outcomes was influenced by the inherent limitations of the official vegetation map products, which reported an overall accuracy ranging between 75% and 83% for the most recent vegetation map of the study regions [56]. However, the advent of next-generation fire remote-sensing products and model methodologies based on machine learning and artificial intelligence holds promise for enhancing the quality and efficacy of post-fire recovery studies [36,111].

Despite the projected changes in disturbance regimes due to climate change, research on post-fire vegetation recovery in the context of drought within the Chilean Mediterranean ecosystem remains scarce. Extensive research has highlighted the profound impacts of climate change on forest structure, disturbance patterns, and carbon storage, particularly in Mediterranean regions [112,113,114]. The vulnerability of central Chile’s Mediterranean forest to wildfires during prolonged droughts highlights the critical intersection between climate stressors and ecological resilience. Long-term observations suggest that climate change has reshaped natural disturbance regimes, increasing ecosystem susceptibility to severe disturbances like wildfires [6,11,115]. This is particularly evident in central Chile, where the interplay between the prolonged Mega Drought and subsequent Hyper Drought has intensified ecosystem vulnerability, compromising short-term vegetation recovery. Prolonged and severe droughts can deplete the soil seed bank and impair resprouting mechanisms, resulting in slower and less predictable recovery processes. Drier conditions may lower the threshold at which disturbances trigger shifts to alternative ecosystem states, as even minor changes in disturbance regimes can be transformative [116,117].

In this study, we adopted a categorical approach to classify drought conditions into three well-documented phases: No Drought (1992–2009) as the climate reference, Mega Drought (2010–2016), and Hyper Drought (2017–2022) [12,15,16,18,19,20]. This approach allowed us to analyze post-fire recovery across distinct drought periods. While it facilitated a landscape-level perspective on forest recovery, we acknowledge its limitations, particularly regarding potential boundary effects for fires occurring near period transitions and the absence of continuous drought measures such as the PDSI. These constraints may influence recovery trajectories, and our findings should therefore be viewed as initial insights into the complex interactions between drought severity and post-fire recovery. Despite these limitations, our findings contribute to a better understanding of recovery thresholds in Mediterranean forests and how prolonged droughts may undermine these thresholds. Understanding the thresholds of forest resilience and resistance to climate extremes, such as droughts, is crucial for forest conservation and ecological restoration efforts [108]. This study provides valuable insights into the complex interplay between drought and fire, emphasizing the need for integrating comprehensive vulnerability assessments into sustainable forest management. Given the shifts in fire regimes, post-disturbance ecosystem trajectories, and increasing aridity—especially in southern Mediterranean regions [20,118,119]—our findings underscore the critical role of post-fire climate in vegetation recovery. The future resilience of Mediterranean forests may face increasing challenges due to the rising frequency of extreme droughts projected for this region, highlighting the need for further research on the long-term implications of these changes.

5. Conclusions

The interaction between drought and wildfire presents significant challenges for the recovery of Mediterranean forest ecosystems, particularly in central Chile, where the magnitude of these disturbances has been exacerbated in the last decades. The vulnerability of central Chile’s Mediterranean ecosystems to prolonged droughts underscores the critical intersection between climate stressors and ecological resilience, emphasizing the need for targeted conservation strategies. This study highlights the urgent need for comprehensive monitoring of forest recovery dynamics in response to such disturbances. Our findings indicate that the level of drought plays a pivotal role in determining post-fire recovery rates. The observed variation in recovery rates among different forest subtypes under varying drought conditions and fire severity emphasizes the need for further investigation into the factors influencing post-fire recovery, as well as their complex interactions. These factors include species-specific traits, the duration of drought, and other ecological considerations. Understanding these dynamics is essential for forecasting future forest resilience in the face of climate change, particularly as drought frequency and intensity are expected to increase.

The insights gained from this study not only advance our understanding of Mediterranean forest recovery but also provide a valuable resource for policymakers and land managers aiming to mitigate the impacts of climate change on this forest.