1. Introduction
The severity and impacts, at the regional and global scales, of the burned area resulting from wildfires have increased over the last few decades [
1]. Climate change has been identified as an important factor that contributes to these increases [
2,
3,
4,
5]. Climate change affects wildfires, both directly, through the weather conditions that affect fire ignition and propagation, and indirectly, through its effects on vegetation and fuels [
6].
Another type of factor responsible for the increased impacts and severity of wildfires identified by other authors was the socioeconomic factor. For example, the abandonment of crops or agroforestry land, the disinvestment in the forestry sector, and other human practices also contribute to the increase in biomass fuel [
7,
8,
9,
10,
11,
12,
13].
Wildfires cause many environmental problems because the physical–chemical and biological properties of soil are affected, the aerosols and other particles emitted during combustion cause air pollution, and the pollutants deposited on the surface of burned areas’ soils can be transported by water and pollute water bodies [
14]. The vegetation, beyond all the benefits it brings to the environment, is of utmost importance for attenuating some of these environmental problems, in particular the soil erosion and water contamination factors mentioned above.
In general, ecosystems are partially modified when a wildfire occurs, the vegetation is burned partially or totally, and the biological activity is perturbed [
15]. The natural regeneration of vegetation after a fire depends on the soil condition (the severity of the fire is important) [
16,
17,
18,
19], the vegetation species present (growth by the rhizomes or not), the bank of seeds available (for germination in the burned areas) [
20], and the weather conditions or climate characteristics.
Wildfires are a big problem in the Mediterranean region due to their social, economic, and environmental consequences [
21]. Many landscapes in this territory have been affected by fire in the past, but fire is considered a natural factor with an important influence on the biological productivity and composition of several ecosystems [
20]. In fact, fire creates open areas, which favors the germination of species by removing the established vegetation, and has direct effects on germination and seed survival [
22]; fire may facilitate the germination and development of several species by changing the mineral environment [
23]. Some plant species are resilient to fire, such as
Quercus faginea, while others are more easily consumed by fire due to their flammable substances, e.g., pine resin. Portuguese forests have experienced major changes in the past few years, mainly because of the conversion to eucalyptus, which can be explained by the increasing demand for this raw material by the cellulose industry. These changes have had an impact on the fire regime of this territory: the events are increasingly intense, occur more frequently, and the extent of the burned area is increasing because of the difficulty of extinguishing the fires [
24].
Many Portuguese wildfires are recorded every year, resulting in a large burned area [
24]. The severity of fire in this territory is often high [
25] because the weather conditions provide excellent conditions for fire propagation due to the particularly high temperatures, low relative humidity, and significant wind speed. Conversely, the high availability of combustible biomass fuel also contributes to fire intensification, and these factors sometimes result in uncontrollable wildfires. The climate in Portugal promotes the occurrence of wildfires, as it has a rainy season during the spring, which is favorable for the development of vegetation, followed by a very warm period that triggers the development of large wildfires [
26].
The assessment of vegetation recovery in burned areas is crucial in land management [
27]. This assessment is important to prevent or remediate different environment impacts, but vegetation recovery is not always a positive factor in the context of wildfire recurrence because these areas are then particularly susceptible to future wildfires. However, vegetation recovery assessment is complex because it depends on a variety of biological and environmental factors and on the interaction between them [
15]. Post-fire regeneration depends mainly on the initial vegetation and on environmental factors—climatic and terrain characteristics—present onsite [
28].
Several methodologies have been proposed for vegetation assessment, namely, methodologies based on monitoring the vegetation state from spectral indices [
17]; for example, the Normalized Difference Vegetation Index (NDVI) [
29,
30,
31], the Soil-Adjusted Vegetation Index (SAVI) [
32,
33,
34], the Leaf Area Index (LAI), the Fractional Vegetation Cover (FVC) [
35,
36], the Regeneration Index [
37,
38,
39], the Normalized Difference Infrared Index (NDII) [
34,
40,
41,
42], and Spectral Mixture Analysis (SMA) [
43,
44]. Other spectral indices, such as the Normalized Burn Ratio (NBR) and Enhanced Vegetation Index (EVI), which combine and extract useful information from several spectral bands [
30], have also been widely used to study fire-induced vegetation changes, including burn severity [
45] and regeneration dynamics [
27,
30,
32,
46,
47]. In recent studies, the factors that determine post-fire regeneration patterns were investigated, and this analysis integrated the improvement of the predictive models of vegetation dynamics [
39,
48,
49,
50,
51,
52,
53].
The recurrence of wildfires in Portugal is higher, in particular in the central region [
24], where few preventive measures have been adopted in forest areas (for example, the construction of paths through the forest for access by firefighters, the obligation of owners to clear the forest, etc.). In these areas, the natural growth of vegetation without restrictions and the biomass fuel available can induce new, large wildfires. In this sense, the vegetation recovery assessment assumes great importance for the management of these territories, particularly the creation of preventive and reactive measures to apply in forest areas. This is a premise advocated by the research community—for example, Lentile et al. [
54] found that indicators of burn severity, and thus potential ecosystem recovery, could prove useful to post-fire planners tasked with strategically rehabilitating areas likely to recover slowly or in undesirable ways.
The main goal of this research is the assessment of vegetation recovery in burned areas of central Portugal (study area,
Figure 1) through remote sensing data, using Landsat 8 OLI images, and biophysical data. The assessment of vegetation recovery was performed differentially by vegetation type (low- or high-growing vegetation and predominant species) and biophysical data (for example: precipitation, temperature, insolation, relief characteristics, etc.).
3. Results
3.1. Burned Area Distribution by LUC
The wildfires of 2017 devastated large forest areas, but many cropland areas were also affected, for example, olive groves or nonperennial crops that are rainfed and irrigated, as shown in
Table 3 by cross-referencing burned areas with LUC (pre-fire). However, three LUC types were predominant in burned areas:
Pinus pinaster forest, eucalyptus forest, and shrubland. These three LUC types accounted for 81.9% of the burned area in the study area in 2017. Other LUC types were also affected by the wildfires (e.g., agricultural crops or urban areas), but are not representative and were grouped under “Other land cover types”.
3.2. Burned Ratio Index
Wildfires occurred in different months, but large wildfires occurred mainly in the summer and at the beginning of the autumn (a period with reduced seed germination capacity and reduced vegetative growth). The short period between these large wildfires did not allow for the regeneration of vegetation in the burned areas. This was observed during field work, where the growth of vegetation was evaluated.
The severity of fire is spatially diversified; for example, the wildfires of October show areas with high fire severity along with other areas with less severe effects. This differentiation was observed particularly in the first NBR results after the fires (
Figure 4), where the central burned areas presented higher values of NBR.
3.3. Differenced Burned Ratio Index
In a short period after the occurrence of wildfires, it was observed that the vegetation did not recover or recovered very little, as can be seen from the analysis of the dNBR
P1 (
Figure 5). The last large wildfires occurred in October 2017 (
Figure 1) and the time that elapsed until the end of the period of dNBR
P1 was very reduced; however, besides this conditioning factor, we must also take into account the season (autumn), as the vegetation does not regenerate easily due to the climatic conditions (progressively cooler temperatures and first rainfalls).
In the following year after the fire, particularly after spring, we observed the regeneration of vegetation in large patches of burned areas; this is shown in the dNBRP2 values, with a greater emphasis on burned areas further north where shrubland predominates. This recovery was slowed by the arrival of autumn (dNBRP3); the recovery resumed and was more marked from February 2019 (dNBRP5), the year in which the regeneration of much of the vegetation in the burned areas stands out, with the exception of pine forests along the coast (dNBRP6).
Analyzing the results of each dNBR, we saw variations concerning the statistical description (
Table 4), with the highest average value in the dNBR
P1 because it was the shortest post-fire period. The lowest value occurred in the dNBR
P6, being negative in this case. These results indicate a greater recovery of the vegetation in this last period, with some areas of more vegetation compared to what occurred before the fire.
The spatiotemporal distribution of dNBR is distinct depending on the post-fire period. The vegetation’s regeneration is more pronounced when considering a longer period, hence the dNBR
P6 is not similar to the remaining results of dNBR (
Figure 6). In the burned areas in the interior, especially in the north of the study area, the vegetation recovery process is more accentuated (areas with negative dNBR), which means that the post-fire NBR values are higher than those observed pre-fire.
The results of the dNBR allowed us to distinguish the regeneration of the vegetation according to the season of the year. In this case, with the transition from winter to spring, the vegetation had a greater regeneration capacity, as shown by the transitions from P4 to P5 and P6.
The longer the period after the fire, the greater the recovery of the vegetation should be. Thus, it is important to analyze in greater detail the dNBR with the longest period—in this case, the dNBRP6.
Certain vegetative types have greater recoverability capacity, a fact observed, for example, in herbaceous vegetation and oak forest (
Figure 7, negative values). In this recoverability, the phenological processes of the vegetation must also be considered, in particular during the spring season, when the environmental conditions (precipitation and temperature) lead to greater development of most vegetation.
In the areas of Pinus pinaster forest, eucalyptus forest, and shrubland, the vegetation present smallest recovery periods. In these cases, the vegetation is essentially arboreal and shrub, and the recovery process differs depending on the size of the vegetation: faster in the shrubs vegetation where the vegetal composition is a differentiating factor (including this class several species that naturally recover after fire, without human intervention), while arboreal vegetation needs much more time to recover.
The recovery also differs according to the forest species present, Pinus pinaster takes longer to regenerate compared to eucalyptus if there are conditions for the regeneration process of Pinus pinaster to occur. When the recurrence of wildfires is high in the pine forests, the availability of seeds for regeneration is compromised because the pine-trees do not have enough time to grow and produce new seeds.
Certain LUC types show a behavior very similar in terms of dNBR, and this is highlighted in the correlation coefficients of dNBR presented in
Table 5. For example, shrubland presents high correlation with herbaceous vegetation, and eucalypt forest with
Pinus pinaster forest.
The results of principal component analyses show different groupings of LUC classes of dNBR, specifically forest species with a minor area of representation (in absolute terms), for example forests of cork oak, holm oak, chestnut,
Pinus pinea, and other conifers. Associated with these LUC classes, other LUC classes are affected by wildfires, in particular the crop land cover types (olive groves and agroforestry). LUC classes with more area affected by wildfires tend to group in the PCA at higher linkage distances (
Figure 8). These results reflect the different regeneration capacity in time of the different species (
Pinus pinaster, eucalyptus, and shrubland).
3.4. The Importance of Biophysical Factors on Vegetation Recovery of Burned Areas
The most important factors affecting the vegetation recovery in burned areas are climate factors and soil characteristics. This is shown by the positive correlation of the analyses with dNBR
P6 (
Table 6).
Temperature is the most important independent variable in vegetation recovery, with a higher positive correlation value between environmental biophysical variables and dNBRP6. These results, analyzed in detail, indicate that lower temperatures influence the vegetation recovery; this is more evident in higher areas, where the temperatures are in general lower compared to areas with lower elevation.
Precipitation is also important in the vegetation recovery process, and this factor is also influenced by the topography of the study area. Burned areas located in the interior and to the north of the study area have more precipitation compared to the rest of the study area. Both latitude and relief also have an important influence on the recovery process.
The continentality gradient is also important in this process. Indirectly, this factor reflects the influence of the prevailing winds from the north and from the Atlantic Ocean that interfere with temperature, precipitation, and air humidity content, all of which are important factors in the germination of seeds and plant growth.
The soil type and some physicochemical characteristics are also important factors. Vegetation recovery is most efficient in humic cambisols and in soils with reduced pH.
In contrast, topographic factors present lower correlation values, although positive, indicating that they also influence the vegetation recovery in burned areas.
The aspect of hillsides is the most important topographic factor on vegetation recovery in burned areas. Hillsides exposed to the north and northwest have more efficient vegetative growth due to the exposure of these hillsides to precipitation and to masses of air rich in humidity that prevail in these directions; for hillsides facing south, insulation is a more important factor and can explain the low humidity in the soils and air, which interfere with vegetation growth.
The elevation has a lower positive correlation. The burned areas with high elevation do not present high regeneration of vegetation post-fire because these areas have low temperatures and are more exposed to the wind (which interferes with the growth of plants). It was also observed that, in areas with low elevation, sheltered from the winds, and with mild temperatures, the regeneration of vegetation is more efficient.
4. Discussion
In the burned areas, the analyzed vegetation recovered more from wildfires that occurred between June and August. The fire severity in these burned areas was lower relative to those areas affected by the wildfires of October. However, the recovery process of burned vegetation was more pronounced here, first, as a result of the species present pre-fire that regenerate relatively fast post-fire (e.g., eucalyptus and shrubland—Genisteae), and second, due to the time between the date of the wildfire and the date of the first Landsat 8 Oli image (post-fire), with more regeneration of vegetation in the areas burned by the first wildfires, relative to those areas affected by the wildfires of October.
Some studies found that regeneration of vegetation occurs within the first two years following fires [
68,
69,
70]. This fact was observed in this research, where some areas have more vegetation relative to the pre-fire situation (especially shrubland), i.e., in these cases, it took approximately 19 months to regenerate all the vegetation. Cerdà and Doerr [
64] found that post-fire vegetation recovery in the Mediterranean territory can be rather rapid due to the adaptation of vegetation to disturbance by fire and, following burning, the low competition for sunlight, nutrient availability, and reduced water losses by transpiration. Other post-fire forest floor conditions such as the amount of bare soil exposed are also important determinants of post-fire vegetation recovery [
54].
The vegetation recovery dynamics in burned areas is different according to the species present before the fire and the vegetation size (trees, herbs, and shrub). Eucalyptus resprouts readily after fire [
71], so eucalyptus forest regenerates relatively quickly, and this growth was observed in the NBR results. However, in areas with this vegetation type, other vegetal species are also present, for example herbaceous or shrub species. The mixed vegetation is important regarding fire severity, especially when species with lower flammability are present [
72,
73]. The burned areas analyzed were composed of diversified vegetal species (before the fire), although there are predominant species (e.g.,
Pinus pinaster or eucalyptus), also present are the species with lower flammability mentioned above (
Quercus rotundifolia,
Q. robur,
Q. faginea,
Q. pyrenaica, and
Q. suber), which contributed to the spatial differentiation of fire severity (which is reduced) [
74], a factor that also influences the vegetation recovery process [
17,
30,
39,
47]. Although some species of
Quercus can be flammable, they are well adapted to fire. For example,
Quercus suber L. is the only Mediterranean species that can regenerate from the crown, even in the case of severe fires, allowing it to recover more quickly than other species. In forests of
Quercus faginea (a minimally flammable species), few trees die due to fires (about 5%) [
75].
NBR results provide a means to differentiate the severity of wildfires [
46], with emphasis on the fire severity that affected the burned areas resulting from the October wildfires. The weather conditions were the primary factor, according to reports from the Portuguese civil protection services, for the high fire severity observed, specifically the high temperatures, wind speed, and low relative humidity (ranges for these variables are available in the report of IPMA [
76]).
The negative values of the last dNBR observed in the northern part of the study area indicate more regeneration of vegetation relative to other burned areas. The fire severity in these areas was high, but the vegetation present pre-fire was essentially shrubland, composed of species that regenerate easily after fire (e.g.,
Ulex and
Genisteae), thus explaining the results obtained. In this sense, we highlight the importance of vegetation resprouting in the burned areas because resprouting provides persistence under disturbance [
77] and enables more efficient recovery of the vegetation land cover [
54,
78].
In contrast, the ashes deposited in the soil surface after a fire provide nutrients that are available to the plants, with this ash representing a substantial part of the nutrient stock [
79]. This factor could also have contributed to the efficiency of vegetation regeneration, but seeds and/or vegetal species that grow from rhizomes must be present. However, losses of nutrients by ash convection (conversion of elements to solid, inorganic ash subject to wind action) increase with fire intensity [
80].
Nutrients resulting from the burning of forest material are available in burned areas if they are not lost by direct volatilization or ash convection [
80]. The ash deposited on the surface of soils (a process in which low-intensity fire has an important role) constitutes a factor to provide nutrients to the plants and the sustainability of the forest [
81]. In this process, the amount of precipitation that falls on the burned areas is an important factor to take into account. When the water retention capacity of the soils is exceeded, surface runoff occurs and these nutrients are carried away by the water [
14,
82]. This factor can affect the recovery of vegetation on the burned areas resulting from wildfires, especially those that occurred in October, because rainfall in the following months is high. In the burned areas resulting from the wildfires of May, June, and July, some plants had already germinated and contributed to the reduction of these losses.
The influence of the seasons on the vegetation recovery is very clear when analyzing the dNBR results. The biggest vegetative growth occurs mainly during the spring, characterized by the mildest temperatures and the availability of water (in soil and in terms of air moisture), factors that allow for the development of plants [
83,
84,
85]. The correlation obtained between dNBR and environmental biophysical variables point to the importance of climate characteristics for vegetation recovery, especially temperature and precipitation. These results are in accordance with other studies that examine the influence of these variables [
28,
43,
50,
51]. Additionally, the continentality gradient plays an important role in the recovery process because it also indirectly reflects the spatial distribution of precipitation and temperature.
A limitation of using NBR to calculate the vegetation recovery index has to do with the identification of those areas where some forest species are predominant and to distinguish them from areas covered with other vegetation with fast growth rates after fire. For example, eucalyptus and Pinus pinaster forests recover at different times, as observed in the results, but the first vegetation appearing after the fire in this forest areas is herbaceous; this vegetation type is also considered during the NBR calculus.
The soil type is also important for vegetation recovery. The results point to the importance of this variable in the recovery processes, and, when analyzed in detail, the most important relationship was observed between humic cambisols and the vegetation recovery process. This soil type has an umbric A horizon, which is thicker and rich in organic matter—favorable factors for plant recovery. However, the soil properties can be affected by fire severity [
86,
87].
Soil
p also has a positive correlation with vegetation recovery. The pH inevitably increases due to soil heating as a result of organic acid denaturation, with significant increases at high temperatures (>450–500 °C) [
88]. This factor favors vegetation recovery and can explain the results obtained.
Topographic factors will have different impacts on the regeneration of different forest cover types [
17,
39]. Effectively, the results show that relief characteristics influence the vegetation recovery, but they are not the most important factor in the recovery process. Several studies found an influence of slope in this recovery; for example, Shakesby [
89] found that steeper slopes tend to retain less water and be more susceptible to soil erosion and nutrient loss, which explains why most soils in these slopes are thinner and less productive. Cerdà and Doerr [
64] observed differential recovery rates on north- and south-facing slopes.
The results of this research are based on the assumption of the importance of spatial planning actions in forest areas affected by wildfires. Usually, the planning actions do not take into consideration the potential regeneration of the vegetation in burned areas, but assume, at least in some cases, that the type of LUC existing pre-fire continues unchanged. Thus, the actions resulting from planning have to be reconsidered, especially in the strategical rehabilitation of burned areas [
54]. The previous assumption is used, for example, in the creation of the official LUC maps of Portugal. The burned areas identified during cartographic creation do not present the same vegetative regeneration index as shown in this research; in some areas, the forest does not regenerate because the species present before the fire, when burned, do not recover partially or completely (e.g.,
Pinus pinaster).
5. Conclusions
The vegetation recovery process does not occur in the same way in the burned areas of the study area. This fact points to the interference of different factors, such as the severity of fire, the vegetation species, and the environmental biophysical characteristics of these areas.
With the computation of dNBR from the NBR product at different times (obtained by Landsat 8 OLI images), the dynamics of the post-fire vegetation recovery were assessed; different spaciotemporal dynamics were observed and different patterns were found. The recovery was more marked in some burned areas, especially in burned areas located in the interior and to the north of the study area, but was also influenced by the season of the year.
The vegetation species present in the burned areas pre-fire are important in the process of vegetative recovery because some species regenerate more quickly compared to other species—for example, eucalyptus regenerates in a short time compared to coniferous, oak, or invasive species, and some species also regenerate by the rhizomes.
Climate conditions (temperature and precipitation), continentality gradient, and the properties of soils (type and pH) are great influences on the vegetation recovery of the burned areas assessed. Topographic factors also influence this recovery, but are not the most important variables in this process.
The results of this research are important to spatial planning actions, in particular for forest planning: first, in the recovery of ecosystems; second, to prevent the occurrence of new large wildfires by adopting new strategies of forestry and the modification of anthropogenic practices (e.g., reforestation with native species resistant to fire—for example, Quercus faginea, Quercus ilex, and Quercus rotundifolia; changes in crop activities that use fire to clean the fields, among others). Thus, we stress the importance of the recurrence of fire to be evaluated, especially in areas with certain plant species that are sensitive to fire and species that do not recover integrally (e.g., a high recurrence of fires in pine forests does not allow for the growth of these species until they start producing seeds).