Using a Cultural Keystone Species in Participatory Monitoring of Fire Management in Indigenous Lands in the Brazilian Savanna

: There is a consensus that fire should be actively managed in tropical savannas to decrease wildfire risks, firefighting costs, and social conflicts as well as to promote ecosystem conservation. Selection and participatory monitoring of the effects of fire on cultural keystone species may be an efficient way to involve local stakeholders and inform management decisions. In this study, we investigated the effects of different fire regimes on a cultural keystone species in Central Brazil. With the support of diverse multiethnic groups of local fire brigades, we sampled Hancornia speciosa (Apocynaceae) populations across a vast regional range of 18 traditional territories (Indigenous Lands and Quilombola Territories) as well as four restricted Protected Areas. We considered areas under wildfires (WF), prescribed burns (PB) and fire exclusion (FE) and quantified tree mortality, canopy damage, loss of reproductive structures and fruit production following a simplified field protocol. Areas with H. speciosa populations were identified and classified according to their fire history, and in each sampled area, adult plants were evaluated. We hypothesized that WF would have larger negative impact on the population parameters measured, while FE would increase plant survival and fruit production. We found that tree mortality, canopy damage, and loss of reproductive structures were higher in areas affected by wildfires, which also had the lowest fruit production per plant compared to PB and FE areas, corroborating our hypotheses. However, we also found higher mortality in FE areas compared to PB ones, probably due to plant diseases in areas with longer FE. Considering these results and that the attempts to exclude fire from fire-prone ecosystems commonly lead to periodic wildfires, we argue that the Integrated Fire Management program in course in federal Protected Areas in Brazil—based on early dry season prescribed fires—is a good management option for this, and likely other, cultural keystone species in the Brazilian savanna.


Introduction
Fire is a global threat to biodiversity, including when it is mismanaged in fire-prone ecosystems [1][2][3][4].Protection strategies based on zero-fire policies that tried to exclude fires from fire-prone ecosystems have debilitated the biodiversity and the provision of environmental services in topical savannas [5,6].The accumulation of cured fine fuel allows for the spreading of disastrous wildfires [2,7].These wildfires, usually concentrated in the late dry season, tend to have high intensity and cause damages to biodiversity [8][9][10][11][12].
Fire 2024, 7, 231 2 of 16 Fire ban policies have long created conflicts between governments and local communities across tropical savannas [13], where fire plays a key ecological and cultural role [1,2].
Fire management programs must be planned, monitored, and adapted to the context of each region considering ecological, cultural, and institutional aspects [2].Avoiding large wildfires and their negative consequences to biodiversity and ecosystem functioning, reducing fire-fighting costs, and decreasing socio-environmental conflicts are among the most common and important goals of fire management programs in landscapes dominated by fire-prone ecosystems [2,3,13,14].
Fire management systems that consider local culture and practices associated with fire uses and allow for participatory decision making can effectively contribute to decrease conflicts and help conserve biodiversity [15].Identifying cultural and economically important species and characterizing the impacts of different fire regimes in them might be a direct way to help increase local communities' interest in institutional fire management programs and decision-making processes.Such species may function as cultural keystone species that help to establish dialogue between environmental agencies and local communities on the complex framework of landscape fire management [16].The management of diverse ecosystems should not be decided based on the ecology or the effects of fire on a single species [17], and therefore, the common interest on cultural keystone species may help to establish common goals and boost collaboration among different stakeholders [16].
In tropical fire-prone ecosystems, many fire management programs have been institutionally established neglecting local ecological knowledge and management systems [13].Directly involving local stakeholders in the monitoring and decision-making process is an important step to break such disrespectful and conflictual management systems.In the Brazilian savanna (Cerrado), after decades of failed attempts to exclude fire from fire-prone ecosystems [3], the Integrated Fire Management (IFM) program was launched in federal protected areas in 2014 [4,18].The IFM program mainly aims to (i) decrease the occurrence and extension of wildfires; (ii) protect fire-sensitive ecosystems from fires; (iii) decrease fire-fighting costs and (iv) reduce conflicts between local communities and environmental agencies [14,19].The program has successfully changed the prevalent fire regime in protected areas from periodic (3-to 10-year intervals) large late dry season wildfires [7,13,20] to frequent, smaller, early dry season prescribed burns [15,19,21].
Cerrado traditional communities have vast ecological knowledge associated with fire to manage landscapes and natural resources [22][23][24].Regarding native fruit production, several indigenous communities discern "good fires" that cause low plant damage and might stimulate fruit production from "bad fires", which happen at the "wrong time" (season), reducing fruit production, especially due to direct damages in reproductive structures [19].Similar descriptions of the use of fire to protect or stimulate the production of native fruits have been recorded in tropical savannas in Brazil [25,26] and elsewhere [27][28][29].
The implementation of the IFM program in Indigenous Lands (IL) and Quilombola (maroon) Territories (QT) has been carried out by the Brazilian Institute of the Environment and Renewable Natural Resources (Ibama) through its Fire National Center (Prevfogo).Since the launching of the IFM program, Prevfogo has been seeking to incorporate traditional knowledge into fire management decision framework [19].For that, Prevfogo specialists carried out systematic interviews and meetings in 31 traditional territories, covering more than 10.4 million hectares in the Cerrado, Pantanal, and Amazonia biomes [19].In addition, we participatorily developed simplified field protocols that allow indigenous and quilombola brigades to perform data collection and assessment of the effects of fire regimes in their territories, following the principles of "Citizen Science" [30].
In this study, we aimed to evaluate the effects of different fire regimes in the survival and fruit production of a cultural keystone tree species, Hancornia speciosa Gomes (known as mangabeira) [31][32][33][34].Counting on a widespread network of local brigades, we considered data on 2806 individuals in 22 different protected areas under different fire regimes.
Based on traditional knowledge, we hypothesize that late-dry season wildfires cause higher canopy damage, fruit loss and mortality to adult H. speciosa trees compared to prescribed fires carried out in early-and mid-dry season.Considering the species's morphology and phenology, our second hypothesis is that both wildfires and prescribed burns increase canopy damage, fruit loss and tree mortality compared to fire exclusion.
Applying the same field protocol in such vast geographical area, on a large number of remote traditional territories creates an opportunity to acquire standardized data to inform fire management decisions.In addition, the protagonism of local peoples in the data gathering helps empower these stakeholders on the process of integrating traditional and technical-scientific knowledge into decision processes.The participatory gathering of these data, as well as the results obtained from them, can contribute to enhancing participatory decision-making processes focused on the conservation of biodiversity, traditional ecological knowledge, and the provision of environmental services provided by traditional peoples and their territories.

Study Area and Fire Regimes
Our study involved 18 traditional territories (17 ILs and 01 QT) where prescribed burns were implemented and the species H. speciosa occurred and is recognized as a cultural keystone species (Table 1 and Figure 1).Most of them are within the Cerrado biome, whereas the ILs Xingu and Pequizal do Naruvôtu are in the transition between the Cerrado and the Amazon biome and encompass Cerrado ecosystems [35].To manage these territories within the federal IFM program, 3386 prescribed burns were performed and registered in them during the study period (2015)(2016)(2017)(2018).Such landscape management with active fire use created a temporal and spatial mosaic of fire regimes and a unique opportunity to assess their effects in the ecosystems and species of interest.
Table 1.Name of the traditional territory, total area, number of burning plans (referring to each of the prescribed burns performed) registered and accumulated burnt area between 2015 and 2018, according to [36]; 6 of the 18 study areas are paired because they are contiguous and the information refers to the combined territories.To evaluate H. speciosa populations in areas under longer fire exclusion periods, which are rare in traditional territories, we sampled areas within four restricted protected areas: Chapada dos Guimarães National Park MT, Serra das Araras Ecological Station/MT, Contagem Biological Reserve/DF and Brasília National Park/DF.

Traditional
We classified fire regime of each sampled population according to the recent fire history: prescribed burns (PB) for areas where the last fire occurred between the late rainy season and mid-dry season, coinciding with the period recommended by the indigenous people for safe fire use; wildfires (WF) for areas where the last fire occurred between the late dry season and mid-rainy season, coinciding with the period that the indigenous people do not recommend the use of fire to manage Cerrado areas [25], Table 2; and fire exclusion (FE) for areas unburnt for more than 5 years (we sampled areas unburnt from 60 and 324 months, i.e., 5 to 27 years since last fire).
Table 2. Number of times per year recommended by the traditional communities for carrying out fruiting and prevention fires and the period in which the prescribed burning and wildfire treatments were defined by different ethnic groups for the 18 traditional territories within the Cerrado and Cerrado-Amazon transition regions.Adapted from [19].Precipitation patterns vary across the study region and between years.Mean precipitation for the Cerrado for 30 years before the study has been 1472 ± 154 mm/year; the precipitation during the study years for the region was, in average from 1143 mm (2015) to 1364 mm (2018, data obtained from [37].This might reflect a tendency of overall rainfall decrease in the Cerrado [38].The dry season extends generally from April to September.Local fire practices also vary across the traditional territories, according to rainfall patterns and flood pulses (Table 2).

Season
Within the IFM program, and for this study, the main goals for fire use by traditional communities were classified as for fruiting (stimulating flowering and fruit production) and preventing forest wildfires, which include the implementation of protection burns at forest edges, trail maintenance and fine fuel management.Fires that occurred outside the recommended period for the Cerrado management were considered to be wildfires [19,25], Table 2.

Study Species
Hancornia speciosa is one of the most important native plants for Brazilian traditional communities [31][32][33][34].The species's latex is used in popular medicine, as it has pharmacological properties that are being studied [39][40][41].The species's main product is the fruit, which is consumed directly and turned into sweets, jellies, and pulp.The annual pulp production recorded in northeastern Brazil was 2173 tons in 2021, generating more than USD 1.14 million in income [42].Although high, these values are likely underestimated since it is mostly marketed locally and informally [39].
Native to different regions of the country, H. speciosa is common in nutrient-poor soils.In the Cerrado, the species occurs mainly on rocky slopes in open formations with an aggregate distribution pattern often called "native orchards", which facilitate their gathering [31,33].Adult trees are usually between 4 and 6 m tall with thin bark and a wide and spread canopy.The trees are deciduous and change foliage during the driest period of the year (July to September).During the rainy season, the leaves sprout, and flowers and fruits emerge [31,34,43,44].The species has two fruiting periods per year [31,43,44]: the first one occurs between April and June, called "flower harvest" or "precocious", characterized by a low production of lower quality fruits.The second fruiting period occurs between October and December, called "button harvest", resulting in higher productivity and fruit quality [31].Hancornia speciosa fruits are also an important food source for several animal species [45,46].
Native H. speciosa orchards persist even in areas subjected to frequent and highintensity wildfires, common in protected Cerrado areas until recent years [7].There are no specific studies on fire effects on the species [41].Fruit harvesters claim that fire can kill or damage trees, affecting fruit production for up to two years due to flower and fruit damages [32], with late dry season fires having more negative impacts [33].

Field Sampling and Statistical Analysis
We sampled savanna areas (Cerrado sensu stricto) with known abundances of the target species.The sampling was carried out by crews of two to four previously trained fire brigade members, who evaluated all adult H. speciosa trees they could spot during their walk along a transect, using the rapid survey sampling method [47,48].During a standardized 20 min survey time, all H. speciosa adult trees were evaluated regarding flower and/or fruit production and canopy damage due to fire.We determined a standard time of sampling instead of other parameters, such as transect length or number of individuals, to increase the possibility of these sampling activities to be incorporated into the fire brigades' routine.We also aimed to increase the quality of the data, since the required activity was determined by time and not by the amount of data collected [49].At least one of the authors was present and supervised data collection to insure standardized forms of tree evaluation and data recording.
The loss of reproductive structures and the mortality rate were sampled only once, in the months following fire passage, while its impacts were still visible to experienced eyes familiar with this cultural keystone species.To assess whether the effects of the fire persisted over the years, impacts to tree crowns (hereafter called severity), the phenological stage, and the fruit production per plant were sampled at intervals from 01 to 12, 13 to 24, and 25 to 36 months after the last fire.In the areas under fire exclusion, the samplings were performed only once, and the time since last fire was recorded in months.
The mortality rate was calculated from the comparison between the number of registered dead trees and the total number of sampled trees.To assess the severity index, all surviving trees were classified according to the canopy damage, indicated by the presence of regrowth and dead branches (Figure 2).To assess the loss of reproductive structures, all trees that lost flowers or fruit due to fire were accounted and compared with the total number of trees in the reproductive stage.To determine the proportion of plants in the reproductive stage, the number of trees that were in the reproductive phenophases (anthesis, flowering and fruiting) was compared with the total number of trees sampled.To evaluate the fruit production per plant, all the fruit trees were classified by the local brigade members as "high" or "low" fruit production and compared with the total number of fruiting trees.
The fire regimes and time since last fire of each sampling point were identified through Prescribed Burn Plans, Wildfires Occurrence Reports and satellite images classified and treated by Ibama/Prevfogo.
The number of H. species trees varied across sampling areas, and we only considered in the analyses transects with a minimum of 10 individuals.In total, we evaluated 113 sampling transects, with data on 2806 H. speciosa trees, distributed in the 18 traditional territories and four restricted protected areas.We considered each transect as a sampling unit and use the percentage of trees in each category for the response variables analyzed, i.e., mortality, crown damage, loss of reproductive structure using poison distribution [53].We used linear mixed effect models [54] to compare the effects of different fire regimes in each of the evaluated parameter.We considered fire regimes as the fixed variable and the sampling locations as a random effect variable.The intervals since last fire were grouped within the respective fire regimes.From the results of the models, we performed the Tukey test to verify whether the response variables showed significant differences across fire regimes.We considered p values < 0.05 as significant differences.We performed all statistical analyses in R software version 4.0.0 ( [55].The results are given in mean and ± standard error.

Tree Mortality across Different Fire Regimes
Hancornia speciosa adult mortality significantly varied across fire regimes (WF: 3.76 ± 2.47% > FE: 2.21 ± 1.68% > PB: 0.47 ± 0.20%).The death of adult individuals caused by fire was a rare event, generally associated with very intense and severe wildfires, characterized as such by those that happen under extreme climatic situations (less than 30% air humidity, temperatures higher than 30 °C, after more than 60 rainless days), were extremely hard to control, spread to fire-sensitive areas and caused damage (death or topkill) even to adult trees of fire-resistant species.In this study, 71% (n = 115) of fireinduced mortality was found in the Xerente IL, in areas hit by such severe wildfires.Mortality of adult H. speciosa individuals was also recorded in areas of fire exclusion, including those areas submitted to long fire exclusion periods of between 20 and 27 years.To assess the loss of reproductive structures, all trees that lost flowers or fruit due to fire were accounted and compared with the total number of trees in the reproductive stage.To determine the proportion of plants in the reproductive stage, the number of trees that were in the reproductive phenophases (anthesis, flowering and fruiting) was compared with the total number of trees sampled.To evaluate the fruit production per plant, all the fruit trees were classified by the local brigade members as "high" or "low" fruit production and compared with the total number of fruiting trees.
The fire regimes and time since last fire of each sampling point were identified through Prescribed Burn Plans, Wildfires Occurrence Reports and satellite images classified and treated by Ibama/Prevfogo.
The number of H. species trees varied across sampling areas, and we only considered in the analyses transects with a minimum of 10 individuals.In total, we evaluated 113 sampling transects, with data on 2806 H. speciosa trees, distributed in the 18 traditional territories and four restricted protected areas.We considered each transect as a sampling unit and use the percentage of trees in each category for the response variables analyzed, i.e., mortality, crown damage, loss of reproductive structure using poison distribution [53].We used linear mixed effect models [54] to compare the effects of different fire regimes in each of the evaluated parameter.We considered fire regimes as the fixed variable and the sampling locations as a random effect variable.The intervals since last fire were grouped within the respective fire regimes.From the results of the models, we performed the Tukey test to verify whether the response variables showed significant differences across fire regimes.We considered p values < 0.05 as significant differences.We performed all statistical analyses in R software version 4.0.0 [55].The results are given in mean and ± standard error.

Tree Mortality across Different Fire Regimes
Hancornia speciosa adult mortality significantly varied across fire regimes (WF: 3.76 ± 2.47% > FE: 2.21 ± 1.68% > PB: 0.47 ± 0.20%).The death of adult individuals caused by fire was a rare event, generally associated with very intense and severe wildfires, characterized as such by those that happen under extreme climatic situations (less than 30% air humidity, temperatures higher than 30 • C, after more than 60 rainless days), were extremely hard to control, spread to fire-sensitive areas and caused damage (death or topkill) even to adult trees of fire-resistant species.In this study, 71% (n = 115) of fire-induced mortality was found in the Xerente IL, in areas hit by such severe wildfires.Mortality of adult H. speciosa individuals was also recorded in areas of fire exclusion, including those areas submitted to long fire exclusion periods of between 20 and 27 years.

Canopy Damage
The fire history significantly affected the canopy structure of H. speciosa adult trees.Populations affected by wildfires had a higher proportion of trees classified in high and very high canopy damage severity classes (high 8.19% ± 2.78 and very high 10.41% ± 3.13), in comparison to trees subjected to fire exclusion (high 2.93% ± 1.43 and very high 0.21% ± 0.21) or managed with prescribed burns (high 2.17% ± 0.95 and very high 1.56% ± 0.59).Fire exclusion and prescribed burns differed very little in severity indices (Figure 3).Populations affected by wildfires had a higher proportion of trees classified in high and very high canopy damage severity classes (high 8.19% ± 2.78 and very high 10.41% ± 3.13), in comparison to trees subjected to fire exclusion (high 2.93% ± 1.43 and very high 0.21% ± 0.21) or managed with prescribed burns (high 2.17% ± 0.95 and very high 1.56% ± 0.59).Fire exclusion and prescribed burns differed very little in severity indices (Figure 3).Hancornia speciosa populations that were affected by wildfires had a higher proportion of trees in the high (10.25 ± 3.44) and very high (22.58± 8.15) severity index classes in the first year after fire, a trend that remained constant in the second (high 3.36 ± 1.82 and very high 3.93 ± 2.43) and third years after fire (high 11.07 ± 6.72 and very high 6.73 ± 3.79).On the other hand, the areas managed with prescribed burns had a low proportion of trees in the high (3.59 ± 1.71) and very high (2.44 ± 1.03) severity index classes in the first year after fire.In addition, the few trees that were damaged recovered by the second (high 0.20 ± 0.14 and very high 0.59 ± 0.30) and third years after fire (high 0.85 ± 0.34 and very high 0.30 ± 0.29, Figures A1-A3).

Fruit Loss and Production
Hancornia speciosa trees affected by wildfires showed significantly higher losses of flowers and fruits (57.14% ± 18.70) than those managed with prescribed burns (4.52% ± 2.55).It is important to note that, due to the fragility of this species's flowers, it is likely that most evaluations have only been able to identify fruit losses.
Fire regime significantly influenced the proportion of H. speciosa trees reproducing.The lowest reproduction rates were recorded one year after fire under wildfire regime (WF 21.46% ± 7.36 and PB 31.48% ± 4.39).After that, H. speciosa trees showed a peak of reproduction in the second year after fire (WF 52.6% ± 10.49 and PB 43.69% ± 5.82).From the third year on, reproduction rates did not show significant differences between these treatments (WF 48.73% ± 9.13 and PB 46.90% ± 7.71).On average, considering all reproductive rates during the 3 years of the study, H. speciosa had the highest reproduction rates in fire exclusion areas (FE 50.50% ± 8.18; WF 41.84% ± 5.72; PB 37.36% ± 3.26, Figure Hancornia speciosa populations that were affected by wildfires had a higher proportion of trees in the high (10.25 ± 3.44) and very high (22.58± 8.15) severity index classes in the first year after fire, a trend that remained constant in the second (high 3.36 ± 1.82 and very high 3.93 ± 2.43) and third years after fire (high 11.07 ± 6.72 and very high 6.73 ± 3.79).On the other hand, the areas managed with prescribed burns had a low proportion of trees in the high (3.59 ± 1.71) and very high (2.44 ± 1.03) severity index classes in the first year after fire.In addition, the few trees that were damaged recovered by the second (high 0.20 ± 0.14 and very high 0.59 ± 0.30) and third years after fire (high 0.85 ± 0.34 and very high 0.30 ± 0.29, Figures A1-A3).

Fruit Loss and Production
Hancornia speciosa trees affected by wildfires showed significantly higher losses of flowers and fruits (57.14% ± 18.70) than those managed with prescribed burns (4.52% ± 2.55).It is important to note that, due to the fragility of this species's flowers, it is likely that most evaluations have only been able to identify fruit losses.
Fire regime significantly influenced the proportion of H. speciosa trees reproducing.The lowest reproduction rates were recorded one year after fire under wildfire regime (WF 21.46% ± 7.36 and PB 31.48% ± 4.39).After that, H. speciosa trees showed a peak From the third year on, reproduction rates did not show significant differences between these treatments (WF 48.73% ± 9.13 and PB 46.90% ± 7.71).On average, considering all reproductive rates during the 3 years of the study, H. speciosa had the highest reproduction rates in fire exclusion areas (FE 50.50% ± 8.18; WF 41.84% ± 5.72; PB 37.36% ± 3.26, Figure 4).The areas affected by wildfires had the lowest proportions of trees classified in the "high" fruit production class in the first year after fire (12.00% ± 10.73), compared to areas managed with prescribed burns (43.46% ± 5.29).The low fruit production in populations subjected to wildfires extended to the following years, resulting in the lowest average after three years of evaluation (24.07%± 5.73).On the other hand, the trees in areas managed with prescribed burns showed higher productivity than those impacted by wildfires, with no significant differences when compared to fire exclusion (PB 42.87% ± 3.68 and FE 44.25% ± 8.24; Figure 5).

Discussion
In this study, employing a citizen science approach, we gathered field data on the effects of different fire regimes in the survival and fruit production of a cultural keystone species across a large region, encompassing 22 protected areas in the Brazilian Cerrado

Reproductive trees (%)
Trees with high fruit production The areas affected by wildfires had the lowest proportions of trees classified in the "high" fruit production class in the first year after fire (12.00% ± 10.73), compared to areas managed with prescribed burns (43.46% ± 5.29).The low fruit production in populations subjected to wildfires extended to the following years, resulting in the lowest average after three years of evaluation (24.07%± 5.73).On the other hand, the trees in areas managed with prescribed burns showed higher productivity than those impacted by wildfires, with no significant differences when compared to fire exclusion (PB 42.87% ± 3.68 and FE 44.25% ± 8.24; Figure 5).The areas affected by wildfires had the lowest proportions of trees classified in the "high" fruit production class in the first year after fire (12.00% ± 10.73), compared to areas managed with prescribed burns (43.46% ± 5.29).The low fruit production in populations subjected to wildfires extended to the following years, resulting in the lowest average after three years of evaluation (24.07%± 5.73).On the other hand, the trees in areas managed with prescribed burns showed higher productivity than those impacted by wildfires, with no significant differences when compared to fire exclusion (PB 42.87% ± 3.68 and FE 44.25% ± 8.24; Figure 5).

Discussion
In this study, employing a citizen science approach, we gathered field data on the effects of different fire regimes in the survival and fruit production of a cultural keystone species across a large region, encompassing 22 protected areas in the Brazilian Cerrado

Reproductive trees (%)
Trees with high fruit production

Discussion
In this study, employing a citizen science approach, we gathered field data on the effects of different fire regimes in the survival and fruit production of a cultural keystone species across a large region, encompassing 22 protected areas in the Brazilian Cerrado and Cerrado-Amazonia transition.Following a simplified protocol, local fire brigades collected data on 2806 adult trees over a three-year period.In addition to field data acquisition in very remote areas rarely visited by researchers, these efforts promoted discussions and empowerment to local fire brigades who are responsible for the management of their traditional territories.
Corroborating our first hypothesis, wildfires caused higher mortality and damage to tree canopies and decreased the proportion of trees with high fruit production compared to prescribed burns.Despite their thin bark, Hancornia speciosa adult trees presented low mortality (<4% on average) compared to other Cerrado tree species [11,52,56,57].Such low mortality was found even in the areas hit by wildfires, which have higher flame heights and intensity, and can result in higher plant mortality [9,12,[58][59][60].
We partially rejected our second hypothesis, since trees subjected to fire exclusion had a higher mortality rate than those managed with prescribed burns, indicating that low-intensity fires may benefit this species.Many trees evaluated in areas under fire exclusion for lengthier periods lost their wide and scattered crown.This plant structure may be a consequence of intense shading due to wood encroachment [5,61].In addition, we observed that the trunks and branches of these trees were infested with pests, parasites and pathogens.Several dead and senescent trees had symptoms that resembled the disease of the "dried branches", whose cause is still undetermined [39].Many trees with leaves affected by the "purple spot" leaf disease [31] were also recorded in the fire exclusion areas, even in shorter periods of fire exclusion areas.Prescribed burns might offer a phytosanitary effect against pests, parasites, and diseases.The technical recommendations for the control of the disease are to prune off the sick branches [39], which may be similar to the result of low to mid-intensity prescribed burns.Similar phytosanitary effects of fires have been described for other culturally important tree species that were more affected by diseases and parasites in fire exclusion areas [62,63].
The relationship between burning period and intensity is not always linear [60,[64][65][66].However, large late-dry season wildfires tend to have higher intensity and cause greater negative impacts on the vegetation, even in tropical savanna ecosystems [4,7,10,12,59,65]).High-intensity fires can consume the entire aerial part of plants (topkill), keeping the individuals in a "fire trap" [67,68].In these cases, plant stature, canopy structure and fruit production are limited by regrowth rates between fire events.This phenomenon was observed several times during field sampling, especially in areas subjected to intense and frequent wildfires, such as the Paresi and Juininha IL, where short H. speciosa individuals with low fruit production dominated some of the sampled populations.Constant damage to aerial plant parts may lead plants to invest in regrowth rather than in sexual reproduction [69].
Both wildfires and prescribed burns caused decreases in fruit production in the first year after fire, as we predicted in our second hypothesis.Hancornia speciosa has a wide canopy, which usually spreads even close to the ground [31,34].Consequently, even the short flames of low-intensity prescribed burns may damage leaves and reproductive structures in the tree branches.This might explain why FE areas had a higher proportion of individuals reproducing compared to PB and WF areas.
The phenological stage of trees during fire events is determinant of fire impacts on both plant survival and reproduction.Hancornia speciosa has two flowering periods over the year-the "precocious" or "flower harvest" coincides with prescribed burns (April to June), whereas the "button harvest", which provides higher production of better quality fruits, coincides with wildfires (October to December) [31,33,34].These flowering periods help explain the more intense fruit losses recorded in areas hit by wildfires.This corroborates the reports from fruit harvesters that state the occurrence of wildfire in the late dry season is extremely harmful to fruit production for up to two years due to flower and fruit damage [33].
Our study has numerous limitations since it was performed on a large regional and environmental range.However, our results indicate that the effects of different fire regimes on this cultural keystone species coincide to the predictions based on the traditional ecological knowledge associated with it.The participatory systematic evaluation of H. speciosa populations performed during this study helped to increase the awareness of local brigades to the effects of different fire regimes on this cultural keystone species as well as to the landscapes within their traditional territories.

Conclusions
We found a larger proportion of fruiting adult trees in fire exclusion areas, which could indicate that excluding fire would be a good management strategy for increasing fruit production in this cultural keystone species.However, this result should not be interpreted alone.Since we also found mortality rates significantly higher in FE areas compared to areas managed with PB.Most importantly, attempts to exclude fire from fire-prone ecosystems, such as savannas, commonly lead to large wildfires [2,7,21,70] due to accumulation of unmanaged fuel loads.Hancornia speciosa trees hit by wildfires presented the greatest canopy damage compared to trees under prescribed burns and fire exclusion treatments, which were not significantly different from each other.
A fire regime determined by prescribed burns, considering environmental and cultural aspects within the Integrated Fire Management framework as performed by Ibama/Prevfogo seems to be the most adequate strategy for protecting H. speciosa native orchards in the studied territories.Such a fire regime has been successfully shown to reduce the occurrence of large wildfires [14,15,21].The IFM approach also reduces conflicts between government agencies and local populations [13] in addition to contributing to the conservation of ethnobiodiversity and other ecosystem services provided by the biome [71].

Figure 1 .
Figure 1.Location of the 18 traditional territories (orange) within the Cerrado biome in Brazil where Hancornia speciosa populations were sampled under different fire regimes.Populations of H. speciosa have also been sampled in four protected areas where longer fire exclusion periods could be found (red dots outside traditional territories).

Figure 2 .
Figure 2. Classes (severity indexes) of canopy damage severity, according to the % of dead branches and regrowth.Based on [50-52].

Figure 2 .
Figure 2. Classes (severity indexes) of canopy damage severity, according to the % of dead branches and regrowth.Based on [50-52].

Figure 3 .
Figure 3. Severity index (mean and standard errors) for canopy damage in H. speciosa trees subjected to different fire regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.The gradual shades in bars indicate the degree of severity, with the lighter shade being very little severity and darkest being very high severity.Different letters indicate significant differences between fire regimes (p < 0.05) according to linear mixed models followed by Tukey tests.

Figure 3 .
Figure 3. Severity index (mean and standard errors) for canopy damage in H. speciosa trees subjected to different fire regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.The gradual shades in bars indicate the degree of severity, with the lighter shade being very little severity and darkest being very high severity.Different letters indicate significant differences between fire regimes (p < 0.05) according to linear mixed models followed by Tukey tests.

Figure 4 .
Figure 4. Proportion of H. speciosa trees in reproductive stage (mean and standard errors) under prescribed burn-PB, wildfire-WF and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods since last fire (0-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences between fire regimes (p < 0.05).

Figure 5 .
Figure 5. Percentage (mean and standard errors) of H. speciosa trees with high fruit production under prescribed burns-PB, wildfires-WF, and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods after fire passage (01-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences (p < 0.05) per sampled period according to a posteriori contrast between fire regimes.

Figure 4 .
Figure 4. Proportion of H. speciosa trees in reproductive stage (mean and standard errors) under prescribed burn-PB, wildfire-WF and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods since last fire (0-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences between fire regimes (p < 0.05).

Figure 4 .
Figure 4. Proportion of H. speciosa trees in reproductive stage (mean and standard errors) under prescribed burn-PB, wildfire-WF and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods since last fire (0-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences between fire regimes (p < 0.05).

Figure 5 .
Figure 5. Percentage (mean and standard errors) of H. speciosa trees with high fruit production under prescribed burns-PB, wildfires-WF, and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods after fire passage (01-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences (p < 0.05) per sampled period according to a posteriori contrast between fire regimes.

Figure 5 .
Figure 5. Percentage (mean and standard errors) of H. speciosa trees with high fruit production under prescribed burns-PB, wildfires-WF, and fire exclusion-FE regimes in 22 protected areas within the Cerrado and Cerrado-Amazon transition regions.Values are separated into three periods after fire passage (01-12 months, 13-24 months, and 25-36 months) and the average of all periods together.Different lowercase letters indicate significant differences (p < 0.05) per sampled period according to a posteriori contrast between fire regimes.

Figure A1 .
Figure A1.Fire severity indexes (mean % ± standard error) in H. speciosa trees, between 1 and 12 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 traditional territories in Central Brazil.

Figure A2 .
Figure A2.Fire severity indexes (mean % ± standard error) in H. speciosa trees between 13 and 24 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 study areas across the Brazilian savanna.

Figure A1 .
Figure A1.Fire severity indexes (mean % ± standard error) in H. speciosa trees, between 1 and 12 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 traditional territories in Central Brazil.

Fire 2024, 7 , 17 Figure A1 .
Figure A1.Fire severity indexes (mean % ± standard error) in H. speciosa trees, between 1 and 12 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 traditional territories in Central Brazil.

Figure A2 .
Figure A2.Fire severity indexes (mean % ± standard error) in H. speciosa trees between 13 and 24 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 study areas across the Brazilian savanna.

Figure A2 .
Figure A2.Fire severity indexes (mean % ± standard error) in H. speciosa trees between 13 and 24 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 study areas across the Brazilian savanna.

Fire 2024, 7 , 231 13 of 16 Fire 17 Figure A3 .
Figure A3.Fire severity indexes (mean % ± standard error) in H. speciosa trees between 25 and 36 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 study areas across the Brazilian savannah.

Figure A3 .
Figure A3.Fire severity indexes (mean % ± standard error) in H. speciosa trees between 25 and 36 months after the last fire.The letters indicate significant differences within the same index between fire regimes according to Tukey tests.The data were collected in 18 study areas across the Brazilian savannah.