FOREST FIRE RISK ZONING FOR THE VILA VELHA STATE PARK AND ITS SURROUNDINGS (PONTA GROSSA, PARANÁ)

Forest fire hazard and risk mapping is an essential tool for planning and decision making regarding the prevention and suppression of forest fires,as well as fire management in general, as it allows the spatial visualization of areas with higher and lower ignition probability. This study aimed to develop a forest fire risk zoning map for the Vila Velha State Park and its surroundings (Ponta Grossa, Paraná State, Brazil), for the period of higher incidence of forest fires (from April to September) and for the period of lower incidence (from October to March). The following risk and hazard variables were identified: human presence, usage zones, topographical features, soil coverage and land use and meteorological conditions. Coefficients (0 to 5) reflecting the fire risk or hazard degree were allocated to each variable in order to construct the maps. The integration of these maps, through a weighting model, resulted in the final risk mapping. The very high and extreme risk classes represented about 38% of the area for both periods. The forest fire risk mapping spatially represented the levels of fire risk in the area, allowing the managers to identify the priority sectors for preventive actions in both fire seasons. Keyworks: Forest protection, fire hazard, protected areas, prevention and suppression. _______________________________________________________________________________


INTRODUCTION
The occurrence of fire influences the patterns and processes of the global ecosystem, including the distribution and vegetation structure, carbon cycle and microclimate (BOWMAN et al., 2011). According to Soares et al. (2017), every ecosystem has potentially a fire regime, defined as a set of recurring conditions of the fire that characterizes it. However, these regimes are changed (SOARES NETO et al., 2016) due to the influence anthropic through modifications in the structure, continuity, type and quantity of the combustible material, as well as as in the frequency of fires in different seasons, under different weather conditions (BOWMAN et al., 2011), resulting in negative effects (SOARES NETO et al., 2016). In this way, changes in fire have been considered a threat to the conservation of biodiversity and ecological processes (MYERS, 2006).
The term forest fire is generally used to refer to fire that spreads freely and consumes different types of vegetation, including scrub, fields and grasslands (SOARES et al., 2017). Caúla et al. (2015) emphasized the need to improve the existing control and detection systems for these events in Brazil. However, Bovio et al. (2017) pointed out that forest fire management does not end with prevention and extinction, but it also concerns the prediction of danger before the fire and post-fire reconstitution. FLORESTA, Curitiba, PR, v. 50, n. 4, p. 1818-1826, out/dez 2020. Kovalsyki, B. et.al. ISSN eletrônico 1982-4688 DOI: 10.5380/rf.v50 i4. 65974 1819 One of the fundamental instruments for this management is the Forest Fire Risk Zoning (ZRIF), which allows spatial visualization of areas according to its potential for occurrence and propagation of the fire (KOPROSKI et al., 2011). In this way, ZRIF offers support for decision making and adaptation of prevention and combat resources and equipment (SOARES NETO et al., 2016), how to intensify surveillance in areas of greater risk, restrict access to these places, build preventive fire breaks, reorganize fire management and assist in the construction of access roads to places of risk (FERRAZ;VETTORAZZI, 1998). Even so, Koproski et al. (2011) state that few Conservation Units (UCs) in Brazil have prevention and control of forest fires with established risk zoning, or even a record of occurrences of these events.
The grassy-woody steppe, or region of Campos Gerais do Paraná, is considered a fire-dependent ecosystem (MYERS, 2006). Predominantly composed of grassland vegetation, interspersed with forests galleries and vegetation islands of mixed ombrophilous forest, it presents a high risk of fire (SEGER et al., 2018). Admittedly a biodiversity hotspot, the region has high biotic potential and exuberant species richness, its main remnants are found mainly in conservation units, since these vegetal formations have been suffering strong anthropic pressure with the expansion of agricultural, livestock and silvicultural activities (DALAZOANA; MORO, 2011).Vila Velha State Park is a fully protected UC, created with the purpose of preserving one of the main remnants of grassy-woody steppe (INSTITUTO AMBIENTAL DO PARANÁ (IAP), 2004).
In addition to preventive planning for forest fires, since 2014 the park management has carried out fire management as a conservation strategy for this ecosystem (IAP, 2018). In this context, the present study aimed to elaborate a Forest Fire Risk Zoning for the Vila Velha State Park (PR) and its surroundings, in order to subsidize the fire management actions to be developed in this conservation unit.

Study area
This research was carried out in Vila Velha State Park, located in the Paraná second plateau, in the region called Campos Gerais, municipality of Ponta Grossa, state of Paraná, between the coordinates 25 ° 12 '34''to 25 ° 15 '35''S and 49 ° 58' 04 '' to 50 ° 03 '37' 'W, with altitudes ranging from 800 to 1000 m (IAP, 2004). To this study an area of influence of 1 km (buffer) from the unit boundary was considered, since in this area activities that are developed present the potential to start a fire. The total study area was 6,178.8 ha.
According to the Köppen climate classification system, the region is characterized as Cfb -temperate climate, without defined dry season. It has an average annual rainfall of 1,554 mm, with January being the month with the highest precipitation (186.5 mm) and August with the lowest precipitation (78.9 mm). The occurrence of frosts is concentrated in the period from May to August (NITSCHE et al., 2019). The main vegetable typologies are: grassy-woody steppe (23.4% of the study area), predominant in the park domain, agriculture (18.1%), located in the area surrounding the unit, and mixed montane rain forest (16.1%).

Data gathering
For the Vila Velha State Park, two forest fire risk zones were made, one for the period of high incidence and another for the period of low incidence of forest fires. These were defined based on information from the municipality of Ponta Grossa (PR), considering monthly values from 2006 to 2014, through a cluster analysis, considering the following variables: number of days with precipitation ≥ 2.4 mm, average precipitation (mm), average relative humidity (%), average temperature (° C) (PARANÁ METEOROLOGICAL SYSTEM (SIMEPAR), 2015); number of days with fires occurring, number of occurrences of fires (SYSBM-CCB, 2015) and average value of the Monte Alegre Formula (FMA).
The park's fragility in regarding forest fires was quantified and qualified through the analyzes of the following parameters: human presence (infrastructure and areas of use), slope, altimetry, slope orientation , vegetation cover and weather conditions. Thematic maps referring to declivity, altimetry and slope orientation were elaborated based on the Digital Elevation Model (MDE), of sheet 25S51, made available by the TOPODATA project (NATIONAL INSTITUTE FOR SPACE RESEARCH (INPE), 2011), which has a 30 meter resolution. Thematic maps of human presence and vegetation cover were based on information made available by IAP and classification of use by means of RapidEye satellite images (MINISTRY OF THE ENVIRONMENT (MMA), 2013). Thematic maps, referring to meteorological conditions, were made based on the indexes obtained from the Monte Alegre Formula (FMA) adjusted for the region by Kovalsyki (2016).
The ZRIFs were obtained by superimposing thematic maps, according to a model of weighting for Forest Fire Risk (RIF). All spatial analyzes and cartographic production were carried out using a geographic information system with the aid of ArcGIS 10 software.

Fire hazard and risk maps
To represent the risk according to human presence, maps were prepared based on the infrastructure and in the use zones. For risk maps according to infrastructure, the following factors were considered: highway, FLORESTA,Curitiba,PR,v. 50,n. 4,p. 1818 -1826, out/dez 2020. Kovalsyki, B. et.al. ISSN eletrônico 1982-4688 DOI: 10.5380/rf.v50 i4. 65974 1820 railroad, roads, fire breaks and buildings in the park areas. Ferraz and Vettorazzi (1998) considered for means of access, an area of influence proportional to the flow of people and vehicles passing through it. This way, influence radii were stipulated, for which coefficients were attributed. These coefficients reflect the risk level, and are based on the literature, as shown in Table 1, with the domain of roads and fire breaks classified as null risk (coefficient 0). For topographic characteristics, aspects related to the slope of the land were analyzed, altimetry and orientation of the slopes, and the hazard coefficients that were determined according to Batista et al.(2002) (Table  3).  In this classification, zero hazard was attributed both to the water category and to the infrastructure (roads, buildings and firebreaks), since these act as a physical barrier for the spreading of fire. The class "Extreme" corresponds to areas with combustible material known as dangerous, referring to the thin material which, under natural conditions, proves to be easy and quick to burn.
The meteorological conditions were represented by means of the fire hazard index of the Monte Alegre Formula -FMA (SOARES et al., 2017) adjusted for the region, as this reflects the probability of occurrence, based on atmospheric conditions. FMA was calculated daily with meteorological data provided by SIMEPAR (2015) for the period between 2006 and 2014 and, through the values obtained, the average for two scenarios: a) months that cover the normal fire season; and b) months that cover the least favorable period of incidence. For each class of the FMA, a coefficient referring to the degree of fire hazard, ranging from zero (null) to 4 (very high) was attributed.
The weighting model used, was adapted from Batista et al. (2002), since this was indicated by the same authors for the elaboration of fire risk zoning in the state of Paraná, due to the number of variables used and because it was elaborated using climatic conditions similar to those of the study area. The classification attributed to zoning included six classes: null, low, moderate, high, very high and extreme, established by the quantile method for each of the scenarios. For the purpose of validating the risk assessments of forest fires, we used the burn scars of areas burned between 2015 and 2018, through the visual interpretation of the false color composition of the Landsat images, from the United States Geological Survey (USGS) Earth Explorer database. For sensors ETM + the Red-Green-Blue composition used was: band 5 (1550-1750 nm), band 4 (760-900 nm) and band 3 (630-690 nm); while for OLI sensor: band 6 (1570-1650 nm), band 5 (850-880 nm) and band 4 (640-670 nm).

RESULTS
It was observed that the period of high incidence of forest fires in the municipality of Ponta Grossa goes from April to September, while the period of low fire incidence was from October to March.

Thematic maps
Thematic maps of risk and danger of forest fires for the Vila Velha State Park and its surroundings are represented in Figure 1. Note: (G): meteorological condition for the period of low fire incidence and (H) for the period of high fire incidence. Table 5 shows the representativeness of each class of danger and risk of forest fires for the overload factors. Slope and slope orientation factors were highlighted due to their representativeness in the low danger, 72.5 and 44.4%, respectively. On the other hand, the factor Areas of use and coverage and land use stood out for the representativeness of the very high and extreme classes, with 50.1 and 61.4%, respectively. Regarding infrastructure and their respective interference radii, 81.7% of the area presented zero risk, that is, 18.3% of the study area is under human influence, of which approximately 48% fell into the risk class very high. The altimetry covered two classes of danger, with the very high class being the most representative (74.7%). It was observed for both periods that the high and very high risk classes were concentrated to the east of Vila Velha State Park, while the moderate and extreme classes were concentrated to the west. Table 6 shows the representativeness of each class of forest fire risk for the two scenarios, both for the area inside the park, as well as the area surrounding the park. For both scenarios, the null risk class represented 1.1% of the study area, referring to the road, lagoons and furnas, while the extreme risk class was concentrated in the areas of influence of the road network.

Forest fire risk zoning
Considering the area of the park for both periods, the low risk class stood out, which had approximately 30% representativeness. This class covered areas of restricted access in which the vegetation had higher moisture content, such as the mixed montane rain forest, mixed alluvial rain forest, pioneer formation of fluvial influence, hygrophile steppe, in addition to areas with rock formations. However, considering the total study area, this risk class had an area reduction of 15.7% for the discharge period incidence of forest fires. Approximately 15.5% of the areas that previously presented low risk, presenting moderate risk.
The very high risk class covered approximately 26% of the area surrounding the park for the two periods, which included areas of agricultural crops and areas of countryside vegetation. The very high risk class had greater representativeness in the period of high fire incidence for the study area, 21.8%, an increase of 9.7% in relation to the other scenario.
For the period of low incidence of forest fires, 51% of the area of Vila Velha State Park fit into the highest fire risk classes (from high to extreme), while for the high incidence of fires, this representation rose to 55.2%.
During the 4-year period, 11 polygons were found with burn scars or forest fires in the study area. In the period of low incidence of forest fires, a total of 163.7 ha, distributed among classes of low risk (5.8%), moderate (30.8%), high (44.1%), very high (6.3%) and extreme (13%). In the period of high incidence of forest fires, 599.2 ha were distributed among the classes of low risk (7.2%), moderate (5.0%), high (30.8%), very high (35.4%) and extreme (21.6%).

DISCUSSION
The validation of forest fire risk zoning was based on burn scars, as 81.8% of these had an area of less than 100 ha, this fact that made it impossible to use hotspots for validation, considering that Pereira et al. (2012) found a failure to detect 81% for areas burned up to 100 ha. For the period of low fire incidence, two scars were observed, which reached predominantly high and moderate risk classes (74.9%), while for the period of high incidence of fires were demarcated in nine areas, which predominantly encompassed the classes of high, very high risk and extreme (87.8%). The recurrent area (burned in January 2015 and August 2017) totaled 40.4 ha, sorted into moderate (6.4%), high (63%), very high (1%) and extreme (29.6%) classes, with the highway margins being the probable starting point for the 2017 fire.
The most influential factor in the risk of forest fires was the human presence, a fact corroborated by other studies, which presented a very high or extreme degree of risk in areas of influence of the road network, such as de Ferraz and Vettorazzi (1998), Koproski et al. (2011) andTetto et al. (2012). Prudente and Rosa (2010) mention that proximity to roads and urban areas contribute to the occurrence of forest fires, the probable ones being places where the fire started, mainly for criminal causes, cigarettes, religious services and bonfires. Another important factor observed was the type of vegetation. The rural typology areas stood out in terms of both scenarios, due to the high risk of forest fires in these locations. According to Seger et al. (2018), the combustible material of grassywoody steppe is classified as dangerous, since it is composed mostly of thin and dry material. According to the