Assessment of vegetation regrowth and spatial patterns and severity factors of wildfires in wildland-urban interface — the case of the large wildfire in Baião (2019)

Portugal is one of the Southern European countries most affected by forest fires, with recurrent events and frequent impacts. The demographic and social changes that have occurred in rural areas have driven land abandonment in recent years, which, in turn, influences forest management and wildland-urban interface (WUI) areas that are related to fires. The aim of this work is to develop a case study in the municipality of Baião based on the large wildfire (LWF) of 2019, by defining and mapping the WUI areas, evaluating the LWF recurrence and severity, and the vegetation regrowth in a period of 2 years in areas with different land uses affected by distinct fire severities. The study was organized into 4 stages: firstly, we mapped the fire occurrences and then the WUI areas. After fire mapping, the recurrence of wildfires was characterized and the last step corresponded to the evaluation of the severity of the LWF of 2019 and the evaluation of the vegetation regeneration, and different land use types. The WUI represents 26.7% of the territory of the municipality of Baião. From 2001 to 2021, the municipality registered 3,770 fire occurrences. In 2019, the LWF of Baião burned an area corresponding to 853 ha, which was affected by a total of 12 fires in the period between 1975 and 2019, resulting in a maximum degree of 11 recurrences for the same area. With regard to the direct impact of fire on vegetation and its subsequent recovery, we can see that 2 years after the LWF, the area occupied by the forest and shrub vegetation, which were hit by high severity, already showed significant levels of vegetation regrowth. As a main conclusion, the study contributes to widen the understanding of the patterns created by fire in different landscapes, and this information is valuable for forest managers to know the consequences (beneficial or not) and be able to plan prevention, restoration, and environmental education actions.


Introduction
In recent decades, fire regimes in Europe have changed due to the increase in the occurrence of extreme events of large wildfires (LWF), mainly in the south of the continent (Lopes et al., 2022;Turco et al., 2016). In the Mediterranean region, despite a downward trend in the annual burnt area (Lopes et al., 2022;Turco et al., 2016), LWFs have been considered one of the main disturbances in the environment with large burnt areas and growing socioeconomic and ecological impacts becoming more frequent lately (Tedim et al., 2015;Doerr & Santín, 2016;Fernandez-Anez et al., 2021;Guo et al., 2022;Lopes et al., 2022), because of a complex interaction of high temperatures, prolonged droughts, land abandonment, and fire suppression policies that have led to increased fuel loads (Fernandes, 2013;Ferreira-Leite et al., 2016;Lopes et al., 2022;Moreira et al., 2020).
Portugal is one of the countries most affected by recurrent wildfires, with significant impacts that are, being responsible for losses in the environment, economy, and human lives (Lopes et al., 2022;Guo et al., 2022;Parente et al., 2018;Meneses et al., 2018;Ferreira-Leite et al., 2017). M ore over, demographic and social changes in rural areas have caused land abandonment over the years, which, in turn, influences forest planning and wildland-urban interfaces (WUI) (Barbosa et al., 2022).
The WUI is an area where houses or urban areas meet or intermingle with wild vegetation or rural areas, and are prone to fires (Lampin-Maillet et al. 2009;. In Portugal, the wildfires in WUIs are gaining relevance, especially in the 21st century, and with particular relevance after the tragic years of 2003, 2005, 2013 and 2017 (Bento-Gonçalves & Vieira 2020).
In 2019, the third largest wildfire in mainland Portugal struck in the municipality of Baião (ICNF 2022), located in NUT 1 III Tâmega e Sousa, in the north of the country. In Baião, some areas of its territory have WUI, largely resulting from the rural abandonment of agricultural and horticultural fields around the villages and houses, a pattern very much associated to aging population, favouring an increase of unmanaged forest and the prevalence of, natural growth, in areas closer to houses (Doerr et al., 2017).
In this context, the use of satellite images enables an efficient study of fire events (Brown et al., 2018;Roy et al., 2013), contr ibuting to a perspective of analysis of pre and post-fire conditions such as the one that happened in Baião in 2019 (Santos et al., 2020). Data from the Sentinel satellite system of the European Space Agency (ESA) allow the in wildland-urban interface -the case of the large wildfire in Baião (2019) nº 47 -2023 assessment of wildfires impacts in greater spatial and spectral resolution provided by medium resolution and open access earth observation systems (Brown et al., 2018;Whyte et al., 2018;Huang et al., 2016).
After a fire, several changes take place in the environment as it burns the vegetation, leaving the soil partially or completely bare, modifies the moisture content, and ash deposition occurs (Parker et al., 2015;Pausas & Keeley, 2009;Santos et al., 2020;Veraverbeke et al., 2010). Remote sensing applications offer viable approaches to describe the changes as they contribute to detect them in the fire-affected landscape (Parker et al., 2015;Santos et al., 2020Santos et al., , 2023Sunderman & Weisberg, 2011;Veraverbeke et al., 2014

Study area
The study area is in Baião, the easternmost municipality in the Porto district, in northern Portugal.
Baião has 17,452 hectares and is located on the Douro -Tâmega interfluve, an area considered to be undergoing geomorphological and even economic transition, between the regions of Trás-os-Montes and Entre Douro e Minho (Leitão, 2011;Soares et al., 2023). In the municipality of Baião we can find the Castelo mountain range, part of the Aboboreira and Marão mountain ranges, and the Douro river at its southern limit ( Figure 1) (Soares et al., 2023;Lucas, 2012).  (Lucas, 2012).

Figure 1
Location of the study area (municipality of Baião), main hills, and LWF 2019 area. Tâmega valleys, which cool as they go up the slopes, leading to high precipitation in the highlands (Leitão, 2011).

Methodology
The study was organized in four stages. The comparisons with the rest of the national territory (Peixoto, 2019). For this work, we considered the following COS 2018 classes shown in Table 1.
The third phase was intended to combine the two previous parameters (Lampin-Maillet et al. 2009;. The calculation allowed the combination of different types of dwelling groups and different With the use of GIS software, the information related to the occurrence of fires had to be organized by individual "layers" and the year of its incidence. This information was converted into raster images,   (Equation 2), which aimed to assess the severity of the burnt area through the relationship between pre-fire and post-fire NBR (Santos et al., 2020;Roy et al., 2006).

Distribution of the number of wildfire occurrences and WUI
The municipality of Baião recorded 3,770 wildfire occurrences from 2001 to 2021 (Figure 2).
When analysing the annual distribution, it was observed that there was a downward trend, that is, a d e crea sin g lin ear tre n d in t h e n u m b e r of occurrences, with a R2 of 0.3068. Figure 3 illustrates the spatial distribution of wildfire occurrences, where is clear that ignitions were distributed throughout the entire territory.     Cartography of WUI areas combining configuration of buildings and vegetation structure in the municipality of Baião.

Figure 5
Land cover distribution according to WUI types in the municipality of Baião.

Geografia
Sarah Moura Batista dos Santos, António Bento-Gonçalves, António Vieira e Georgia Teixeira filling the "voids", there was a decrease in the number of cattle that used to make a valuable contribution to preventing rural fires by foraging shrubs when grazing (Leitão, 2011).

The large wildfire of Baião in 2019
According to data provided by the ICNF, up to   Based on the application of the dNBR spectral index to the Sentinel-2A images, we could distinguish the different degrees of severity experienced by the action of the fire (Figure 9a). For the area affected by the LWF, 40% corresponded to a high severity class, 25.25% to a moderately high severity class, 16.6% to a moderately low severity class, 12.95% to a low severity class, and 5.13% of the area was not burnt ( Table 3).
The spatial pattern of the fire severity in the burnt area is often determined by vegetation, topography, weather and duration of the fire (Guo et al., 2022;Fang et al., 2015;Fernández-García et al., 2022;Lentile et al., 2006). Weather is typically considered to play a key role in the distribution of wildfire severity, while terrain and vegetation are the main factors that affect the heterogeneity of spatial patterns of wildfire severity at local scale (Wu et al., 2013;Guo et al., 2022). Topography influences fire behaviour by changing microclimatic conditions and forest characteristics (Guo et al., 2022), leading to different spatial patterns of fire severity. Certain aspects of vegetation characteristics, such as type, structure, and load, have impor tant effects on w ildfire occurrence and behaviour (Guo et al., 2022;Birch et al., 2015) With regard to the slopes of the study area, it can be seen that 5% of the total area, where, during the period under study, the vegetation returned to photosynthetic activity, presented slopes between 0 and 5 degrees, 15% presented slopes between 5 and 10 degrees, 21% slopes between 10 and 15 degrees, 44% slopes between 15 and 25 degrees and 15% between 25 and 44.5 degrees.
In the evaluation of vegetation regeneration after 6 months of the wildfire (Figure 9b), we identified that 40.7% of the area showed high vegetation growth and 20.8% low vegetation growth, 22.4% had a response equivalent to a burnt area, and 17.8% of the area still has the characteristics of a burnt area (Table 4).
One year later (Figure 9c), we identified that 60.83% of the area showed high vegetation growth,

Figure 9
Severity and regrowth maps obtained using dNBR derived from Sentinel-2A images.
Assessment of vegetation regrowth and spatial patterns and severity factors of wildfires in wildland-urban interface -the case of the large wildfire in Baião (2019) nº 47 -2023 20.61% low vegetation growth, 14.97% had a response equivalent to an unburnt area and, 3, 59% of the area still has the characteristics of a burnt area (Table 4). However, within two years after the wildfire (Figure 9d), 76.27% of the area registered high vegetation growth, 12.15% low vegetation regrowth, 8.97% continued with the spectral behaviour of an u n b u r n t a r e a a n d 2 . 61% w i t h b u r n t a r e a characteristics.
In terms of the type of land use and occupation concerning 6 months after the wildfire, 6.3% (    This is valuable information for forest managers to understand the consequences (beneficial or otherwise) and plan prevention, restoration, and environmental education actions. The LWF area in the Baião municipality burned recurrently (maximum of 11 recurrences), and is an area heavily affected by fires (84% of the area has burned at least once).
Even with good vegetation recovery rates, it is urgent to develop studies on the multiple and varied impacts of fires. Two years after the fire, 100% of the area occupied by the forest and shrub classes, which were affected by high severity, already showed significant levels of vegetation regrowth.