Applying a Fire Exposure Metric in the Artificial Territories of Portugal: Mafra Municipality Case Study

Abstract

Portugal’s increasing wildfire frequency has led to home destruction, large areas burned, ecological damage, and economic loss, emphasizing the need for effective fire exposure assessments. This study builds on a Canadian approach to wildfire exposure and evaluates wildfire exposure in the Portuguese municipality of Mafra, using artificial territories (AT) as a proxy for the wildland–urban interface (WUI) and integrates land use land cover (LULC) data with a neighborhood analysis to map exposure at the municipal scale. Fire exposure was assessed for three fire transmission distances: radiant heat (RH, <30 m), short-range spotting (SRS, <100 m), and longer-range spotting (LRS, 100–500 m) using fine resolution (5 m) LULC data. Results revealed that while AT generally exhibited lower exposure (<16% “very high” exposure), adjacent hazardous LULC subtypes significantly increase wildfire hazard, with up to 51% of LULC subtypes classified as “very high exposure”. Field validation confirmed the accuracy of exposure maps, supporting their use in wildfire risk reduction strategies. This cost-effective, scalable approach offers actionable insights for forest and land managers, civil protection agencies, and policymakers, aiding in fuel management prioritization, community preparedness, and the design of evacuation planning. The methodology is adaptable to other fire-prone regions, particularly mediterranean landscapes.

1. Introduction

Several regions of the world are experiencing an increase in seasonally dry forests due to climate change, becoming more prone to wildfires. It has created an urgent need to rethink landscape management practices to reduce wildfire hazards and boost the resilience of forest ecosystems [1]. Wildfires have become increasingly dangerous and threaten life, safety, and property globally [2,3]. Over the last two decades, many wildfires have occurred in wildland–urban interface (WUI) areas, affecting people and their livelihoods [4]. Consisting of areas where structures and human development meet or intermingle with wildland vegetation, the WUI has gained increasing research attention due to heightened wildfire hazard. Geospatial methodologies, including zonal-based and point-based approaches, have been successfully validated for mapping WUIs at different scales, for example the municipal scale in Portugal [5], and the landscape scale in the USA [6] and Poland [7].

Growing expansion of WUI areas globally and adverse weather conditions have contributed to significant structure losses from high intensity wildfires that exceed fire suppression capabilities. Countries such as Portugal (2003, 2005, and 2017), the USA (2007, 2013, and 2017), Greece (2007 and 2018), Australia (2009), Chile (2016), and Canada (2023) have experienced severe impacts from such wildfires, with numerous civilian fatalities and widespread damage [2,8,9,10]. Over the past decade alone, more than 10 million people worldwide have been affected during and following wildfire events [11]. In Europe, particularly in the Mediterranean region, extreme wildfire events (EWE) have increased in frequency [12]. In recent decades, Portugal and Spain have been the most affected European countries, with 46 and 36 EWE, respectively, between 2000 and 2022 [13].

A study documenting wildfire fatality in Southern Europe (i.e., Portugal, Spain, Sardinia in Italy, and Greece) revealed that most occurred during the peak fire season under adverse weather conditions and linked to WUI areas [14]. Examples include the 2017 Pedrogão Grande and the October wildfires in Portugal and the 2018 Mati fire in Greece. The Mati wildfire, one of the deadliest in Europe, resulted in 103 fatalities, a loss of life exceeded only by the 2009 bushfires in Australia where 180 perished [15,16]. The Mati wildfire burned 1431 hectares, spreading quickly due to high temperatures exceeding 37 °C and strong winds gusting at speeds greater than 90 km/h [16]. The town of Mati was destroyed, with 1650 dwellings lost and over 140 injured, in addition to previously noted fatalities. Similarly, in Portugal, the Pedrógão Grande wildfire in June 2017 and the October wildfires that same year destroyed over five hundred structures [17]. During this unprecedented wildfire season, six percent of Portugal’s land area (557,400 ha) burned, causing economic losses exceeding 1.4 billion EUR in human assets and more than 120 fatalities [18]. Ignitions from spotting damaged 60% of the structures in the town, as embers were transmitted to receptive building features (e.g., roofs, windows, and other house openings) [8]. During these devastating wildfires, 92% of fatalities occurred within the WUI [19].

These catastrophic wildfires have prompted governmental and authority-driven efforts to improve fuel management focused on WUI settlements, isolated houses, and roads [19]. Additionally, citizen preparedness initiatives have been launched to help prevent large wildfires [20,21]. For example, the Aldeia segura-Pessoas seguras (Safe village–Safe people) program in Portugal supports community preparedness to face wildfires, protecting population clusters, forests, buildings, and goods in the WUI [20]. This program involves identifying hazard areas and shelter points and managing protection zones [22]. Likewise, scientific attention on wildfires and expanding WUI areas in Portugal has increased [19,23,24]. Land and fire managers can use fire hazard maps to create fire-adapted municipalities where threats to life and property are minimized [25]. In Europe, particularly in Portugal, many approaches have been developed to assess fire hazard in and around communities, such as for Funchal city and Pedrógão Grande village in Portugal, the Pan-Europe performance-based design approach addressed micro-level homeowner scale fire scenarios [3]. Other approaches include wildfire simulation modeling to assess wildfire exposure to communities in Portugal [18] and statistical modeling methods to map hazards on mainland Portugal [26].

Traditional fire hazard systems use weather indices derived from daily weather measurements that are analyzed with other variables in geographical information systems, statistical methods, and simulation modeling techniques [27]. Most of these hazard systems include socioeconomic and human factors with inputs often derived from outdated information, making it difficult to predict future conditions [28]. Dependency on unpredictable weather data and the limited applicability of local WUI-specific exposure pose further challenges to using these complex systems at the operational municipal level. Beverly et al. [29] created a unique community assessment geospatial approach for calculating a wildfire exposure metric in Canada based solely on the composition and arrangement of fuel, regardless of weather conditions or other abiotic influences. The approach targets site-level assessments in regions with high exposure to evaluate the potential for a given location within the built environment to receive ignitions from surrounding hazardous fuel. That study demonstrated that exposure assessments alone were sufficiently detailed to inform community fire management activities, such as improving WUI safety and fuel treatment planning. Other studies document similar geospatial methodologies for mapping wildfire exposure using LULC with the objective of informing effective fire prevention measures [28,30,31,32].

In Portugal, the official municipal methodology to assess wildfire exposure is detailed within the municipal forest fire defense plan (PMDFCI). The plan specifies multicriteria methods for calculating fire hazard from variables such as wildfire probability, risk susceptibility, vulnerability, and economic value [33]. The complexity of multicriteria methods can be prohibitive at the municipal scale in Portugal where resources are limited, prompting consideration of simpler, easier-to-apply methodologies. The success of the neighborhood analysis approach developed in Canada, its adaptability for municipal-scale assessments, and its provision of detailed LULC specific exposure assessment make it an innovative solution with potential to motivate action by empowering land managers with information to prioritize wildfire hazard areas in the WUI in Portugal.

This paper aims to map exposure to hazardous LULC in Portugal at the municipality scale using the neighborhood analysis methodology developed by Beverly et al. [29]. Our analysis targets specific fire transmission distances, integrates fine-resolution LULC data, and creates maps that can be replicated as tools to prioritize wildfire prevention measures. Specifically, we aim to do the following: (a) assess exposure levels in both artificial territories (AT) and land use land cover (LULC) in a Portuguese municipality for three ignition processes, namely radiant heat (RH), short-range spotting (SRS), and longer-range ember spotting (LRS); (b) design exposure maps for AT surrounding LULC for the municipality and its parishes; (c) assess the proportionate distribution of exposure values in the different parishes of the municipality and rank them by exposure. Identifying WUI vulnerability and wildfire exposure is crucial for planning and preparedness programs [34] and we expect the results of our study could offer a low-cost, time-saving alternative methodology for official government assessments relied upon by land and fire managers, civil protection agencies, municipalities, communities, and individuals who can be affected by wildfires. Finally, while this study focuses on Portugal, the methodology is adaptable to other regions with varying LULC types, especially those with Mediterranean climates or expanding WUI areas globally.

2. Materials and Methods

The municipality of Mafra in Portugal was selected for a case study application of local exposure maps to validate the neighborhood analysis methodology originally developed in Canada [29]. Exposure at a location is assessed by the composition of surrounding LULC types that contain vegetation capable of generating ignitions. Exposure is calculated for three ignition distances associated with distinct ignition processes: radiant heat (RH), short-range spotting (SRS), and longer-range spotting (LRS). While the spatial neighborhood calculation and exposure metric methodology were adapted from the original study, our research introduced several modifications to tailor the approach to the Portuguese context. LULC classifications were adjusted to reflect the specific LULC types found in Portugal. Local datasets were integrated to improve the accuracy of the exposure classification, consisting of fine-resolution LULC data provided by the Mafra Municipal Centre of Civil Protection. Finally, thresholds for exposure levels were customized to the Portuguese context.

2.1. Study Area

Mafra is a municipality in the Lisbon district of Portugal (Figure 1), which covers 29,158 ha, subdivided into 11 parishes [33], with a collective population of 86,515 based on the 2021 official census [35]. The municipality has a temperate climate with mild, dry summers. The average annual temperature is 17.4 °C, reaching an average maximum temperature of 24.9 °C in the critical season (June–September). An average annual rainfall of 774 mm occurs throughout the year, peaking in the fall and winter months. These conditions favor vegetation growth, which is fire-prone during dry summer conditions. Topographic conditions are rugged and mountainous. The municipality has experienced several destructive wildfires, particularly in 2003, during which fires burned 2756 ha, corresponding to almost 10% of Mafra territory; 2004 and 2005, with more than five hundred hectares burned each; and 2017, with four hundred hectares burned within the municipality [28]. More recently, the last large fire occurred in 2022, with 295 hectares burned [36]. The population density is high (298 persons per square km) [35], which is more than double the national average (114 persons per square km), and the municipality belongs to the Lisbon metropolitan area with a large commuter population to the Portuguese capital Lisbon, located 40 km away. Although Mafra’s burned area is not large compared with other Portuguese regions, the combination of high population density and high vegetation load makes this municipality a high wildfire hazard area suitable for validating the proposed methodology.

According to the municipality’s 2019 LULC map, Mafra is composed mainly of agricultural land (39%), followed by forest (22%), shrublands (25%), artificial territories (15%) and a small component of other LULC types (5%). Most relevant LULC subtypes within agricultural land include dryland areas with temporary crops (59%), orchards/vineyards (27%), permanently artificially irrigated crops (11%), and other agricultural land (3%). Most relevant LULC subtypes within forest are eucalyptus (40%), mixed forest species (37%), Quercus sp. (10%), maritime pine (7%), stone pine (2%), and other hardwoods and forest species (≤1%). We retrieved AT from the 2019 LULC Mafra municipal map and used it as a proxy for the WUI. According to COS 2018, AT is defined as a human intervention area. This class includes built areas, industries, commercial areas, tourism, infrastructure, road and rail networks, service facilities, gardens, and equipment storage areas [37]. Previous studies (i.e., [32]), successfully used WUI as a perimeter of the built-up area drawn by the directorate general of the territory (DGT).

2.2. Fire Exposure Metrics and Analysis

We used a fine-resolution (5 × 5 m) LULC raster map from 2019 provided by the Mafra municipal Centre of Civil Protection as the main base map for this study. Since the original Canadian methodology used different LULC classifications, we adapted the classification scheme to match the Portuguese LULC context. We reclassified the LULC into a binary hazardous/non-hazardous grid based on expert judgement regarding the fire hazard of each LULC subtype. Hazardous LULC subtypes were classified as “1” and non-hazardous as “0”. This classification relied on input from 10 forest and land managers at the Mafra Municipal Centre of Civil Protection, along with national fire and forest experts. We advised the experts that for the purpose of this study, hazardous LULC subtypes are defined as those capable of generating ignitions over a given fire transmission distance: 0.1 to 30 m (radiant heat, RH), 0.1 to 100 m (short-range spotting, SRS), and 100.1 to 500 m (longer-range spotting, LRS). Each expert was then asked to independently group the LULC subtype into two groups (hazardous and non-hazardous, and we adopted the statistical mode (i.e., the most frequently assigned classification) as the final value. Although no ties occurred in the assessments, our predefined criterion was to adopt the more conservative, hazardous classification in the event of a tie. As previously mentioned, the distance ranges for fire transmission were maintained from the original Canadian study because they align with findings from Portuguese wildfire spotting analyses [38], indicating they are appropriate for assessing wildfire exposure in Mediterranean landscapes. Studies have shown that while extreme spotting distances can occur, they are rare and typically linked to extreme fire conditions [29,39]. In Portugal, most ignitions occur within 2 km of the nearest road (98%) and 85% of them within 500 m [31]), while maximum spotting distances of 400–450 m have been predicted using models like Albini’s [40]. Computed exposure levels were classified into five equal interval bins adapted to the Portuguese context: 0–20% (very low), 20.1–40% (low), 40.1–60% (moderate), 60.1–80% (high), and 80.1–100% (very high). We opted for equal-interval classification due to its simplicity, transparency, and consistency with stretched symbology’s that display continuous data. Equal intervals mirror the pattern of raw values, facilitating meaningful analysis without altering the exposure landscape. This follows the rationale provided by Beverly et al. [41] for preserving information in raw data when mapping exposure for real-world settings. Simplifying exposure levels into classes facilitates a detailed spatial assessment around AT, particularly near houses, which are critical assets in fire-prone areas. As demonstrated by [8], houses with poor construction materials and neglected maintenance are more vulnerable to ignition, especially when embers land on receptive building features or when nearby flammable materials ignite. Exposure maps identify areas where vegetation conditions and housing characteristics may jointly facilitate fire spread.

Based on expert classification of LULC subtypes throughout the study area, hazardous fuels were identified as (

Table 1. Expert classification of LULC subtypes into hazardous (1) and non-hazardous (0) categories. The expert group consisted of 10 forest and land managers from the municipality of Mafra.

LULC Subtypes	Radiant Heat (RH) 0.1–30 m	Short-Range Spotting (SRS) 0.1–100 m	Longer-Range Spotting (LRS) 100.1–500 m
Artificial territories	0	0	0
Dryland areas with temporary crops	1	1	0
Permanently artificially irrigated crops	0	0	0
Herbaceous vegetation	0	0	0
Quercus sp.	1	1	1
Others—agricultural land	0	0	0
Maritime pine	1	1	1
Eucalyptus	1	1	1
Other forest species	1	1	1
Other hardwoods	1	1	0
Orchard/vineyard	0	0	0
Mixed forest species	1	1	1
Stone pine	1	1	1
Shrubs	1	1	0
Coastal dunes and beaches	0	0	0
Riparian vegetation	1	0	0
Sparse vegetation	0	0	0
Clear cuttings	1	1	0
Barren lands	0	0	0
Water bodies	0	0	0

): oak forests, maritime pine, eucalyptus, stone pine, shrubs, mixed forest species, hardwoods, riparian vegetation, clear cuttings, and dryland areas with temporary crops. All other areas were classified as non-hazardous, which included artificial territories, such as continuous and discontinuous urban areas, construction sites, industrial areas, commercial areas, roads, dams, electrical lines, and ports. Some LULC subtypes with vegetation cover, such as locations with irrigated crops, herbaceous, or orchards/vineyards, were also classified as non-hazard.

We measured exposure to hazardous subtypes in LULC with a special focus on AT using neighborhood analysis in the ArcGIS Pro Spatial Analyst toolset [42]. The methodology is reproducible using open-source tools such as the R package “fireexposuR” in R CRAN R-4.3.4 [43] developed by [44] and GIS tools such as QGIS, making it accessible at no cost to a broad range of users and adaptable to different regions without requiring proprietary software. For each focal cell in the 2019 LULC raster map, the neighborhood analysis summed the binary classification of hazardous LULC subtypes for all cells within the specified radius around each assessment cell. The neighborhood analysis evaluates surrounding cells within a defined radius to identify spatial patterns. We estimated exposure at each cell using the proportion of neighborhood cells that contained hazardous LULC subtypes, following the methodology of Beverly et al. [29,41].

Individual exposure assessments were generated for each fire transmission distance range: RH (0.1–30 m), SRS (0.1–100 m), and LRS (100.1–500 m). We used an annulus radius of 20 and 100 cells for LRS and a circular radius of 6 and 20 cells for RH and SRS distances, respectively (Figure 2). Fine-resolution data was essential for accurately calculating exposure to radiant heat.

We applied a negative buffer around the analysis area to minimize edge effects. For the outer buffer zone beyond the 2019 Mafra LULC map, we used the national 2018 LULC raster map (COS 2018, resolution 100 m), freely available through the DGT-SNIG website [45]. Shapefiles were converted into raster format to ensure compatibility. After calculating exposure, we applied buffers corresponding to each ignition distance (30, 100, and 500 m) and then clipped the results to match the study area boundary, ensuring completeness and continuity of exposure estimation throughout the full extent of analysis.

Using the lower-resolution COS 2018 map for the external buffer ensured full coverage while acknowledging a trade-off in spatial precision at the borders. While this approach allowed for complete spatial coverage, the exposure assessment at the edges of the 2019 LULC map is less precise due to shifts in LULC resolution from fine (5 m) to coarse (100 m).

We classified each parish’s exposure within the AT and surrounding LULC subtypes into five “low to very high” levels. The composition of LULC subtypes surrounding these AT areas determined the amount of vegetation capable of producing ignitions. We conducted several field visits from January to March 2023 to validate the resulting exposure maps. Five villages of the eleven Mafra parishes were selected based on two criteria: areas with very-high wildfire hazards assessed by the PMDFCI and areas selected to implement the Aldeia segura-Pessoas seguras program. Visited villages were: (1) Bocal de Baixo (parish “Venda do Pinheiro e Santo Estêvão das Galés”), (2) Picão (parish “Enxara do Bispo Gradil e Vila Franca do Rosário”), (3) Lagoa (parish “Santo Isidoro”), (4) Boco, Valverde, Pipo (parish “Igreja Nova e Cheleiros”), and (5) Urzal (parish “Carvoeira”). During field visits, we assessed LULC subtypes and structural vulnerability to validate the exposure classifications generated in our maps. Specifically, we examined whether high exposure areas identified in the spatial analysis corresponded to hazardous LULC subtypes on the ground. Observations included the proximity of structures to hazardous LULC subtypes or the location of structures in low and very high exposure areas. Field visits were carried out with the local forest and land manager, who validated and discussed the exposure classifications obtained.

Finally, we overlaid the official hazard map provided by the Mafra Municipal Centre of Civil Protection, with our RH exposure classes. To validate the correspondence between the official hazard classes and our RH exposure classes, we used 3000 random points and performed a spatial join in ArcGIS Pro. Since the official hazard map does not account for ember distances, RH ignition was considered the most appropriate ignition type for validation.

3. Results

The municipality of Mafra has exposure to wildfire from all three fire transmission distances (i.e., ignition processes) included in our analysis: RH (0.1–30 m), SRS (0.1–100 m), and LRS (100.1–500 m). Exposure levels ranged from “very low” to “very high” across all three ignition processes. Nearly all of Mafra’s territory exhibits some level of exposure (

Table 2. The proportion of artificial territories (AT) and land use land cover (LULC) area in the municipality of Mafra by exposure class.

Exposure Level	Radiant Heat (RH) 0.1–30 m		Short-Range Spotting (SRS) 0.1–100 m		Longer-Range Spotting (LRS) 100.1–500 m
	Proportion LULC	Proportion AT	Proportion LULC	Proportion AT	Proportion LULC	Proportion AT
0–0.2 (very low)	0.42	0.79	0.31	0.59	0.58	0.71
0.2–0.4 (low)	0.07	0.12	0.15	0.20	0.23	0.21
0.4–0.6 (moderate)	0.06	0.07	0.15	0.13	0.13	0.06
0.6–0.8 (high)	0.07	0.02	0.14	0.06	0.05	0.01
0.8–1 (very high)	0.37	0.00	0.25	0.02	0.00	0.00

). Substantial areas of LULC (37%) have very high exposure to RH, whereas AT exhibited none. Most AT land (79%) has very low exposure to RH, compared with 42% of LULC, i.e., a quarter of LULC lands have “very high” exposure to SRS, while only 2% of AT falls into that class. The majority of AT (59%) has “very low” exposure to SRS. Most areas in LULC (58%) and AT (71%) have “very low” exposure to LRS, with no areas identified as having “very high” exposure. This suggests that LRS poses a lesser threat to both LULC and AT within the municipality.

The exposure maps for Mafra reveal distinct patterns that differ by the fire transmission distance assessed. AT lands were assessed as having predominantly lower exposure to RH (Figure 3A). Exposure to SRS in AT zones exhibited more variation, with exposure levels ranging from “very low” to “high” (Figure 3B). In contrast, the LRS map indicates minimal risk to AT, with most areas assessed as having “very low” or “low” exposure (Figure 3C). These spatial patterns support the identification of neighborhoods and infrastructures most exposed to each ignition process, informing more targeted mitigation efforts.

Summaries of exposure levels by the parish (

Table 3. Proportion distribution of exposure class by parish, LULC or AT, and fire transmission distance (SRS, 0.1–100 m; LRS, 100–500 m).

	Exposure Class
Parishes			Very Low >0–≤0.2	Low >0.2–≤0.4	Moderate >0.4–≤0.6	High >0.6–≤0.8	Very High >0.8–≤1
A-Encarnação	RH	LC	0.48	0.07	0.06	0.07	0.31
		AT	0.72	0.16	0.09	0.03	0.00
	SRS	LC	0.36	0.15	0.15	0.15	0.18
		AT	0.47	0.25	0.17	0.09	0.02
	LRS	LC	0.97	0.03	0.01	0.00	0.00
		AT	1.00	0.00	0.00	0.00	0.00
B-Santo Isidoro	RH	LC	0.35	0.08	0.07	0.08	0.41
		AT	0.77	0.14	0.07	0.01	0.00
	SRS	LC	0.23	0.16	0.17	0.16	0.27
		AT	0.51	0.25	0.16	0.07	0.01
	LRS	LC	0.39	0.37	0.18	0.06	0.01
		AT	0.03	0.91	0.05	0.00	0.00
C-Ericeira	RH	LC	0.68	0.05	0.04	0.04	0.19
		AT	0.95	0.03	0.01	0.00	0.00
	SRS	LC	0.59	0.13	0.10	0.07	0.11
		AT	0.87	0.08	0.03	0.01	0.00
	LRS	LC	0.59	0.29	0.11	0.01	0.00
		AT	0.82	0.17	0.01	0.00	0.00
D-Carvoeira	RH	LC	0.62	0.07	0.06	0.06	0.19
		AT	0.87	0.08	0.04	0.01	0.00
	SRS	LC	0.52	0.15	0.13	0.10	0.09
		AT	0.75	0.13	0.07	0.03	0.01
	LRS	LC	0.70	0.29	0.01	0.00	0.00
		AT	0.92	0.07	0.01	0.00	0.00
E-Azueira e Sobral da Abelheira	RH	LC	0.57	0.06	0.05	0.05	0.27
		AT	0.89	0.07	0.03	0.00	0.00
	SRS	LC	0.48	0.14	0.11	0.09	0.18
		AT	0.77	0.14	0.06	0.03	0.00
	LRS	LC	0.50	0.23	0.18	0.08	0.01
		AT	0.77	0.17	0.05	0.01	0.00
F-Mafra	RH	LC	0.44	0.07	0.06	0.07	0.35
		AT	0.85	0.10	0.04	0.01	0.00
	SRS	LC	0.31	0.17	0.16	0.14	0.22
		AT	0.67	0.19	0.10	0.03	0.01
	LRS	LC	0.39	0.30	0.22	0.09	0.00
		AT	0.77	0.16	0.05	0.01	0.00
G-Igreja Nova e Cheleiros	RH	LC	0.38	0.09	0.08	0.08	0.37
		AT	0.73	0.17	0.09	0.01	0.00
	SRS	LC	0.24	0.18	0.19	0.17	0.22
		AT	0.44	0.27	0.20	0.08	0.01
	LRS	LC	0.70	0.17	0.10	0.03	0.00
		AT	0.74	0.19	0.06	0.00	0.00
H-Enxara do Bispo, Gradil e Vila Franca do Rosário	RH	LC	0.35	0.07	0.06	0.07	0.45
		AT	0.73	0.15	0.09	0.03	0.00
	SRS	LC	0.23	0.15	0.15	0.15	0.33
		AT	0.48	0.23	0.17	0.10	0.03
	LRS	LC	0.72	0.10	0.09	0.07	0.02
		AT	0.68	0.21	0.09	0.01	0.01
I-Milharado	RH	LC	0.31	0.07	0.07	0.08	0.46
		AT	0.68	0.18	0.11	0.03	0.00
	SRS	LC	0.18	0.14	0.17	0.19	0.33
		AT	0.35	0.27	0.23	0.12	0.03
	LRS	LC	0.79	0.15	0.05	0.01	0.00
		AT	0.77	0.22	0.01	0.00	0.00
J-Malveira e São Miguel de Alcainça	RH	LC	0.30	0.05	0.05	0.06	0.53
		AT	0.83	0.11	0.06	0.01	0.00
	SRS	LC	0.21	0.09	0.12	0.14	0.44
		AT	0.62	0.18	0.13	0.06	0.01
	LRS	LC	0.36	0.36	0.21	0.07	0.00
		AT	0.72	0.24	0.03	0.01	0.00
K-Venda do Pinheiro e Santo Estêvão das Galés	RH	LC	0.39	0.06	0.06	0.07	0.42
		AT	0.80	0.12	0.06	0.02	0.00
	SRS	LC	0.28	0.13	0.13	0.15	0.31
		AT	0.59	0.19	0.13	0.07	0.02
	LRS	LC	0.41	0.37	0.15	0.07	0.00
		AT	0.44	0.43	0.10	0.02	0.00

) reveal significant variation across the municipality. For RH, the parish of “Malveira e São Miguel de Alcainça” shows the highest LULC exposure in the “very high” class (53%), while “Ericeira” shows the lowest AT exposure, with 95% of its AT classified as “very low”. Notably, no parish had AT with “very high” RH exposure, indicating low vulnerability of the built environment to radiant heat ignition.

Exposure to SRS was notable for “Malveira e São Miguel de Alcainça”, where 44% of LULC was assessed as having “very high” exposure. However, AT in this parish exhibited predominantly lower exposure, suggesting that although the surrounding landscape is hazardous, the built environment is not directly exposed. The Parishes of “Ericeira” and “Azueira e Sobral da Abelheira” exhibited the lowest AT exposure to SRS, with 87% and 77% of AT, respectively, assessed as having “very low” exposure.

Exposure to LRS across all parishes is consistently low. “Encarnação” parish has the lowest exposure with 97% of LULC and 100% of AT (100%) assessed as having “very low” exposure. In all parishes, “very high” exposure to LRS was limited to ≤1% of AT, confirming that longer-range ember transport poses minimal risk to artificial territories in Mafra. This pattern indicates that RH and SRS remain the dominant fire transmission processes for prioritizing prevention actions in the municipality.

Interestingly, exposure profiles vary not only between parishes but also between LULC and AT within the same parish. For example, while “Milharado” and “Malveira e São Miguel de Alcainça” have high LULC exposure to RH and SRS, AT in these parishes have predominantly lower exposure to RH and SRS. In contrast, parishes like “Ericeira”, “Carvoeira”, and “Azueira e Sobral da Abelheira” consistently exhibit low exposure in both LULC and AT, suggesting lower overall wildfire susceptibility.

Ranking of parishes by the proportion of area classified as “very high” wildfire exposure (>0.8 to ≤1) is shown in

Table 4. Parish rank by the proportions of LULC and AT associated with “very high” exposure (>0.8 to ≤1).

	Radiant Heat (RH) 0.1–30 m		Short-Range Spotting (SRS) 0.1–100 m
Parishes of Municipality of Mafra	Ranks	Exposure Values	Ranks	Exposure Values
Malveira e São Miguel de Alcainça	1	0.53	1	0.44
Milharado	2	0.46	2	0.33
Enxara do Bispo, Gradil e Vila Franca do Rosário	3	0.45	2	0.33
Venda do Pinheiro e Santo Estêvão das Galés	3	0.42	3	0.31
Santo Isidoro	4	0.41	4	0.27
Igreja Nova e Cheleiros	5	0.37	5	0.22
Mafra	6	0.35	5	0.22
Encarnação	7	0.31	6	0.18
Azueira e Sobral da Abelheira	8	0.27	6	0.18
Carvoeira	9	0.19	7	0.09
Ericeira	9	0.19	8	0.11

. “Malveira e São Miguel de Alcainça” was ranked on top for both RH (0.53) and SRS (0.44), followed by “Milharado” and “Enxara do Bispo, Gradil e Vila Franca do Rosário”, which also have elevated exposure levels (≥0.33). It is evident that certain parishes, particularly in inland zones, are more prone to ignition due to land use and vegetation patterns that support short-range fire transmission. In contrast, coastal parishes like “Ericeira” and “Carvoeira” consistently exhibit the lowest exposure levels, likely benefiting from less fire-prone LULC subtypes.

Across all parishes, AT exposure in the highest class remains minimal or near zero for all fire transmission types, reinforcing that built areas themselves are typically less hazardous but may still be at risk when adjacent to high-exposure LULC. This spatial distinction underlines the importance of targeting surrounding vegetation in fire prevention efforts. Figure 4 illustrates the composition of LULC surrounding AT in the parish of Milharado, one of the highest-ranked areas, where dominant LULC subtypes such as eucalyptus and shrublands contribute significantly to “very high” exposure to SRS. The magnified inset highlights LULC types clustered around built structures, signaling priority areas for preparedness planning and fuel reduction treatments.

Field validation carried out after the exposure assessment supported its accuracy. The field validation was qualitative, based on observations. Field visits confirmed that several parishes with “high” and “very high” exposure levels identified in the spatial analysis corresponded to hazardous LULC subtypes and fuel continuity on the ground. In all of the field-visited parishes, notably “Santo Isidoro” and “Venda do Pinheiro e Santo Estêvão das Galés”, the assessed exposure levels aligned with observed conditions as they exhibited the highest exposure, with minor discrepancies due to new developments and isolated structures that had recently emerged and were not updated in the AT database. Local civil protection authorities attended the field trips and confirmed the reliability of the exposure classification, reinforcing the usefulness of the methodology for identifying high-risk areas requiring prevention measures.

The comparison between the official Mafra hazard map (Figure 5A) and the 2019 RH (0.1–30 m) exposure map (Figure 5B) highlights key differences in the assessment approaches. The official hazard map, which incorporates factors such as slope but does not account for ember travel or wind-driven fire spread, presents a more spatially limited pattern, with “high” and “very high” hazard zones presenting in narrow linear patterns, concentrated primarily along ridges and steep terrain. In contrast, the RH exposure map (Figure 5B), which specifically evaluates radiant heat within a 30 m radius, reveals contiguous areas of varying sizes with very high exposure, distributed in a spatially varying patchwork throughout the parish. The exposure assessment identifies a much larger land area with a “high” or “very high” (red) classification, particularly near built environments. This suggests that the two approaches provide fundamentally different information, and localized fire hazard may be underestimated by the official hazard map, particularly in WUI areas where radiant heat and ember spotting are critical ignition pathways.

4. Discussion

Mapping wildfire exposure at the municipal scale provides critical insights for improving forest and land use planning, community preparedness, and wildfire mitigation strategies [5]. By integrating LULC data with a neighborhood analysis approach based on Beverly et al. [29], this study provided a detailed exposure assessment for the municipality of Mafra, Portugal. Using AT as a proxy for the WUI, the analysis identifies areas where built environments are vulnerable to wildfire ignition, which offers valuable guidance for decision makers seeking to implement more targeted prevention and preparedness strategies. Our results reveal significant variability in wildfire exposure across Mafra, particularly between AT and surrounding LULC. High exposure areas are predominantly located within LULC areas that are rich in continuous vegetation. Up to 41% of LULC land area has very high exposure above >80%. In contrast, AT areas, which are characterized by less flammable materials and landscape fragmentation, generally exhibit very low exposure levels with less than 8% of AT land area assessed as having high exposure >60%. However, AT areas adjacent to hazardous LULC subtypes remain at risk due to ember transport and SRS [29], which can be expected in fire-prone Mediterranean landscapes. Small areas can contain high densities of structures with correspondingly high potential for fire damage. Areas classified as “high” and “very high” exposure should be targeted for fuel management and defensible space initiatives. Home ignitions, often caused by high densities of embers falling on receptive building materials like roofs or gutters [5,46] can be mitigated through strategic landscape prevention measures in these critical areas.

While AT was used in this study as a proxy for the WUI, the spatial patterns observed highlight typical WUI vulnerabilities, where built environments interface with wildland areas. For example, parishes like “Malveira e São Miguel de Alcainça”, “Milharado” and “Enxara do Bispo” emerged as high priorities for hazard reduction due to their elevated exposure to RH and SRS [29]. In contrast, parishes like “Encarnação” and “Ericeira” with extensive “very low” exposure areas, present opportunities for future urban growth or environmental protection.

Our findings align with previous studies emphasizing the critical role of LULC in wildfire spread [47,48,49,50]. Several studies in Portugal and other mediterranean regions have identified eucalyptus, pine forest including Quercus sp., and maritime pine as flammable LULC subtypes [49,50,51,52,53,54,55,56]. However, most of these studies focused on natural landscapes, omitting built environments from exposure assessments. By using AT, this study addressed that gap, enabling a more comprehensive analysis of wildfire exposure. This is consistent with recent studies promoting integrated hazard models that consider both natural LULC and built environment [5,32,57]. Furthermore, using a neighborhood analysis to quantify exposure at different fire transmission distances (RH, SRS, and LRS), adds depth to traditional hazard assessments, which often overlook the spatial dynamics of ember distances and heat transfer [29].

Although our results were compared with Mafra’s official hazard map, the two approaches are based on fundamentally different criteria. The official map incorporates slope and LULC but does not consider wind or ember transport, key elements in wildfire spread. In contrast, our RH based exposure assessment focuses on proximity to hazardous fuels, offering a more localized and structure-relevant perspective. Inconsistencies between the two maps are not surprising, given the differences in assessment criteria, and highlight the added value of neighborhood-based exposure analysis in capturing risks to built environments.

This study contributes to the evolving field of wildfire hazard assessment by applying an exposure metric from Beverly et al. [29], which is promoted widely by Canada’s FireSmart Program. The FireSmart-endorsed tool has been used to guide wildfire planning and protect values at risk, at spatial scales ranging from individual homeowners to municipal and national extents [29]. Exposure assessments can also be applied and customized for different contexts, for example as an input for directional vulnerability mapping [58], for regional validation efforts [59], and for optimizing fuel treatments [60]. Similarly, Portugal has embraced simplified exposure metrics for community wildfire hazard assessments, leading to successful validations in projects focused on municipal-level planning [61].

The generated exposure maps are a valuable tool for municipal and civil protection agencies, and can be used to prioritize areas for fuel management as a local prevention measure, as well as to inform community preparedness and planning. These maps can also inform evacuation planning by identifying critical road networks and potential bottlenecks [61], which is essential for safeguarding people during wildfires. Areas with high exposure should be prioritized for fuel treatments and public awareness campaigns, while low-exposure areas could be considered for safer urban expansion. Resilience-building strategies should focus on AT near hazardous LULC, as these areas are most susceptible to ember transport and RH [8,57].

We believe incorporating this methodology into existing community-based wildfire prevention programs, such as Portugal’s Aldeiras seguras (Safe villages), could enhance preparedness efforts, particularly in parishes identified as high risk from ember transport and RH. Complying with national regulations that require vegetation clearance around buildings, such as the 50 m buffer law [62], remains a critical preventive measure that directly alters wildfire vulnerability [57].

The simplified methodology used in this study enhances replicability and scalability across different regions, but it also has several important limitations. The exposure assessment relies on static land use and limited land cover (LULC) data [63], which may not accurately reflect temporal changes in vegetation or land management practices. Furthermore, the approach does not incorporate dynamic fire behavior drivers such as wind direction and speed, topographic variation, or real-time weather conditions, all of which influence fire spread and intensity. Exposure maps can, however, be interpreted day-to-day in relation to weather and ignitions, which are highly variable and cannot be known with any degree of certainty over the strategic planning horizons that exposure assessments inform. Omission of fire behavior modeling within the exposure assessment framework could result in under- or overestimation of actual exposure, particularly in complex terrain or under extreme fire weather scenarios. These limitations highlight the need for future integration of dynamic fire modeling to complement LULC-based assessments. Another minor limitation of this study is the mixed resolutions of the LULC datasets used (5 m and 100 m) and the likelihood that some edge effects were introduced in border zones. Field validation confirmed the accuracy of our results, which suggests edge effects did not undermine our analysis; however, future studies could explore ways to reduce edge effects, particularly in border areas near high-risk AT. We also acknowledge that the use of AT as a WUI proxy may not capture finer heterogeneity in peri-urban vegetation structure or informal settlements.

The exposure assessment method used in this study aligns with broader wildfire risk reduction frameworks, such as the EU forest strategy and the Sendai framework for disaster risk reduction, which emphasize proactive, community focused wildfire management [64,65]. Beyond its application in Mafra, this methodology can be adapted to other geographic regions with varying fire regimes and LULC types. Given its reliance on LULC classifications and spatial neighborhood analysis, this approach is particularly well suited for wildfire prone areas experiencing WUI expansion, including other Mediterranean regions, such as California or Australia. By refining local land cover classifications and integrating updated land management data, exposure assessments could be completed in many different regions at different scales, from municipal to national fire management planning. Future follow-up studies would help test the applicability of this methodology in diverse fire regimes. Future exposure assessments could also be used to evaluate prevention and mitigation efforts that aim to reduce wildfire hazard in high exposure parishes. Longitudinal studies could be used to track changes in fire exposure over time. Finally, incorporating social vulnerability assessments and road network exposure [66] would likely support more equitable and effective fire management practices, ensuring that high-risk but resource-limited communities receive adequate attention.

5. Conclusions

This study presented a refined wildfire exposure assessment method by integrating LULC classifications with a neighborhood analysis approach. The resulting assessments provide a detailed view of fire exposure dynamics by targeting three specific fire transmission distances (RH, SRS, and LRS). The study objectives were to (a) assess exposure from RH, SRS, and LRS in artificial territories and LULC; (b) create exposure maps at municipal and parish levels; (c) rank parishes by exposure. Most AT had low or no exposure; however, AT adjacent to hazardous LULC subtypes remained vulnerable, particularly due to ember transport processes like SRS. Areas with the highest hazard were identified, using AT as a proxy for the WUI. Field validation confirmed the reliability of the methodology for application in Portugal.

The generated exposure maps are a cost-effective, time-saving tool for forest and land managers, civil protection agencies, and policymakers. Exposure assessments can inform targeted fuel management, community preparedness, and urban planning while offering the flexibility to integrate into existing wildfire prevention frameworks like Aldeia segura-Pessoas seguras. Replicability and adaptability of exposure maps make them valuable for use in other Mediterranean regions or fire-prone landscapes.

Our methodology provides a scalable framework that could be integrated into official wildfire hazard assessments and planning strategies, reinforcing its relevance for both municipal-level mitigation and broader national policies. By streamlining exposure mapping and focusing on practical applications, our study supports more efficient wildfire prevention efforts, while laying the groundwork for future research into dynamic fire behavior and community vulnerability.

Also, the method for assessing fire exposure in forest landscapes, particularly in heterogeneous environments like Portugal, should be expanded beyond vegetation cover types. The original method, developed for Canadian landscapes, may have limitations in structurally diverse ecosystems. Future studies could explore the use of fuel load, weather data, and fire behavior modeling to enhance the accuracy and applicability of hazard estimates.