The Potential of Landscape Metrics for Estimating Forest Fire Risk in Poland

Forest fires (and more broadly – wildfires) are extensively studied globally due to their far-reaching consequences on nature (e.g., Ward et al. 2020), society (e.g., Kountouris, Remoundou 2011, Paudel 2023), the economy (e.g., Sadowska et al. 2021, Poduška, Stajić 2024), and human life (e.g., Kalogiannidis et al. 2023). Countryman (1972) stated that fire behaviour is determined by the environment in which it occurs, highlighting topography, fuel, and meteorological conditions in this context. Subsequent studies conducted, among others, in temperate-zone forests have confirmed the relationship between fires and both topographic conditions (e.g., Adámek et al. 2015, Ciesielski et al. 2022) and meteorological factors (e.g., Krikken et al. 2019, Sutanto et al. 2020, Venäläinen et al. 2020, Mandal et al. 2021). Regarding fuel, Countryman (1972) emphasised the significant influence of human activity, such as roads and settlements near forests – a finding later corroborated by the literature review of Costafreda-Aumedes et al. (2017) and, in the context of Polish forests, by Kolanek et al. (2021) and Milanović et al. (2023). Countryman (1972) also mentioned forest management as a factor potentially affecting forest fires. Lee et al. (2009) demonstrate that forest stand type, tree composition and the spatial arrangement of forest cover patches influence fire spread intensity in Samchuck, South Korea. Research on the forest structure in relation to fire behaviour has yielded significant results, regardless of the region in which it was conducted. For example, some research found that fires are more likely in homogeneous woody areas (Lloret et al. 2002) and tend to be lower in forests with high patch density and diverse tree cover (Lee et al. 2009, Román-Cuesta et al. 2009). Greater variability in forest structure may enhance resistance to external factors, including fire (Koontz et al. 2020). Fire resistance may stem from the interaction between fire behaviour and forest structure, although Koontz et al. (2020) note that the scale of this effect is local, evident in the smallest tested area (90 m × 90 m). In studies broadly examining the influence of the spatial structure of landscape on wildfires, other factors have also been considered in addition to stand type, such as tree age (Zhu et al. 2006) or humidity (Davies et al. 2016).

The landscape context in forest studies has been analysed in terms of the impact of wildfires on landscape structure (Hudak et al. 2004, James et al. 2007, Hanberry 2020, Bountzouklis et al. 2022, Donager et al. 2022) and vice versa. Gonzalez et al. (2005) noted that it is reasonable to assume a relationship between the spatial distribution of different forest stands and fire risk; therefore, the spatial distribution of more or less susceptible to fires forest stands should be considered. Beverly et al. (2021) mentioned that the proximity of land cover elements to one another can either facilitate or limit fire spread. Lloret et al. (2002) and Gonçalves et al. (2012) demonstrated that the landscape-level structure of vegetation and its changes affect the spatial patterns of wildfire frequency and size. However, in these studies, forest habitats were grouped only into general categories (e.g., dense forest/open forest), without analysing the more detailed structure of stands (e.g., in terms of humidity, age, or dominant tree species).

Landscape metrics, which are algorithms that quantify specific spatial characteristics of patches, classes of patches, or entire landscape mosaics (McGarigal, Marks 1995), have been used to study forest environments, typically analysing changes in the cover or structure of individual habitat types (Southworth et al. 2002, Rutledge 2003, Żmihorski et al. 2010, Kayiranga et al. 2016, Ersoy Mirici et al. 2020, Netzel et al. 2024). Geri et al. (2010) quantified changes in forest patterns based on three forest classes (broad-leaved, conifer, and mixed) in a Mediterranean area by comparing historical and recent forest maps. del Castillo et al. (2015) used landscape metrics to relate forest management to nature conservation, with analyses based on nine classes (forest stand type and species composition). Hardy et al. (2023) applied landscape analyses to investigate the share of old forests and forest fragmentation regarding forest management. Landscape metrics have also been utilised to study forest fires. Costafreda-Aumedes et al. (2013) sought to predict human-caused fire occurrence in Spain for >5 years using landscape ecology metrics across 10 km × 10 km grids based on the Forest Map of Spain. Additionally, Viedma (2008) assessed the relative influence of topography and fire on landscape patterns within a large forest in Spain. Alexandre et al. (2016) categorised forest areas into three flammability categories, subsequently employing landscape metrics. Dutt and Kunz (2024) used landscape metrics to monitor changes in forest cover post-disasters, aiding in recovery assessments and management decisions.

In Polish forests, a mosaic of trophic types and humidity types of forest habitats exists, which may directly influence their susceptibility to fire. Coniferous species dominate 68.7% of Poland’s forest area (partly due to the preference of these trees by the timber industry) and Scots pine (Pinus sylvestris L.) occupies 58.1% of the forest area (Report on the State of Forests in Poland 2018). However, does the forest structure itself affect the incidence of fires? Aszalós et al. (2022), through analysing the diversity of stands in selected European countries, discovered that the forest structure impacts sensitivity to disturbances like the bark beetle. Even straightforward diagrams from Poland in 2007–2017 illustrate differences between the features of stands in all forest areas and those where fires occurred (Fig. 1). Based on this, herein we verify a research hypothesis that the spatial structure of forests (understood as the spatial differentiation of forest stands’ characteristics within forests) significantly influences the occurrence of fires. In other words, we aim to check whether increased spatial fragmentation of forest stands causes the occurrence of fires to rise.

Fig. 1.

Differences between the structure of forest stands in general forest areas (outer ring) and in stands where fires broke out (inner ring); data from the entire territory of Poland, 2007–2017. Units are given in percentages.

Landscape metrics may play an important role in forest management by providing quantitative measures that describe the structure of forest ecosystems. Gonçalves et al. (2012) tested several landscape metrics (e.g., mean, dominant, relative patch richness, area weighted mean shape index, mean perimeter-area ratio, mean patch fractal dimension, edge density, median patch size, patch density) as potentially useful to conduct a fire risk analysis. Their results pointed to the usefulness of these metrics to assess the role of landscape characteristics on fire size distribution, which was the reason for us to take up the topic. Our research aimed to verify the hypothesis that the spatial differentiation of tree stand structures (precisely, State Forests division subareas) is related to the occurrence of forest fires. The hypothesis is formulated on the basis of two elements:

• differences between the structure of forest characteristics in all forest subdivisions in Poland and in subdivisions where fires were noted (Fig. 1);

• the fact that the spatial structure of forest habitats in Poland largely reflects human forest management, making it possible to conduct modelling of forest structure in ways that enhance fire prevention.

Forest structure is influenced by the origin, history and manner of forest management. Żmihorski et al. (2010) suggest that forest management and economic interests may be related to the share of forest characteristics in central Poland. For over 100 years (since 1924), forests in Poland have been managed by State Forests. Since 1991, the Forest Act (1991) has been regulating the management of stands in individual forest districts, which are the basic management units of the State Forests administration. In accordance with the provisions of this Act, after every 10 years, a document (called forest management plan [FMP]) is created for each forest district that specifies the scope, objectives, and methods for conducting economic activities in individual forest subareas, which are the smallest units of forest, distinguished by trophic, humidity, and age types of trees. Therefore, the results of the above studies may provide valuable information that will help in planning future activities to significantly reduce the number of fires breaking out in forest districts. Especially, given that landscape metrics are already being utilised in the management of areas affected by various hazards (e.g., Buchholtz et al. 2023).

Our findings confirm the research hypothesis and suggest that some of the characteristics of forest patches in the vicinity of fire outbreaks differ from those in randomly selected areas. At the landscape level, regardless of the habitat feature under consideration – whether age class, trophic type, or humidity type – forest patches surrounding fire ignition points were generally smaller and exhibited longer edge lengths. This suggests that areas where forest fires occurred exhibit higher landscape fragmentation compared with randomly chosen locations. This may imply that such areas are more susceptible to external disturbances because greater forest fragmentation is associated with a higher density of roads running through forest areas, which may affect the density of forest fires (Kolanek et al. 2021). These results are similar to those of Bonner et al. (2024), who suggest that the complexity of forest structures enhances the overall damage caused by fires. Also, Noble et al. (2025) showed that fire density was the greatest in small, complex forest patches.

Moreover, patches in fire-affected zones displayed greater shape complexity, indicating more irregular and non-standard contours. It is a result similar to the findings of Hébert-Dufresne et al. (2018), who demonstrated that patches with more irregular boundaries are more susceptible to the effects of edge fire.

For trophic and humidity-related characteristics, patch size variability and the NumP were higher in the vicinity of fire sites than in one of the randomly selected groups. However, some of these differences were statistically insignificant, potentially indicating a lesser relevance of these metrics. Overall, at the landscape level, areas surrounding forest fires appear to be more highly fragmented than randomly selected areas, particularly in relation to habitat features associated with trophic and humidity characteristics. These findings are consistent with research from the Amazon forests (Armenteras et al. 2013, Silva-Junior et al. 2018, Silva-Junior et al. 2022, Noble et al. 2025), which has demonstrated that the majority of forest fires – and the most severe ones – tend to occur within the first kilometre of forest edges, underscoring the increased vulnerability of fragmented forests to fire. In Poland, where the vast majority of forest fires are caused by human activity (NFFIS 2018), forest edges are more vulnerable to human penetration than areas deeper in the forest (Kolanek et al. 2021).

Landscape-level conclusions are supported by the analysis of differences observed within individual classes of forest habitat characteristics. For most age classes (0–20, 21–40, 41–60, and >100 years), forest patches in the areas surrounding fire sites were smaller and had shorter perimeters compared with at least one control group. A higher NumP indicates greater fragmentation of younger (<60 years) and the oldest forest stands in the vicinity of fire outbreaks. This pattern may be explained by forest management practices implemented by the State Forests of Poland, which involve relatively small-scale final felling of the oldest stands, followed by regeneration activities (State Forests 2012a).

The absence of significant differences in the 61–80 and 81–100-year age classes may indicate a higher degree of landscape stability during this successional phase. This could be associated with the fact that for most forest habitat types in Poland, the rotation age (i.e. the minimum age at which a stand can be harvested to achieve an optimal balance between costs and benefits; State Forests 2023) is typically >100 years (for the pine, which is the most dominant species in Poland; Report on the State Forests in Poland 2018). Therefore, forest stands within the 61–100-year age range are generally not subject to active large-scale management interventions that can change the forest site type.

In Central Europe, the highest fire risk was observed in young and middle-aged pine stands, whereas no increased fire risk was reported for maturing, mature or old-growth forests (Slavskiy et al. 2023). It is well established that the most significant risk of fire occurs in the most disturbed and transformed habitats – whether due to natural disturbances such as droughts, wind events, or bark beetle outbreaks (Berčák et al. 2023), as a consequence of forest management practices (Zald, Dunn 2018), or high density of roads (Kolanek et al. 2021). These factors may help explain the observed results.

Other significant differences observed within individual age classes – such as lower patch size variability in the 21–40 age class, higher patch compactness in the 41–60 age class, and more homogeneous patches in the >100 age class in the vicinity of fires – may result from the specific forest management practices applied within individual forest districts. Without considering FMPs, it is challenging to draw reliable conclusions about the underlying causes of these patterns.

For the mixed and deciduous forest stands that dominate in Poland (Report on the State Forests in Poland 2018) and dominate within sub-areas where fires occurred, patch size, as well as edge length and edge density, were lower in the A1 group, suggesting a more fragmented spatial structure in areas surrounding fire events. Lower variability in patch size and shape in A1 within mixed and deciduous stands indicates greater homogeneity of patches in the forest landscape in the surroundings of fire sites. This may be attributed to the relatively uniform size of final fellings and regeneration areas in intensively managed forests. By contrast, greater patch complexity was observed in group A1 subareas of pine-dominated stands, which are known to be more fire-prone (Krawchuk et al. 2006).

At the class level, the most crucial differences in the context of forest fire occurrence appear to be associated with humidity classes, given that one of the primary factors contributing to fire susceptibility is the nature of the fuel material (Davies et al. 2016, Shi et al. 2024). In dry forest stands, patches in the vicinity of fire locations were larger and less compact compared with those in other areas. Conversely, in more humid stands (fresh and boggy), patches were generally smaller. This is a significant finding, as dry stands are more susceptible to fire ignition and subsequent spread. In certain stands (e.g., fresh and boggy), the shape variability of patches was lower in areas surrounding fire sites, while the overall shape complexity was higher. Taken together, at the habitat class level, the landscape structure in fire-affected areas (group A1) is characterised by higher fragmentation, smaller patch sizes, and, depending on the habitat type, variable shape complexity. Similar patterns have been discovered by other researchers (Armenteras et al. 2013, Numata et al. 2017). These effects are particularly evident in young, mixed, and moist forest stands, where the landscape is more fragmented, and patches are either more irregular or more homogeneous, depending on the metric analysed. The observed characteristics of fragmentation, patch size, and patch number may serve as a basis for identifying areas more susceptible to fire outbreaks, thereby enabling more effective prevention strategies and faster firefighting response.

While landscape-level analysis provides a broad overview of structural instability, class-level analysis helps to pinpoint which forest types and developmental stages are susceptible to these effects. Both approaches demonstrate that landscapes surrounding fire occurrences are more fragmented and more complex in shape. However, the class-level results reveal that this effect is not uniform; somewhat, it varies depending on stand age, habitat type, and dominant tree species composition. Our conclusions emphasise the potential utility of landscape metrics in identifying areas that may be particularly susceptible to forest fires, without attempting an in-depth analysis of the underlying drivers shaping landscape structure. Such an analysis would require more comprehensive research, including detailed assessments of individual forest districts and their forest management strategies as outlined in operational planning documents.

The findings of this study demonstrate that landscape analysis can be integrated into forest management planning, particularly in anticipating and mitigating fire risk, especially since the research review conducted by Uuemaa et al. (2012) also showed that landscape metrics can be a tool for fire control. Further research is planned to explore the predictive potential of landscape metrics for assessing forest fire risk in a more detailed and systematic manner.

At both the landscape and class levels, the results were consistent: whenever a metric showed higher (or lower) values in areas where forest fires occurred compared with randomly chosen locations at one level, the same trend was observed at the other. They indicate that areas in the vicinity of fire occurrences exhibit greater forest structural fragmentation compared with the control groups, as reflected by smaller, more numerous, and more irregularly shaped forest patches. Specifically, the following patterns were observed:

• –

  At the landscape scale, regardless of the analysed forest habitat attribute, patches surrounding fire sites were consistently smaller and more irregular, suggesting increased fragmentation, particularly in relation to trophic and moisture-related characteristics.

• –

  At the level of individual habitat classes, similar patterns were observed in greater detail.

  Young stands (0–40 years) and mixed and fresh to boggy habitats were especially affected. In these classes, patches were smaller, more fragmented, and more numerous, indicating heightened landscape fragmentation.

These findings underscore the potential significance of forest structure in shaping fire risk. The implications could be significant, particularly in the context of Poland, where a single, state-owned organisation (State Forests) is responsible for managing nearly all forested areas. These results could inform national forest management strategies and be incorporated into fire risk forecasting systems.

Strategically planned forest management can help mitigate forest fire risk or, at a minimum, identify areas more vulnerable to fire, thereby enhancing the effectiveness of prevention measures and emergency response planning.