Introduction

Minimizing negative impacts of wildfire is a major societal objective in fire-prone landscapes around the globe (Gill et al. 2013). Due to a range of factors including climate change and a history of fire suppression, wildfire activity is increasing in many ecosystems and may out-pace current efforts to constrain wildfire and restore burned landscapes (Calkin et al. 2014; Dennison et al. 2014; Balch et al. 2017; Keyser and Westerling 2017; Fusco et al. 2019). For example, in the western United States, burned area from large wildfires (> 405 ha) averaged more than 1.25 million ha per year between 2010–2019 and 2.5 million ha burned in 2020 (Higuera and Abatzoglou 2020). The likelihood of very large wildfires (> 20,234 ha) is predicted to increase under changing climate conditions (Stavros et al. 2014). Wildfires spread across the landscape, responding to a suite of factors including combustible vegetation, weather, and topography, thus designing fire management strategies at the scale of a large landscape or extensive ecosystem can be a considerable challenge. Models can be used to better understand how fire spreads, how landscape characteristics and vegetative fuels facilitate or impede spread, and the ways in which pre-emptive fire management can limit fire-related losses.

Fire behavior models rely on physical (Sullivan 2009a), empirical (Sullivan 2009b), and mathematical (Sullivan 2009c) concepts and data to simulate the process of fire spread. The specific goals, complexity, and outputs of these fire behavior models vary widely. For example, coupled atmosphere-fire models can be used to simulate large fires and their spread in real time with the goal of supporting active firefighting and associated mitigation actions (e.g., WFS-SFIRE; Kochanski et al. 2013). Other fire simulation models characterize and summarize the potential risk and behavior of fires, and this information used in planning management actions (e.g., fuels reduction) and hazards/risk planning. Simulations of many possible individual fires can be used to calculate pixel-based probable rates of spread, conditional flame length estimates, and burn probability statistics (e.g., FlamMap; Finney 2006; Finney et al. 2011). These models provide insight into the interactions among fuels, weather or climatic conditions, and fire outcomes. However, it is challenging to conduct fire behavior simulations over large landscapes due to computational constraints. Further, while rate-of-spread (ROS) maps describe expected properties of fires, connectivity outputs describe the landscape and the spatial arrangement of pixels with similar properties that could facilitate contiguous fire spread. The connectivity outputs are therefore useful for visualizing the linkages among those areas with high rates of spread based on the density of potential connections.

Because fire is a contagious process, connectivity models provide a useful lens to extend and complement existing fire behavior models (Rodriguez Gonzalez et al. 2008). Connectivity of fuels has been associated with increased wildfire size and shorter fire return intervals (Miller and Urban 2000; O’Donnell et al. 2011; Balch et al. 2013; Fernandes et al. 2016). Extending this understanding beyond fuels to also incorporate the process of fire spread, we define ‘fire connectivity’ as the landscape’s capacity to transmit fire from one point on the landscape to another. Circuit-based connectivity methods, which have been widely applied within landscape and wildlife ecology (Dickson et al. 2019), offer a compelling method for understanding the patterns of potential fire connectivity (Gray and Dickson 2015). Circuit methods provide a cumulative estimate of connection probability based on the mathematics of electrical current flow (McRae et al. 2008) and can be used to model and visualize connections across large landscapes. Current flows across a cost surface of pixels representing landscape conductance that approximates a random walk and evaluates the relative ease of moving through many possible paths of wildfire progression. Circuit-based analyses have been used to analyze local-level fire connectivity among cheatgrass patches in the Mojave Desert (Gray and Dickson 2016) and compared with empirical and simulated fires (Gray and Dickson 2015). However, circuit-based assessments of fire connectivity have not been developed at ecoregional- or biome-scales. Assessing potential paths for wildfire across such broad scales may help to identify how different fuel treatment design and placement strategies could increase fire suppression efficiency.

Fire connectivity analyses can be used to identify opportunities for management actions to disrupt potential fire spread, and fuel breaks are a key management tool to accomplish the objective of reducing wildfire spread at landscape-scales. Fuel breaks are typically constructed as linear strips along roads in which fuels are reduced, redistributed, or enhanced (e.g., by increasing fuel moisture), with the primary goals of altering fire behavior (i.e., reduce fire-line intensity) and providing safe access for fire crews to stage or actively combat fire fronts (Agee et al. 2000; Syphard et al. 2011; Shinneman et al. 2019). The placement of fuel breaks has been historically designed to protect valued resources such as human infrastructure or ecologically important areas at high risk of fire and placed in locations that support fire suppression operations and logistics. However, the siting of new fuel breaks is often assessed at relatively local scales, typically determined by local land management offices. Although locally relevant siting is important and practical, decisions of where to develop new networks or extend existing networks could additionally be supported by fire connectivity maps. Comprehensive, broad-scale assessments of fuel-driven fire connectivity across large landscapes can help optimize designs of fuel break networks to more efficiently disconnect large tracts of flammable vegetation.

In the western United States fuel breaks are a large component of wildfire management efforts (Hoover 2017; Bureau of Land Management 2020). The Bureau of Land Management has approved plans to significantly expand the existing fuel break network through implementation of ~ 17,700 km of new or enhanced fuel breaks across the Great Basin region over the next several years (Shinneman et al. 2019, Bureau of Land Management 2020). Future fuel break implementation plans could be supported by broad-scale pattern analyses that evaluate fuel break configuration options in the context of fire connectivity. Fuel break placement and maintenance could be prioritized, for example, in locations that optimally disconnect areas of high potential fire connectivity and contiguous, highly flammable fuels.

We developed an approach to assess the spatial patterns for fire connectivity in and among areas with high potential rates of fire spread. To do so, we applied circuit-based analyses using an omnidirectional circuit theory algorithm (Omniscape; McRae et al. 2016; Landau et al. 2021) to assess fire connectivity across a large region. Specifically, we sought to identify pathways that connect areas of high rate of fire spread in the Great Basin. We expected that fire connectivity and relative flow patterns would reflect the heterogeneity of the landscape and provide insights into the variation of potential fire spread. To our knowledge, this is the first application of omnidirectional circuit theory to characterize fire connectivity and use results to identify areas where targeted management actions could disrupt key linkages among areas with high potential rates of fire spread. We developed fire connectivity pattern metrics, summaries, and examples to explore how such analyses could support spatial fuel management such as fuel reduction treatments, installing new fuel breaks and networks, and prioritizing the maintenance of existing fuel breaks.

Methods

Study region

The sagebrush-steppe ecosystem and sagebrush-dominated landscapes in the western United States are considered imperiled, due in large part to nonnative annual grass invasion that affects the frequency, severity, and size of wildfires (Noss and Peters 1995; Shinneman 2020). The Great Basin is comprised of many plant species (sagebrush, perennial grasses) that are slow to recover after fire and it supports numerous wildlife species of concern (Wisdom et al. 2005; Rottler et al. 2015). We focused on two main ecoregions of the Great Basin that are central to the planned Bureau of Land Management (BLM) (2020) fuel break expansion: the Northern Basin and Range (Ecoregion Level III, 10.1.3) and Central Basin and Range (10.1.5, Bureau of Land Management 2020). The majority of land in the Great Basin (~ 76%) is managed by federal agencies, with high potential for coordinated, broad-scale landscape management actions (e.g., among thirty-five BLM field offices that intersect the study area; Fig. 1).

Fig. 1
figure 1

Existing and planned fuel breaks in the Great Basin. Analysis of the future fuel break network includes both the existing and planned fuel breaks. Line thickness of fuel breaks are exaggerated for visibility (most fuel breaks are less than 200 m in width; Shinneman et al. 2019). Bureau of Land Management field office administrative boundaries are delineated by white lines

Fire connectivity model

We used an omnidirectional circuit theory approach to model the probability of fire spread within the Great Basin shrub-steppe ecosystem. We performed the analysis using the Omniscape connectivity modeling algorithm (v0.5.6, Landau et al. 2021) in Julia programming language (v1.5.2, Bezanson et al. 2017). The Omniscape algorithm is designed for efficient computation over large landscapes. It uses a moving window approach to connect any pixels meeting specified thresholds, in contrast to the related Circuitscape algorithm (McRae and Shah 2009) that calculates connections among discrete patches or nodes placed in specific locations, i.e., in patches or map edges. We used the moving window approach to connect pixels within the window as sources and grounds in order to identify and characterize the pathways or linkages among areas with high rates of spread. Requirements for the Omniscape algorithm include a resistance or conductance surface, threshold value for pixels to be considered sources, and the selection of a moving window size.

We developed a conductance surface for our connectivity analysis using a modeled fire rate of spread (ROS). The ROS surface was calculated using the FlamMap fire mapping and analysis system (Finney 2006) at 30 m resolution using Scott and Burgan’s 40 fire behavior fuel models (fuel models represent classes of live/dead vegetation types for input into fire behavior models, Scott and Burgan 2005) with fuel disturbances updated to 2019 conditions from the base LANDFIRE 2016 dataset (https://landfire.gov/getdata.php). To model fire behavior that is likely to be targeted for fire suppression activities, we used 15 mph upslope winds and 97th percentile fuel moisture conditions (3%, 4%, and 5% fuel moisture for 1 h, 10 h, and 100 h fuels, respectively, and with 45% live herbaceous and 90% live woody fuel moistures; Keane 2015). We used these conditions to reflect landscape fire conditions where suppression activities are feasible; with greater winds suppression activities are less effective or are abandoned because they are too dangerous to firefighters. Our attempt here is to use fuel and weather conditions where fuel breaks could be used for active suppression efforts. Fuel breaks and other fuel treatments were not explicitly included in the underlying ROS calculation and so passive effects of fuel breaks (i.e., reduced current flow due to treated fuel conditions within breaks) were not considered in those calculations. Hence, results describe potential fire connectivity in the absence of fuel breaks and related fuel treatments. We aggregated the resulting ROS surface to 270 m resolution for computational efficiency. To reduce barrier effects, non-target landcover (e.g., irrigated agriculture) was assigned a value by randomly sampling the ROS values from the target landcover cells, and subsequently masked out of final outputs. We also created a buffer around the study region to reduce potential edge effects. The buffer extended to the radius of the moving window (described below) and was filled with the same distribution of values as the non-target landcover pixels (see Supplemental Information Fig. 1).

Omniscape calculates linkages among pixels within a moving window by calculating the cumulative current flowing through each pixel (Landau et al. 2021). Current originates from source pixels with high rates of spread in the conductance map and flows to target pixels within the moving window. We set the source threshold parameter equal to the 75th percentile of the ROS values across the Great Basin (ROS = 39 chains/hour) to represent only those areas with high rates of spread. Any pixels greater than this threshold were considered sources and targets in the moving window analysis. To identify potential fire connectivity patterns that are relevant to the spread of large wildfires, we set the moving window radius at 78.03 km (270 m × 289 cells) to encompass the maximum radius achieved by a wildfire or wildfire complex in the study region from 2000 to 2019 (~ 78 km in 2007, Welty and Jeffries 2020). The full Omniscape initialization file can be found in Supplemental Information (Fig. 2).

Fig. 2
figure 2

Omnidirectional circuit theory output showing potential fire connectivity in the Great Basin based on cumulative current density, visualized with 20 quantiles

To characterize fire connectivity in terms of the number of potential paths that connect high rate-of-spread areas across the full extent, we calculated the cumulative current density across the landscape. The cumulative current density calculates the sum of potential paths over which a process can spread for each pixel on the landscape. For visualization, we conducted a standard raster log-transformation and binned the map into 20 quantiles (McRae et al. 2016; Landau et al. 2021). To differentiate paths of highly constrained flow (e.g., pinch points of flow) from those of diffuse flow (able to move among many paths), we used the normalized cumulative current density output; this output is calculated by dividing cumulative density by the current flow expected under null resistance conditions whereby sources remain the same but all resistance values are set to 1 (Landau et al. 2021). This normalized cumulative current density represents the relative flow of a process and characterizes whether it is impeded (relative flow < 1), diffuse (relative flow ~ 1), or channelized (relative flow > 1). We used the standard deviation of relative flow values (SD = 0.2362) to classify the values into five categories (Table 1) and adapted the interpretation of relative flow classes from McRae et al. (2016) to describe fire connectivity (Table 1). We also include the expected relationships between the relative flow classes and fuel break placement (Table 1).

Table 1 Relative flow classes of fire connectivity derived from Omniscape normalized cumulative current (NCC) density values, adapted from McRae et al. (2016) and Jones (2016)

Fuel break distribution and density

To understand the current and planned fuel break network in the context of fire connectivity results, we overlaid the existing (Shinneman et al. 2018) and planned (Bureau of Land Management 2020) fuel break networks with maps of potential fire connectivity and relative flow. We used a 9 × 9 cell focal window (~ 1 km) around pixels with fuel breaks to calculate the mean connectivity value and summarize local fire connectivity around fuel breaks. We used a Spearman rank correlation test to evaluate how fuel breaks were distributed relative to the potential fire connectivity classes (Koen et al. 2014; Gray and Dickson 2015). We also considered the density of fuel breaks in a region may contribute to the effectiveness of the fuel break network, particularly compared with the relative flow classes that represent locally relevant patterns. We calculated local fuel break density as the number of pixels that were occupied by fuel breaks within the local neighborhood of 8 km. The neighborhood size represents the 95th percentile radius of fires or fire complexes that occurred between 2000 and 2019 (Welty and Jeffries 2020). We used polygons for hydrological units (HUC10 watersheds) and BLM field office management units (Fig. 1) to represent general ecologically and management-relevant regions within which to summarize and compare fuel break network characteristics. We compared average fuel break density and dominant relative flow class for the existing fuel break network, as well as the planned network (i.e., including both existing and planned fuel breaks).

Results

Regional potential fire connectivity

The cumulative current density output illustrated spatial patterns of relatively higher and lower potential fire connectivity (Fig. 2). Distinct areas of higher fire connectivity represent many cumulative pathways for fire to spread and to connect areas with predicted high rates of spread (Buchholtz and Kreitler 2023). These well-connected areas and high rates of spread could therefore indicate greater potential for extensive wildfires. Areas with lower fire connectivity represent lower cumulative density of pathways for fire to spread. At a broad scale, these fire connectivity patterns highlight linkages among areas expected to have landscape and fuel characteristics that can support fires capable of spreading to proximate highly flammable areas. Fire connectivity results do not represent individual fires nor fire behavior explicitly, and results do not express the likelihood of ignition or risk of fire.

We found that existing fuel breaks were evenly distributed across the twenty potential fire connectivity classes (Spearman rank correlation test: S = 1544, rho = − 0.1609, p < 0.496; Fig. 3). Local patterns varied, with the area near existing fuel breaks occurring in areas of high, moderate, as well as low potential fire connectivity patterns (Figs. 3 & 5). The future fuel break network, which included both existing and planned fuel breaks, showed a strong negative rank correlation with potential fire connectivity, (Spearman rank correlation S = 2654, rho = − 0.9955, p < 0.0001; Figs. 3, 5) with more fuel breaks planned in lower fire connectivity classes.

Fig. 3
figure 3

Distribution of area classified as within fuel breaks across potential fire connectivity classes for the existing and future fuel break networks. Note, connectivity classes are equal in area across the study region

Regional patterns of relative flow

Using normalized cumulative current density as a measure of potential relative flow of fire across the Great Basin, we found that the landscape is characterized by a distinct fire connectivity gradient where relative fire flow is either impeded, diffuse, or channelized (Figs. 4, 5). Relative flow patterns were highly heterogeneous and varied across watersheds and regionally (Buchholtz and Kreitler 2023). Areas with relatively impeded or channelized flow emerged as a result of the landscape configuration and spatial variation in modeled fire ROS. Some fuel breaks were placed in areas with relatively impeded flow (Fig. 5A), while others were placed or planned for areas that intersected with relatively channelized flow (Fig. 5B).

Fig. 4
figure 4

Omnidirectional circuit theory output showing relative flow patterns in the Great Basin based on normalized cumulative current density. The output represents how the landscape shapes cumulative patterns, for example, by impeding or channelizing flow paths. White lines indicate Bureau of Land Management (BLM) Field Office boundaries

Fig. 5
figure 5

Rate of spread, potential fire connectivity, and relative flow patterns all provide different information about how fire may spread across the landscape. These patterns can, in turn, be used to provide context for the placement of fuel breaks with the goal of efficiently limiting fire spread. A. (left to right) A region of relatively moderate rate of spread. This region spans a gradient of potential fire connectivity values indicating variation in the density of linkages among high-spread areas for fire. Relative flow varies as well. Existing fuel breaks in this region exist primarily in the low potential fire connectivity and the already impeded areas. B. (left to right) A region of heterogeneous rates of fire spread. Here, potential fire connectivity values may be relatively similar in pattern to the rates of spread, while the relative flow classes highlight areas with channelized flow where impeded regions serve to channelize the flow. There are few existing fuel breaks, and the proposed fuel breaks cross many of the channelized flow paths

The dominant pattern of relative flow varied across the 35 BLM field office units in the Great Basin (Figs. 3, 6, Supplemental Table 1). Impeded or diffuse-impeded flow characterized the majority of 4 field offices, indicating landscape constraint or barriers to flow. Fourteen field office regions were dominated by diffuse flow, indicating many directionally unconstrained potential fire paths. Six field offices were dominated by diffuse-channelized flow and eleven by channelized flow, representing where potential paths linking areas with high rate of spread were more concentrated than expected.

Fig. 6
figure 6

The A mean connectivity class and B dominant relative flow class for each BLM field office, grouped by state

Most watersheds were dominated by diffuse relative flow (n = 604). Watersheds characterized by mostly diffuse-channelized relative flow were limited (n = 24, Table 2). The highest fuel break densities were in watersheds with high amounts of diffuse-impeded relative flow areas; conversely, lower fuel break densities were in watersheds dominated by channelized relative flow (Table 2, Fig. 7). Average local (8 km) fuel break density was higher in watersheds with diffuse and diffuse-impeded flow patterns than in ones with more channelized flow patterns. However, there was a broad range of densities across the landscape, especially within the diffuse-impeded relative flow class. The maximum local fuel break density by watershed was 22.7 km2 per 100 km2 in the current network and was 29.0 km2 per 100 km2 in the future fuel break network (i.e., including both existing and planned fuel breaks). Both maximum densities occurred in watersheds dominated by diffuse-impeded relative flow.

Table 2 Placement of fuel breaks in the context of fire connectivity and dominant relative flow classes, summarized within watersheds
Fig. 7
figure 7

Density of local fuel breaks in Great Basin watersheds, grouped by the dominant relative flow class for each watershed. Box plots depict the minimum, first quartile, median, third quartile, and maximum values for each relative flow class for the current and future fuel break networks

We summarized potential fire connectivity and relative flow patterns across BLM field office units by state (Fig. 6, Supplemental Table 1). Field offices with higher connectivity classes represent regions with high fire ROS and strong connectivity among areas with high ROS. Field offices with lower connectivity classes represent regions that may have lower fire ROS and are distant or poorly connected to areas with high ROS.

Discussion

Altered wildfire regimes can pose considerable social, economic, and environmental challenges, and different analytical tools are needed to understand, plan, and manage them. Our approach to modeling the contagious nature of the fire process by quantifying potential fire connectivity permitted the identification of areas on the landscape that might facilitate or impede fire spread. The spatially heterogeneous fire connectivity patterns produced by our model reflect landscape variation of fuels and topography that influence rate of fire spread, and may reveal potential concentrations of fire spread via pathways capable of connecting highly flammable areas. Moreover, by assessing the distribution of existing and planned fuel breaks within our model framework, we’ve provided an opportunity to examine where these fuel treatments might most effectively disrupt fire spread vis-à-vis areas with either potentially high fire connectivity or highly concentrated fire connectivity relative to surrounding conditions.

Fire connectedness among regions with fuel and landscape characteristics that accelerate fire spread (Fig. 2) represent potential pathways that can facilitate fire spread across large portions of the study area, potentially resulting in larger fire sizes. Variation in fire connectivity patterns indicate that regions with higher or lower connectivity values may benefit from using different fuel break strategies to disconnect fuels. For example, BLM field offices in north-central and western Nevada exhibited high fire connectivity (Figs. 2 and 6A). Areas such as these, with generally high potential fire connectivity, may require a greater concentration of fuel breaks to reduce or disrupt fire compared to areas with lower fire connectivity (although ecological trade-offs should be considered; see Shinneman et al. 2019). In regions with lower fire connectivity, managers may consider other strategies to reduce fuels and wildfire risk, including using local patterns of relative flow (e.g., Figs. 4 and 5) combined with local expertise to spatially refine fuel treatment strategies to disrupt specific areas with highly concentrated fire connectivity potential on the landscape.

Indeed, the relative flow patterns highlight the diverse ways that landscape patterns could facilitate or impede the transmission of fire. Different types of relative flow (diffuse, impeded, channelized) may indicate different levels of constraint on fire spread due to spatial configurations of landscape characteristics (e.g., topography) and fuels that either support or inhibit high rates of fire spread. In areas of diffuse flow, fire may be less directionally constrained and move primarily in response to local wind, weather, and fire conditions. Higher fuel break densities arranged in a network design may provide opportunities to access and combat active fires that could come from multiple directions in these types of environments (Maestas et al. 2016). Accordingly, we found that the density of fuel breaks was spatially correlated with diffuse or diffuse-impeded relative flow areas on average (Table 2), perhaps because these areas have gentler terrain with existing road networks that facilitated fuel break construction. By contrast, fire spread may be more directionally constrained in areas with relatively channelized flow. In these areas, the landscape configuration of fuels limits possible fire paths, and thus fewer but more strategically placed fuel breaks that bisect channelized pinch points could disconnect linkages among high ROS areas. Gray and Dickson (2016) also used a circuit model approach to suggest that the placement of a few, well-placed fuel breaks might be highly effective at reducing potential fire spread among highly flammable areas. These types of results highlight the potential utility of a circuit-based approach to inform installation locations for fuel breaks, particularly those placed in highly channelized flow areas that could yield disproportionate benefits.

Existing fuel breaks were evenly distributed across fire connectivity classes. Future fuel breaks, though, are generally planned for areas with moderate or lower fire connectivity thus increasing the density of the fuel break network in those areas. Across our study area landscape, we estimated that the average density of fuel breaks will roughly triple with the implementation of the BLM planned fuel break network. However, based on our assessment, key areas of diffuse flow are expected to still lack fuel break networks in the future, as will many isolated locations with channelized flow in which a few fuel breaks could be highly strategic.

Ideally, a combination of fire modeling, local knowledge, and logistical considerations would result in the lowest fuel break density that achieves the highest possible return on reduction of wildfire, particularly to protect valued resources in areas where fire is no longer ecologically beneficial due to extensive exotic annual grass cover and dramatically decreased mean fire return intervals (Fusco et al. 2019). Although there may be reasonable ecological, logistical, or fire-risk reasons for some areas to not have fuel breaks, further work is needed to describe and understand the underlying patterns driving fuel break placement. There may be opportunities to increase effectiveness of fuel break networks by including fire connectivity as one of the criteria that informs fuel break design and placement. Fire connectivity results could be paired with fire behavior or cumulative fire risk models to enrich information used in the design of fuel break networks (Rodriguez Gonzalez et al. 2008). Precise placement will depend on many other considerations as well, including locations of existing and accessible roads, fuel break maintenance logistics, locations of cultural and natural resources, and conservation objectives (Syphard et al. 2011; Shinneman et al. 2019).

The omnidirectional circuit theory approach can complement existing fire behavior models by assessing the spatial patterns associated with areas capable of burning with high rates of spread, but it has limitations and assumptions that must be recognized. The fire connectivity analysis does not model individual fires or their behavior and should not be interpreted to estimate fire probabilities or assess cumulative fire risk. Our approach did not explicitly incorporate likelihood of ignition or burn probability; therefore, it is possible to have areas where there would be high connectivity for fire to spread, but low probability for a fire to start, and therefore low likelihood that fire would spread across a modeled fire connectivity linkage there. However, even if likelihood of fire ignition is low, preventative actions may be appropriate in areas of higher potential fire connectivity for areas that would have substantial negative consequences from wildfire (e.g., the wildland urban interface). Ignition and burn probability could be incorporated into future research using circuit theory methods, for example, by using those variables to inform relevant source layers for the circuit analysis. Alternatively, our output highlighting fire connectivity potential could be combined post-hoc with maps of ignition rates (see Balch et al. 2017) and output from large fire behavior models that are often used to generate fire risk assessments and to help locate effective fuel treatments (e.g., FSim; Finney et al. 2011). For example, areas with high potential fire connectivity could be superimposed on or included in assessments of wildfire risks to highly valued resources based on burn probability and fire intensity, as well as analyses of wildfire risk transmission across landscapes (Ager et al. 2012, 2014; Calkin et al. 2011; Scott et al. 2013; Dillon et al. 2015; Gonzalez-Olabarria et al. 2019; Kreitler et al. 2020). In addition, this information could be incorporated into new developing spatial fire planning efforts for large fire containment such as Potential Wildfire Operation Delineations (PODS; Thompson et al. 2016; Wei et al. 2018). Further research is needed to understand the degree to which fire connectivity analyses can practically improve fuel break strategies and effectiveness.

Wildfire will continue to be a considerable ecological and economic challenge in the sagebrush biome, and it requires that we develop tools to understand and mitigate wildfire from a landscape perspective. We adapted theoretical and analytical connectivity methods used to describe gene flow and animal movement, with resulting novel outputs for identifying linkages among quick-burning patches and visualizing directional constraints at larger scales that may shape how wildfires spread. Where large wildfires are common, identifying areas with high fire connectivity potential can provide insights into strategic locations for fuel treatments to intentionally disconnect and reduce fire spread. The circuit theory approach can provide a new understanding of how landscapes facilitate wildfire spread, complementing existing data and models, and offering opportunities to further investigate connectivity-based approaches to fire management.