Using Dynamic, Fuels-Based Fire Probability Maps to Reduce Large Wildfires in the Great Basin

ABSTRACT Spatial and temporal dynamics of rangeland fuels is a primary factor driving large wildfires. Yet detailed information capturing variation in fine fuels has largely been missing from rangeland fire planning and fuels management. New fuels-based maps of Great Basin rangeland fire probability help bridge this gap by coupling dynamic vegetation cover and production data from the Rangeland Analysis Platform with weather and climate data to provide annual forecasts of the relative probability of large wildfire. In this paper, we review these new fuels-based maps and discuss implications for prefire planning, preparedness, and strategic fuels management. Examining patterns of fire probability through time reveals high spatial and temporal variation in risk of large wildfires across the Great Basin. Certain areas are chronically impacted with high fire probability most years, while others have more sporadic or low probability of large fire annually. Maps confirm previous research that the recent increase in large fire risk in the region is highly associated with invasive annual grasses, but total aboveground herbaceous production (including perennials) continues to be a primary predictor of fire probability. Fuels-based fire probability maps can be used alongside existing data sources and prioritization frameworks by fire and fuels managers to inform questions of 1) what kind of fire year might this be, 2) where large fires are most likely to occur given an ignition, and 3) where resources should be focused. We provide examples of how maps can be used to improve prefire preparedness and planning to enhance suppression, facilitate annual targeting of fine fuels reductions, and support land use planning for implementation of landscape-scale fuels management. Proactively incorporating this new information into rangeland fire and fuels management can help address altered fire regimes threatening the region's wildlife and working lands.


Introduction
Increasing wildfire size and frequency in North America's sagebrush ( Artemisia spp.) biome is imperiling ecosystem services nec-✩ This work was supported in part by the USDA −Natural Resources Conservation Service and Agricultural Research Service. The findings and conclusions in the publication are those of the authors and should not be construed to represent any official USDA or US government determination or policy. The USDA is an equal opportunity provider and employer. Proprietary or trade names are for information only and do not convey endorsement of one product over another.
* Correspondence: Natural Resources Conservation Service, West National Technology Support Center, 1201 NE Lloyd Blvd, Portland, OR 97232, USA E-mail address: jeremy.maestas@usda.gov (J.D. Maestas). essary to sustain wildlife and people ( Coates et al. 2016 ;O'Neil et al. 2020 ;Crist et al. this issue). Shrublands of the Great Basin have been disproportionately impacted by altered fire regimes due to interacting effects of invasive species, climate, land use and management ( Westerling et al. 2003 ;Balch et al. 2013 ;Fusco et al. 2019 ). Fire suppression and other factors have kept most fires (97%) from becoming large ( > 400 ha; Havlina et al. 2014 ), but those that escape initial attack can become megafires that account for most acres burned (Crist et al. this issue). Concern over impacts has led to calls for new strategic approaches to proactively reduce risks of large fires ( Murphy et al. 2013 ;Crist et al. 2019 ).
Wildland fire behavior is governed by weather, topography, and fuels. Besides addressing human-caused ignitions as a pri-  ( Smith et al. this issue ) and associated fire perimeters (purple) from the Monitoring Trends in Burn Severity dataset ( Eidenshink et al. 2007 ) for representative yr: 1993, 1998, and 2006. mary source of wildfire (Balch et al. 2017), fuels are one of the only things that can be managed proactively to potentially alter fire behavior and outcomes. In contrast with forested ecosystems where fuels are consistently abundant but vary in moisture content, semiarid and arid rangelands are considered relatively fuel limited where climate-driven variation in vegetation production drives availability of fuels to carry fire ( Abatzoglou and Kolden 2013 ). On sagebrush rangelands of the Great Basin, antecedent precipitation and accumulation of herbaceous fine fuels are primary predictors of large fires and total area burned each year ( Pilliod et al. 2017 ). Fine fuels and flammability of sagebrush rangelands have been highly modified by widespread invasion of exotic annual grasses, such as cheatgrass ( Bromus tectorum L.), where even low amounts on the landscape ( > 1% cover at 250-m spatial resolution) can increase the risk of wildfire ( Bradley et al. 2018 ). While invasive annual grasses justifiably receive much attention, fire is analogous to a generalist herbivore-indiscriminately consuming plant species ( Bond and Keeley 2005 )-so native herbaceous vegetation must also be considered in fire occurrence and risk ( Pilliod et al. 2017 ).
Probabilistic maps simulating fire risk are an important tool commonly used in fire and fuels planning and management ( Short et al. 2020 ), but current products lack detailed and timely information on vegetation cover and biomass that largely drive spatial and temporal variation in fire activity. However, recent breakthroughs in remotely sensed vegetation mapping lend new insights into rangeland fuels ( Jones et al. 2018Reeves et al. 2020 ;Rigge et al. 2020 ;Pastick et al. 2020 ;Allred et al. 2021 ). The Rangeland Analysis Platform (RAP; https://rangelands.app/ ) is a free online tool that provides continuous 30-m resolution estimates of vegetation cover (annual grasses and forbs, perennial grasses and forbs, shrubs, trees, bare ground) and production (herbaceous aboveground biomass) from the 1980s to present ( Jones et al. 2018Allred et al. 2021 ). Coinciding with availability of Landsat satellite data, RAP provides herbaceous production estimates at 16-d intervals, allowing land managers to better understand spatiotemporal change in fine fuels throughout a given year . Smith et al. (this issue) combined RAP's high-resolution (30m) vegetation functional group cover and production data (annual and 16-d estimates), with moderate-resolution (4 km) weather and drought data, in a machine learning framework to produce new dynamic, fuels-based maps of the relative probability of large ( > 400 ha) wildfire in Great Basin rangelands ( Fig. 1 ). Leveraging a 32-yr time series of spatial data, these new maps accurately cap-ture spatiotemporal variation in large fire probability, thereby improving predicative ability of fire forecasting on rangelands in the region ( Smith et al. this issue ). In contrast with other national probabilistic fire products, which estimate burn probabilities by simulating fire weather and fire behavior using static vegetation maps ( Short et al. 2020 ), the fire probability forecasts presented in Smith et al. (this issue) were developed specifically for Great Basin rangelands with the understanding that fire activity in these arid and semiarid ecosystems, unlike forests, is primarily fuel limited ( Abatzoglou and Kolden 2013 ). New dynamic maps of fire probability are made available each year in the spring to forecast upcoming fire conditions ( https://rangelands.app/great-basin-fire/ ). Historical predictions (back to 1988) of large fire probability and actual fire occurrence are also provided, allowing users to visually compare observed fire trends with forecasted fire probability over time and space. These maps complement existing national products by providing more fine-scale and dynamic tools for fire and fuels management planning and action to reduce fire size and frequency.
In this paper, we review fuels-based fire probability maps produced by Smith et al. (this issue) and elaborate on implications for prefire preparedness and fuels planning and management on rangelands of the Great Basin. First, we interpret and discuss general map patterns and trends of fire probability, along with underlying vegetation data, observed over the past 2 to 3 decades. Next, we illustrate how the maps can be used by land managers to inform expectations for the upcoming fire year, point to where large fires are most likely to occur given an ignition and where fuels management should therefore be focused, and help answer fundamental questions of "Why here?" and "Why now?" for proposed management actions. Finally, we present examples of applications in prefire planning and fine fuels management to spur thought on how maps can be incorporated into real-world scenarios.

Interpreting patterns of large fire probability in the Great Basin
High spatial and temporal variation in fire probability emphasizes need for flexibility Examining annual maps of large fire probability over the 32yr period produced by Smith et al. (this issue) reveals high spatiotemporal variation, especially across the northern and eastern Great Basin. Consistent with previous precipitation-based models of fire risk ( Pilliod et al. 2017 ), maps produced by Smith et al. (this issue) show the extent and location of areas exhibiting moderate-to-high probability of large fire changes dramatically  ( Eidenshink et al. 2007 ).
from year to year. This variation is illustrated well by comparing predictions of large fire probability in 1993 to 2006 (see Fig.  1 ). These patterns are primarily determined by precipitation and the subsequent accumulation of fine fuels in the previous 1 −3 yr ( Balch et al. 2013 ;Pilliod et al. 2017 ;Smith et al. this issue ). Productivity of arid and semiarid rangelands is known to be extremely variable due largely to fluctuations in precipitation timing and amount ( Holmgren et al. 2006 ). Forecasted fire probability was strongly correlated with total area burned and number of large fires each year at the scale of the Great Basin ( Smith et al. this issue ). However, actual fire occurrence is contingent on other factors as well, such as ignitions, weather, suppression, and drawdown of firefighting resources, as illustrated by actual large fire occurrence data from the Monitoring Trends in Burn Severity dataset ( Eidenshink et al. 2007 ). For example, fuels were conducive to widespread fire activity in 1998 but the fire season was relatively cool and wet and few large fires occurred (see Fig. 1 ). That said, the predictive implications of the model are relevant to managers whether or not predicted on-the-ground fire patterns eventually emerge, because fuels are one of the few factors that can be preemptively addressed by land managers to mitigate wildfire probability.
High interannual variation in fire probability over time, particularly in the northern and eastern Great Basin, underscores the need for land managers and agencies to be flexible in addressing wildfire risk in the region. Annual forecast maps provide a preseason estimate of large fire risk that can help managers request sufficient fire-suppression and fine fuels management resources and target those actions to mitigate risks when and where they arise. However, latitude in local decision making and regional resource allocation will likely be required to capitalize on this information in a timely manner. Like other domains of natural disaster preparedness, managers can use annual fire probability forecasts alongside other predictive tools to plan accordingly in fire-prone years.

Spatial and temporal context informs management approach
While interannual variation in fire probability is high, patterns emerge when examining large fire probability and actual fire occurrence data over time, highlighting the importance of landscape context in applying appropriate prefire and fuels management strategies. To visualize patterns of recent fire probability through time, we summarized pixel-level time series of forecasted fire probability between 20 0 0 and 2021, calculating the mean and trend in fire probability ( Fig. 2 ). Trend was estimated with the nonparametric Theil-Sen slope estimator ( Theil 1950 ;Sen 1968 ) and represents the median annual change during the period 20 0 0-2021. We also calculated the pixel-level burn frequency from 1984 to 2019 using all wildfires in the Monitoring Trends in Burn Severity dataset (see Fig. 2 ; Eidenshink et al. 2007 ). Resulting maps lend credence to the notion that certain portions of the Great Basin are chronically impacted by higher fire risk and likely require drastic changes in management to preserve ecosystem functions and services. Great Basin land managers and researchers have long recognized the impacts of increased fire activity in areas of abundant invasive grass fuels and called for alternative management paradigms to more effectively mitigate changes in fine fuels ( Coates et al. 2016 ;Perryman et al. 2018 ;Davies et al. 2021 ). Mean fire probability maps show that rangelands most often at risk of large fires occur in the northern and eastern Great Basin (see Fig. 2 ), which supports previous research documenting similar spatial patterns ( Brooks et al. 2015 ;Pilliod et al. 2017 ;Short et al. 2020 ;Pilliod et al. 2021 ). Most areas exhibiting an increasing trend in fire probability are situated in and around areas with higher mean fire probability and previously burned areas, suggesting some spatial order to changing fuel conditions and associated fire risk (see Fig. 2 ). Perhaps of equal importance, the mean fire probability across most of the Great Basin rangelands is either low or moderate (see Fig. 2 ). Therefore, a nuanced and targeted approach to placing the right fuels management strategies in the right places is warranted.
As an example, fire probability maps provide context for guiding appropriate prefire and fuels management actions that align well with an existing prioritization framework for addressing invasive annual grasses already being implemented by partners across the sagebrush biome ( NRCS 2020 ;WGA 2020 ;NRCS 2021 ;Creutzburg et al. 2022 ;Wollstein et al. 2022 ). Described as "Defend the Core, Grow the Core, Mitigate Impacts," this spatially explicit framework emphasizes proactive management to protect and expand large and intact rangeland areas ("cores") while focusing more intensive actions in heavily invaded regions to mitigate the worst of the invasive grass-fire cycle ( NRCS 2020 ;WGA 2020 ;Creutzburg et al. 2022 ). Thinking about this framework from a wildfire and fuels perspective shows utility of fire probability maps in determining strategic actions. For example, in areas of consistently low fire probability, fuels management would likely focus on enhancing monitoring and preventative management to keep fire risks low, such as Early Detection and Rapid Response to prevent invasive annual grass establishment and defend rangeland cores  ). In areas with moderate probability of large fire, management strategies might focus on opportuni- ties to defend and grow cores by prepositioning of firefighting resources in fire-prone years or targeted herbicide application along roadsides and other disturbance corridors to reduce invasive annual grass cover and seedbank ( Sebastion et al. 2017 ), thereby decreasing herbaceous vegetation flammability and opportunities for human-caused ignitions (Balch et al. 2017). In areas of chronically high fire probability, more intensive and long-term fire and fuels management infrastructure, such as strategic fuel break networks or regularly planned targeted grazing, may be more appropriate to mitigate impacts of frequent fire on life and property. Smith et al. (this issue) show that perennial and annual herbaceous vegetation were primary drivers of large fire occurrence on Great Basin rangelands, while shrubs and trees were relatively unimportant predictors ( Fig. 3 ). This is somewhat surprising given that herbaceous fuels are thought to be less important than woody fuels with extreme fire weather conditions under which large fires often occur ( Strand et al. 2014 ). This observation also stands in stark contrast to the emphasis of contemporary fuels management activities on shrubs and trees in the region ( Taylor et al. 2013 ;Crist et al. this issue ). Variable importance metrics confirm that the composition of the herbaceous understory is less important than total herbaceous biomass for predicting large fires ( Smith et al. this issue ). These findings highlight the need for fuels management to increase focus on herbaceous fine fuels when the goal is to minimize risk of large fires, especially on rangelands in the northern and eastern Great Basin. Current fuels management focused on woody plants remains important for many reasons, such as altering fire behavior to enable safe firefighting and decreasing fuel loading to reduce fire severity and postfire impacts , but a heightened emphasis on herbaceous fine fuels will be needed to lower the probability of large fires on rangelands.

Herbaceous vegetation drives large fire probability on rangelands
To further explore patterns of herbaceous vegetation driving large fire probability, we used RAP data Jones et al. 2021 ) to calculate and depict trends in total herbaceous aboveground biomass production, annual herbaceous aboveground biomass production, and perennial herbaceous aboveground biomass production for the period 20 0 0-2021 using the nonparametric Theil-Sen slope estimator ( Fig. 4 a). We also plotted total aboveground production for annual herbaceous, perennial herbaceous, and total herbaceous vegetation from 1988 to 2021, identifying years in which > 404 685 ha (1 million acres) burned in the Great Basin according to the Monitoring Trends in Burn Severity dataset (see Fig. 4 b; Eidenshink et al. 2007 ). Resulting figures illustrate spatial and temporal patterns of vegetation change underlying large fire probability models. While perennial herbaceous production has largely remained steady in recent decades, annual herbaceous production has become an increasing component of total aboveground production through time (see Fig.  4 ), particularly in the northern and eastern Great Basin. Notably, the amount of total aboveground production attributable to annual herbaceous vegetation exceeded that of perennial herbaceous vegetation for the first time in 2016 (see Fig. 4 b). Plotting production data alongside historic burn data illustrates pulses in herbaceous production that precede large fire years (see Fig. 4 b).
Invasion of cheatgrass and other exotic annual grasses is widely recognized as the most influential factor altering historic fire regimes on sagebrush rangelands ( Brooks et al. 2004 ;Brooks et al. 2015 ;, and their role in elevating fire risk and lengthening fire seasons is well documented ( Balch et al. 2013 ;Bradley et al. 2018 ;Fusco et al. 2019 ). Smith et al. (this issue) also found increasing cover and production of herbaceous annuals and declining bare ground corresponded with increases in fire probability, likely owing to increased fuel continuity and flammability. Perennial herbaceous plants have historically comprised the bulk of total aboveground production across the Great Basin, and this remains the case across most of the region. However, increases in annual herbaceous production indicate we may be reaching an inflection point where annuals exceed perennials in total production at a regional scale (see Fig. 4 b). Collectively, these findings support a primary focus on prevention, reduction, and mitigation of annual grass invasion to restore historic fire regimes in the region. Given the scope of the annual grass problem and challenges with managing it  ), a strategic approach to targeting management actions is prudent ( NRCS 2020 ;WGA 2020 ;NRCS 2021 ;Creutzburg et al. 2022 ).
While invasive fine fuels are clearly a priority for management emphasis, perennial herbaceous vegetation continues to strongly influence fire occurrence and size ( Davies and Nafus 2013 ;Pilliod et al. 2017 ;Smith et al. this issue ), so fuels management must consider the total accumulation of herbaceous vegetation (i.e., including perennial). Perennial herbaceous vegetation plays a keystone role in maintaining sagebrush ecosystem resilience to fire and resistance to invasion ( Chambers et al. 2007 ;Chambers et al. 2014 ); therefore, fine fuels manipulations must be carefully tailored and strategically targeted so as not to undermine perennial plant health ( Crist et al. 2019 ). Local land managers must weigh trade-offs and risks when planning herbaceous fuels management since loss of perennial plants makes sites more vulnerable to invasive annual grass conversion ( Chambers et al. 2007 ). Stopping all fire is not possible and is not typically an objective in native rangeland ecosystems, but properly planned and strategic management of total herbaceous fine fuels is an important part of mitigating fire risk. Using techniques such as dormant season grazing Davies et al. 2022 ) and virtual fencing technologies  ) may help achieve fine fuel reduction goals while minimizing unintended impacts on desired perennials. Trends in herbaceous fine fuels chiefly responsible for variation in large fire probability across rangelands of the Great Basin, including total herbaceous aboveground biomass production, annual herbaceous aboveground biomass production, and perennial herbaceous aboveground biomass production derived from the Rangeland Analysis Platform Jones et al. 2021 ). Trend estimates were calculated for the period 20 0 0-2021 using the nonparametric Theil-Sen slope estimator. B, Total aboveground production (lbs) plotted annually for annual herbaceous, perennial herbaceous, and total herbaceous vegetation from 1988 to 2021. Vertical dotted lines represent years in which more than 404 685 ha (1 million acres) burned in the Great Basin according to the Monitoring Trends in Burn Severity dataset ( Eidenshink et al. 2007 ).

Using fire probability maps to improve prefire planning and fuels management
Fuels-based mapping of large fire probability provides a powerful new tool to support strategic fire and fuels planning and resource allocation in the Great Basin, especially when combined with national wildfire risk assessments ( Short et al. 2020 ;. Increasing availability of high-resolution vegetation data ( Jones et al. 2018Rigge et al. 2020 ;Pastick et al. 2020 ;Reeves et al. 2020 ;Allred et al. 2021 ) now enables next-generation quantification of fuels on rangelands that bolsters fire and fuels effort s at local and regional scales ( Smith et al. this issue ). Here, we present examples of how Smith et al.'s (this issue) fire probability maps can be used alongside existing data sources to improve prefire forecasting and planning to enhance suppression, facilitate annual targeting of fine fuels reductions, and support land use planning for implementation of landscape-scale fuels management strategies.

Prefire forecasting to enhance preparedness and suppression
Fire managers currently leverage a variety of data sources and predictive services to generate outlooks for the upcoming fire season and continually assess fire danger. The National Fire Danger Rating System is a widely used standardized system for rating fire danger in the United States, with origins in forested ecosystems, that relies in part on fuels data. Geographically complete and timely information on fine fuels that drive rangeland fires has been historically inadequate, but new dynamic rangeland vegetation datasets help fill this gap ( Jones et al. 2018 ;Allred et al. 2021 ). Perhaps the most novel contribution of the forecasting framework presented in Smith et al. (this issue) is the integration of dynamic vegetation data with drought, weather, and climate to provide a quantitative metric of fire probability months ahead of the fire season across the entire Great Basin. Predictions provide early forecasts of large fire probability for the upcoming season, helping to support allocation of fuels management and suppression resources across and between large geographic regions. While forecasts can change as spring conditions shape fuel amount and moisture, early forecasts still perform well and can be useful in securing and mobilizing resources as early as possible ( Smith et al. this issue ). New maps also provide fine-scale spatial information to support integrated fire management planning by federal, state, and local fire managers (e.g., Rangeland Fire Protection Associations) to improve preparedness and fuels management to engage the fire before it starts ( Stratton 2020 ;Wollstein et al. 2022 ).
In 2021, the Burns Interagency Fire Zone and Vale Bureau of Land Management (BLM) in southeast Oregon began piloting the use of fire probability forecasts to augment prefire planning. Preparedness level assessments are a key component of local operating plans that inform fire restrictions, provide situational awareness, support resource requests, and feed other aspects of fire management. On sagebrush rangelands, traditional fire danger rating systems based on weather-derived indices useful in forest fire management often do not correlate well with rangeland fire history, primarily because fire risk is much more heavily influenced by fluctuations in fine fuels. Fire planners in the Burns Interagency Fire Zone and Vale BLM bolstered vegetation information in their Fire Danger Operating Plan by leveraging fire probability forecasts and historical data to comprehensively account for variation in fine fuel accumulations within specific Fire Danger Rating Areas ( Fig. 5 ). Incorporating new fire probability maps and data into planning enabled BLM fire planners to assess the predicted severity of the upcoming fire season by comparing with historical condi- tions (see Fig. 5 a) and visualize patterns of forecasted fire probability within Fire Danger Rating Areas to support on-the-ground observations of fuel accumulation and better focus resources (see Fig.  5 b). Combining spatially complete fire probability data with traditional indices (e.g., National Fire Danger Rating System, live fuel moisture data, and current and anticipated fire workload) allowed fire planners to better communicate weekly fire risks to fire management staff, agency administrators, and cooperators and improve preparedness (Casey O'Connor, personal communication). Furthermore, information aided decision makers with determining where to prioritize suppression resources and guide fire restriction discussions across a large geography. Incorporation of fuels-based fire probability data in similar assessments elsewhere could allow for more consistent and data-driven comparisons of fire risks and resource needs across the entire Great Basin.

Supporting planning of targeted fine fuels reductions
Fine fuels management is an important tool for strategic rangeland fire risk mitigation, particularly in areas of moderate-tohigh probability of large fire. Reduction of herbaceous vegetation reduces fine fuel continuity, flammability, fire spread, and engagement of shrub fuels ( Strand et al. 2014 ;Davies et al. 2017 ). Livestock grazing, mowing, prescribed burning, and herbicide treatments are commonly used methods for fine fuels reduction ( Diamond et al. 2009 ;Maestas et al. 2016 ;Shinneman et al. 2018 ). However, such actions can increase invasive annual grasses ( Davies et al. 2012 ;Pyke et al. 2014 ) and negatively affect desired plant communities, wildlife habitat, and other ecosystem attributes if applied haphazardly ( Keeley 2006 ;Reisner et al. 2013 ;Shinneman et al. 2020). Careful and strategic application of fine fuels reduction in space and time is needed to minimize undesired effects and foster greater stakeholder support for proposed actions ( Maestas et al. 2016 ;Crist et al. 2019 ). Annual fire probability forecast maps provide a nuanced view of fuel dynamics across the landscape to support timely and targeted fuels reduction. With this context, stakeholders are better positioned to assess trade-offs between vegetation removal and values at risk.
Targeted grazing to reduce fine fuels ( Diamond et al. 2009 ;Great Basin Restoration Initiative Workgroup 2010 ;Clark et al. this issue) is one practice that could be informed by annual fire probability forecasts. For example, the BLM recently worked with permittees and other partners on the T Lazy S Ranch in northern Nevada to demonstrate the utility of targeted grazing to protect priority habitats for greater sage-grouse (Centrocercus urophasianus) and mule deer (Odocoileus hemionus) ( BLM 2016 ). A 34-km long grazed fuel break was strategically placed adjacent to at-risk priority habitats where fire risks are high ( Fig. 6 ). In 3 out of 4 yr since being implemented, wildfires intersected the grazed fuel break, enabling firefighters to put out these fires before they could spread to adjacent priority habitats (Pat Clark, personal communication; see Fig. 6 ). Annual forecasts of fire probability provide detailed spatial data that can help managers determine where to prioritize targeted grazing and justify why such actions are needed in years of high fuel accumulation. Static fire probability mapping is already incorporated into BLM's fire risk assessment and 5-yr fuels budgeting process  ), but Smith et al.'s (this issue) annual forecast products can be used at more local levels to help managers allocate limited resources to specific project areas based on predicted fire probability.
In portions of the northern and eastern Great Basin chronically impacted by invasive annual grasses and high risk of large fire, a sustained and large-scale fuels management approach to protect ecosystem services and rural communities is needed ( Perryman et al. 2018 ;Davies et al. 2021 ). Excluding suppression costs, the BLM spent $210 million on rehabilitating burned federal lands over the past decade alone ( Pilliod et al. 2021 ). Since 1990, roughly 26% of those restoration seedings have at least partially reburned, potentially undermining recovery goals and taxpayer-funded investments ( Pilliod et al. 2021 ). Mitigating fire risks in and around these areas likely requires integrated management strategies ( Perryman et al. 2018 ;Davies et al. 2021 ), such as implementation of strategic fuel break networks ( Shinneman et al. 2019 ;Clark et al. this issue ), incorporation of dormant season grazing Davies et al. 2022 ), and enhanced prefire planning and suppression ( Stratton 2020 ;Wollstein et al. 2022 ). Yet considerable uncertainty surrounds the efficacy of fuels treatments and questions remain about unintended ecological effects ( Shinneman et al. 2019 ). Existing frameworks based on ecosystem resilience and resistance ( Chambers et al. 2017 ), resist-accept-direct decision making ( Schuurman et al. 2022 ), and geographic strategies for invasive annual grass management ( NRCS 2020 ;WGA 2020 ;Creutzburg et al. 2022 ) can help land managers prioritize the right actions in the right places to mitigate ecological transformation in an era of climate change. When combined with these frameworks, fire probability maps provide additional science-based context for evalu- Figure 6. Targeted grazing fuel break (black polygon) on the T Lazy S Ranch allotment in Elko County, Nevada, depicted relative to the 2021 fire probability ( Smith et al. this issue ), wildfires (purple) since targeted grazing implementation Eidenshink et al. 2007 ), greater sage-grouse priority areas for conservation (crosshatch), and prevailing wind direction. Photograph shows the Welch Fire (right of fence) that ignited July 2021 and intersected the fuel break (left of fence), which had been intensively grazed to a stubble height of 5 −8 cm in the spring. Since 2018, three wildfires intersected the fuel break (red circles) allowing firefighters to put them out before spreading to high-value habitat resources. (Photo courtesy: Pat Clark.) ating risks and trade-offs of potential fuels management actions ( Wollstein et al. 2022 ).
Overlaying proposed fuel break networks on fire probability maps could help managers determine the most appropriate landscapes to prioritize this practice, where the goal is to mitigate risk of large fire. Fire managers in the sagebrush ecosystem have identified strategically placed fuel breaks as a necessary fire-suppression tool to reduce fire size and frequency by improving firefighter access and minimizing response times, providing safe and strategic anchor points for suppression, and compartmentalizing wildfires and constraining fire growth. The BLM recently put forth a plan to construct and maintain a strategic network of linear roadside fuel breaks on up to 17 703 km (11 0 0 0 mi) across the Great Basin ( BLM 2020 ), nearly doubling the amount of fuel breaks in the region ( Shinneman et al. 2018 ). Overlaying the BLM's proposed fuel breaks with mean fire probability provides data-driven context on large fire risk that stakeholders can combine with other resource data to better assess trade-offs ( Fig. 7 ). Targeted implementation of fuel breaks in areas of chronically high fire risk, particularly those with low resilience to fire and resistance to invasion, represents a scenario in which potential benefits of proposed actions in terms of protecting unburned sagebrush, limiting annual grass conversion, and maintaining prior restoration investments likely outweighs risks, especially when combined with a long-term commitment to monitoring and maintenance ( Crist et al. 2019 ;Pilliod et al. 2021 ;Crist et al. this issue ). Conversely, the mean fire probability map also highlights landscapes typically at low risk of large fire where ecological and economic trade-offs of fuel break construction may be more debatable (see Fig. 7 ).

Implications
With fire probability forecasts now publicly available before each fire season in the Great Basin, the convergence of factors conducive to large rangeland fires should be met with a greater level of awareness and preparedness. Proactively incorporating this new information into rangeland fire and fuels management can help address altered fire regimes threatening the region's wildlife and working lands. Fuels-based fire probability models further re- inforce that accumulation of herbaceous vegetation plays a primary role in the incidence of large rangeland fires, emphasizing the need for heightened management focus on fine fuels. Increasing contributions of annual herbaceous vegetation to total aboveground production are concerning and support current efforts to prioritize and rapidly scale up strategic management of invasive annual grasses, especially considering native rangelands are transitioning to an annual grass −dominated state at an average rate of > 2 300 km 2 per year ( Smith et al. 2022 ). Patterns of fire probability through time provide place-based context showing some areas to be chronically impacted by large fire risks while others are only sporadically impacted or consistently low. Extreme interan-nual variation in fire probability across the region highlights the importance of flexibility in budgeting, resource allocation, and land use decision making.
Dynamic fuels-based forecasts of rangeland fire probability augment existing wildfire risk assessments by providing critical context on changing fuel conditions, thereby helping land, fuels, and fire managers to determine when and where to strategically target appropriate management responses. With uncertainty surrounding efficacy of fuels management, continued collaboration between researchers and field practitioners is needed to coproduce evaluations of such actions in order to quantify outcomes and improve the adaptive management feedback loop ( Naugle et al. 2019 ). Large-scale fuels management actions are already occurring or planned in the Great Basin, so abundant opportunities exist to couple land treatments with monitoring and evaluation to learn while doing.
While we focus here on demonstrating the utility of fuelsbased fire probability mapping in the Great Basin, implications of this work extend to other ecoregions impacted by invasive grasses and altered fire regimes. Spatially complete remote-sensing vegetation mapping now enables scientists and managers to integrate dynamic fuels data into fire forecasting and risk assessments across the continental United States (see https://rangelands. app/ ). Incorporating dynamic fine fuels data into fire probability modeling could be especially helpful in managing diverse ecosystems affected by fire-prone invasive grasses, from warm deserts  to forests ( Fusco et al. 2019 ;Kerns et al. 2020 ). Dynamic products capable of reflecting the spatial and temporal variation in fuels and fire risk are an increasingly important adaptation tool as climate change and invasive species continue to affect fire regimes in novel and surprising ways.