Research

INTRODUCTION

Wildfire in California is an already pervasive and worsening problem, made more challenging by the expense and difficulty of managing fuels at a large scale (Thompson and Anderson 2015, Abatzoglou and Williams 2016, Bowman et al. 2020, Hunter and Taylor 2022). In the 18th century, the introduction of European livestock and global trade to California brought seeds of invasive grasses and herbs that permanently displaced native plant species in many parts of the state (Reiner 2007, Bell et al. 2009). These Mediterranean species now dominate much of California’s rangelands, increasing fine fuel biomass and continuity, fire spread rate, and flame lengths (Reiner 2007, Bell et al. 2009, Sugihara et al. 2018, Ratcliff et al. 2022). Fire regimes maintained by Native Californians were disrupted or eliminated by colonization and genocide. By the 20th century, indigenous, natural, and agricultural burning were supplanted by aggressive fire suppression strategies (Stephens and Sugihara 2006, Sugihara et al. 2018). These strategies contributed to fuels accumulation and increased fire occurrence and severity throughout California (Steel et al. 2015). The situation has worsened with the spread of residential development into fire-prone wildlands, where human activity increases fire risk (Hammer et al. 2007, Kramer et al. 2018, Radeloff et al. 2018, Syphard et al. 2021, Li et al. 2022). This intermixing limits options available to reduce hazardous vegetation, notably constraining prescribed fire and herbicide application (Brown et al. 2018, McBride and Kent 2019).

Livestock grazing is increasingly identified as a fuel-management tool in the wildlife urban interface (WUI); however, research has historically focused on targeted fuel reduction by short term use of goats and sheep and the development and maintenance of fuel breaks by sheep and cattle (Taylor, Jr., 2006, Nader et al. 2007, Lovreglio et al. 2014, Bailey et al. 2019, Clark et al. 2023). These services are typically dependent on the availability of a contractor and can cost $927 to $2903 per ha, or $375 to $1175 per acre, providing short term, high intensity treatments that reduce brush and fine fuels (Ingram 2021, J. Walker, unpublished manuscript). Whereas this lends itself to removal of fuels on a small-scale to protect structures and inhabited areas, reducing overall wildfire risk is a challenge that extends over a much larger area (Syphard et al. 2011, 2021).

Cattle grazing is widespread in California, occurring over approximately one-third of the state (Huntsinger and Bartolome 2014, Brown et al. 2018, Saitone 2020). Whereas an estimated 1.8 million beef cattle grazed California’s rangelands in 2017, the impacts of existing cattle operations on fuels and wildfire in California have only recently begun to be assessed by the scientific community (Ratcliff et al. 2022, 2023, Siegel et al. 2022).

Cattle grazing manipulates fuel and influences other fire factors through the reduction of fine fuel loads and can cause long term alterations to vegetation structure and species composition (Ford and Hayes 2007, Nader et al. 2007, Ratcliff et al. 2022). Absent fire, dense shrub cover can develop on grasslands, increasing fire hazard and intensity and reducing options for fuel management (Ford and Hayes 2007). Whereas cattle herbivory is less effective at brush reduction than grazing or browsing by sheep or goats (Taylor, Jr., 2006), the cessation of cattle grazing in combination with fire suppression can increase shrub encroachment in coastal California (McBride and Heady 1968, Russell and McBride 2003). Anecdotally, areas where grazing has been retained are described as preferred staging areas for wildland firefighting crews because of lowered fire hazard, a near absence of hazardous fuels, and vegetation structure similar to those of protective fuel breaks (M. Turbeville, unpublished manuscript).

The interactions between cattle grazing, climate patterns, and long term processes like vegetation cover change make it more challenging to assess annual cattle grazing as a fire hazard reduction tool compared to high intensity, short term fuel reduction treatments (Nader et al. 2007, Ratcliff et al. 2022, Siegel et al. 2022). Some small-scale research has quantified cattle-fuel interactions (Diamond et al. 2009, Leonard et al. 2010), but short term experimental approaches are inadequate for capturing the full scope of impacts that cattle grazing has on wildfire over time and across large areas (Ratcliff et al. 2022, Siegel et al. 2022).

Recent large-scale studies show cattle grazing to be effective in reducing fire risk and mitigating fire behavior (Rouet-Leduc et al. 2021, Ratcliff et al. 2022, Siegel et al. 2022). In 2017, cattle grazing removed approximately 3.5 billion kg of non-woody plant material, or 195 to 1143 kg per grazed ha, or 174 to 1020 lb per grazed acre, from California, enough to influence fire behavior and reduce fire hazard and flame lengths (Ratcliff et al. 2022). Cattle grazing in the North Bay and Central Coast of California reduced annual burn probability by 0.8–3.6 percentage points (Siegel et al. 2022). These findings suggest cattle are currently serving a fire hazard reduction role in many areas of California and have potential as an effective fuel-management strategy, even when livestock producers focus on cattle production rather than fuel-management objectives.

Understanding the current and potential role of cattle grazing for fuel reduction at temporal and spatial scales appropriate for management, planning, and decision making is essential for its possible adoption into regional and statewide fuel-management strategies (Syphard et al. 2021). We explored the current and potential roles of cattle grazing in fire risk management at a multi-county scale and within priority treatment areas in Napa and Sonoma counties. We used mixed-effects regression models to generate study-area-wide average annual burn probabilities under a suite of different grazing scenarios to answer: (1) What is the current influence of cattle grazing on burn probability in Napa and Sonoma Counties, California?; (2) How do average annual burn probabilities change under different intensities and spatial extents of cattle grazing?; and (3) How can cattle grazing be applied to most effectively reduce fire probability in the study area and protect priority landscapes?

METHODS

Study area

Napa and Sonoma County rangelands are composed of grasslands, shrublands, and woodlands in the Sierran Steppe and Coastal Chaparral ecoprovinces (Bailey 1995). During the Spanish-Mexican era (1769–1846) the most widespread land use was domestic livestock grazing, including horses, cattle, sheep, and goats, which continued into the 1860s (Cleland 1941). Much of these grazing lands were replaced by cropland agriculture, vineyards, and orchards over the next hundred years (Grossinger et al. 2008). Expanding residential and agricultural development now occurs within a widespread legacy of fuel buildup due to fire suppression, initiated and perpetrated by the Anglo-American occupation of California (Stephens et al. 2018).

The study area of 452,903 ha encompasses the rangelands of Napa and Sonoma counties, excluding the California Coastal Steppe-Mixed Forest-Redwood Forest Province because of its drastically different climate and fire regimes (Stephens et al. 2018, Siegel et al. 2022; Fig. 1). A mosaic of urban and grazing lands and a high risk of rangeland conversion make these counties good candidates for studying the role of cattle grazing in wildfire hazard reduction. Sonoma County alone lost 3445 ha of rangelands to residential conversion from 1984—2008, and an additional 3208 ha was converted to vineyards (Cameron et al. 2014). The region also provides adequate temporal and spatial distribution of wildfire and grazing data for econometric analysis (CAL FIRE 2022; Siegel et al. 2022). Because our focus was on cattle grazing, grazing used hereafter refers to cattle grazing unless otherwise specified.

Data

We compiled data at 46,864 points at 200 m x 200 m spacing throughout the rangelands of the study area (187,456 ha). This spacing was selected as a compromise needed to retain adequate points for analysis because of relative scarcity of wildfire perimeters while mitigating for unobserved spatial autocorrelation. Table 1 summarizes data sources and generated variables.

We obtained land cover classification data (NLCD) for 2001 and 2016 from the National Land Cover Database (LaMotte 2016, Dewitz 2019). These were used to limit analysis to points categorized as grassland or shrub/scrubland in 2001 that remained categorized as grassland, scrub/shrubland, or forest in 2016. This removed points that transitioned from rangeland to development or cropland, vineyards, and orchards during the study period of 2001–2017.

Long term grazing intensity data was provided by Siegel et al. (2022). It was collected via phone surveys of large private landowners with holdings over 202.3 ha, or 500 acres, with grass/shrublands throughout three major ecoregions of California. The dataset describes whether or not a property was grazed and the stocking rate aggregated to an annual level, i.e., animal unit years per grazed ha or AUY per grazed ha, estimated per year from 2001–2017 for each property. One standard animal unit month is 780 lb or 355 kg of dry matter forage, and one animal unit year is 9360 lb or 4246.63 kg of dry matter forage (Society for Range Management 2003). Animal units are the amount of dry matter forage a 1000 lb or 454 kg cow and calf are estimated to consume over a given period of time, and when calculated over an area, usually acre or hectare, are a common measure of stocking rate, or intensity of use. Data from the 22 properties that fell within the study area were included in our analyses. This amounted to 1789 sample points at 200 m x 200 m spacing.

We generated fire-history variables using fire perimeter data from 1998–2017 from the California Department of Forestry and Fire Protection’s Fire and Resource Assessment Program database, which contains measured and estimated fire perimeters from 1900 to the present (CAL FIRE 2022). For each year included in the analysis (2001–2017), this data was used to determine whether or not a point burned in that year (t) and whether or not it had burned in any of the previous three years (t-1, t-2, or t-3).

Climate variables were accessed from the TerraClimate global dataset using the climateR package and aggregated in R (Abatzoglou et al. 2018, Johnson 2022). TerraClimate provides monthly climate data at ~4 km spatial resolution. Because California’s plant phenology and fire seasons are both strongly driven by temporal distribution of climate variables, we aggregated them seasonally. Seasons were defined as winter, i.e., December to February; Spring, i.e., March to May; Summer, i.e., June to August; and Fall, i.e., September to November. For each year, variables were assembled for seasonal precipitation, mean maximum wind speed, and maximum and minimum temperatures.

We obtained percentage vegetation cover data using Google Earth Engine to access the cover dataset of the Rangeland Analysis Platform (Allred et al. 2021). This product consists of annual estimates of percent cover by functional group at 30 m x 30 m resolution.

Data from the United States Census Bureau and United States Geological Survey (USGS) were analyzed in ArcGIS Pro to generate temporally fixed spatial factors, i.e., distance from roads, elevation, slope, and solar radiation aspect index (SRAI), and population density estimates for 2000 (U.S. Geological Survey 2013, U.S. Census Bureau 2018). Roads included primary and secondary roads, as well as rural roads, trails, and private roads (U.S. Census Bureau 2018).

Three datasets were used to develop and test-grazing scenarios. The important farmlands product from the California Department of Conservation’s 2020 Farmland Mapping and Monitoring Program (FMMP) was used to determine the extent of grazing lands in the study area (California Department of Conservation FMMP 2020). Grazing lands are defined by the FMMP as “land where the existing vegetation is suited to the grazing of livestock.” CAL FIRE’s “reduce wildfire threats to communities” dataset assigns scores to priority landscapes on a scale of one to five, with one considered the lowest priority for risk reduction treatment and five the highest. These scores are assigned based on housing density and wildfire threat, and were used to develop the priority scenario and stratify scenario comparisons (CAL FIRE 2018). Finally, the USGS Watershed Boundary Dataset was used to provided moderately sized, geographically intuitive subdivisions of the dataset for use as hypothetical treatment units (U.S. Geological Survey 2018).

All datasets used in this analysis except for the grazing survey data are free and accessible to the public via Google Earth Engine, R, or as downloadable shapefiles from their respective sources (R Core Team 2023).

Models

The models developed in this paper are derived from the logistic mixed-effects burn-probability model specified in Siegel et al. (2022). Logistic mixed-effects models take advantage of repeated observations of the same locations to control for unobserved characteristics of sample points, enabling better detection of changes over time. Here, we use these models with causal inference methods to estimate a series of counterfactual predictions of how altering grazing, i.e., the driver of interest, in the observed data would have changed burn probability. The selection of covariates for causal inference differs from that of predictive models (Hernán et al. 2019, Arif and MacNeil 2022). Instead of selecting the fewest predictors to most successfully predict the outcome, expert knowledge of the structure of the system being modeled is used to select covariates required to answer how grazing will influence burn probability (Table A1).

In the context of this analysis, burn probability refers to the estimated probability of a location burning in each year from 2001–2017. These estimates are generated by altering the drivers within the system while controlling for other variables to investigate how these changes influence the response variable, i.e., whether or not a location burned for each year in the dataset. This definition of burn probability is widely used in wildfire studies that employ econometric techniques (Syphard et al. 2013, Starrs et al. 2018, Siegel et al. 2022). The average annual burn probability is the burn probability for each location for each year averaged over the study period.

We used two models to understand the impact of grazing on burn probability. First, we used the burn model to estimate how burn probability is influenced by grazing and other covariates. This was sufficient for answering questions about existing grazing conditions and burn probability in the study area. However, a second model, the shrub model, was needed to explore the full impact of grazing on burn probability under hypothetical grazing scenarios.

The shrub model estimates percent shrub cover in relation to grazing and other covariates. When examining burn probabilities under hypothetical grazing scenarios, we used the shrub model first to replace remotely sensed percent shrub cover data with estimated percent shrub cover values calculated with the relevant hypothetical grazing data. These adjusted datasets were then used with the burn model to estimate the influence of grazing on burn probability under the hypothetical grazing scenarios (Fig. 2). Whereas the model specified by Siegel et al. (2022) utilized pre-regression matching to reduce standardized mean differences between grazed and ungrazed properties, they found the resulting regression results did not differ meaningfully from those produced using the full, unmatched dataset. As a result, we chose to specify both models in this analysis using the full dataset, enhancing the ability of our simulations to account for heterogeneity across the landscape.

Burn model

To estimate the influence of grazing on burn probability, we fit a logistic regression model to the grazing survey dataset using Stata, dropping highly correlated explanatory variables (|Pearson’s correlation coefficient| ≥ 0.66, p-value < 0.05; StataCorp 2021, Siegel et al. 2022) (Table A2). Variables included were those expected to influence vegetation dynamics and burn probability. Table 1 provides variable names, sources and resolution information, whereas Table A1 includes additional information on variable selection.

(1)

The binary term “grazed” was included to capture unobserved differences between grazed and ungrazed properties that might influence burn probability but are not attributable to the impacts of livestock use alone as estimated by stocking rate (Siegel et al. 2022). Examples could be increased human presence, more developed infrastructure, and more water resources. u_i is the site-specific random effect and e_it is the error term for each point in each year. The interaction between AUYperGrazedAcre*ShrubCover allowed for different responses to grazing across varying amounts of shrub cover. Latitude and longitude were included and interacted to control for spatial autocorrelation (Schleicher et al. 2017, Siegel et al. 2022).

For both the burn model and shrub model, cluster-robust standard errors were calculated at the property level to control for unobserved effects of the different surveyed properties and pseudoreplication (Abadie et al. 2023). Table 2 lists the regression coefficients and cluster-robust standard errors for the burn model.

Shrub model

One mechanism through which cattle grazing can affect burn probability is by altering vegetation structure and composition over time, like shrub distribution and density. Dense shrub cover can increase fire hazard and has historically encroached into areas where fire and grazing have been removed from the landscape (McBride and Heady 1968, Russell and McBride 2003, Ford and Hayes 2007, Parker et al. 2016). To account for the influence of cattle grazing on fire probability via its effect on shrub cover, the remotely sensed shrub cover data had to be adjusted for each scenario to reflect the impact of grazing.

To estimate the relationship between grazing and percent shrub cover, we regressed percent shrub cover on grazing data, burn data, and a subset of topographic and climatic variables, dropping highly correlated explanatory variables (|Pearson’s correlation coefficient| ≥ 0.66, p-value < 0.05; Table A2). This was executed using the plm package in R using the regions of the study area for which we had existing grazing data to specify the model (Croissant and Millo 2008).

(2)

Table A3 lists the regression coefficients and cluster-robust standard errors for the shrub model.

The shrub model was used to estimate adjusted percent shrub cover values for hypothetical grazing scenarios, enabling us to account for the influence of grazing presence and grazing intensity on shrub composition under different treatment conditions.

Estimating the current influence of cattle grazing

We applied information from the grazing dataset and interviews with University of California Cooperative Extension professionals to assign grazing intensities to FMMP grazing lands in the study area (California Department of Conservation FMMP 2020), creating a baseline scenario, or estimate of current grazing in the study area (Fig. 3a).

Using the predict function in Stata, the log-odds of burn probability was estimated for each point for each year in the dataset using the burn model (StataCorp 2021). In R, the log-odds were converted to annual probabilities and averaged over the 17-year dataset, resulting in an estimated average annual burn probability for each point for the study period (R Core Team 2023; Fig. 3b). The process was repeated for a no-grazing scenario, in which burn probability was estimated when grazing intensity and extent was set to zero, with shrub cover adjusted to match using the shrub model.

Estimating average annual burn probability under different intensities and extents of cattle grazing

Grazing scenarios were generated using the shrub model and burn model for two additional patterns of grazing use in the study area.

Full extent of study area

Full-extent scenarios applied grazing to the entire study area. These scenarios were run for 0.02, 0.41, 0.74, and 1.05 AUY per grazed ha (0.01, 0.167, 0.3, and 0.424 AUY per grazed acre), with 0 already represented by the no-grazing scenario. 0.41 was chosen as a representative stocking rate for production grazing, or a standard rate for cattle operations in the region that expect to sell calves for a profit. The highest reported stocking rate in the survey dataset was 1.05 AUY per grazed ha, which served as an extreme high stocking rate scenario. A more moderate comparison was 0.74, with 0.02 included as an extremely low but still grazed comparison.

Priority treatment areas

Priority scenarios added grazing to areas identified as priority three or greater under CAL FIRE’s “Wildfire threats to communities” ranking system in addition to areas grazed under the baseline scenario (CAL FIRE 2018; Fig. 4).

Priority scenarios were generated for 0.41, 0.74, and 1.05 AUY per grazed ha, with non-overlapping grazing lands remaining at the baseline stocking rates. Probabilities in priority scenarios were assessed both as part of the greater study area and in comparison to non-priority areas.

Scenario efficiency and key treatment units

The change in overall burn probability for each full extent and priority scenario were assessed in relation to the baseline scenario. Return for each scenario was calculated for relative comparison of scenario efficiencies.

(3)

To demonstrate how these models might be used to decide where to apply cattle grazing as a treatment strategy, subwatersheds (12-digit hydrologic units) ranging from 4856 to 17,806 ha, or 12,000 to 44,000 acres, were used to delineate hypothetical treatment units (USGS 2018). Grazing efficiency, here defined as the decrease in a point’s average annual burn probability when going from no grazing to grazed at 0.41 AUY per grazed ha, and CAL FIRE priority landscape scores were averaged across sample points within each treatment unit (CAL FIRE 2018). The change from no grazing to 0.41 AUY per grazed ha was selected as representative of applying a stocking rate suitable for calf-cow cattle production to ungrazed areas. Grazing efficiency and priority score were cross-referenced to identify key treatment units for the use of cattle grazing to have the greatest impact on burn probability while also prioritizing community protection from fire.

RESULTS

Estimating the current extent, intensity, and influence of cattle grazing

We estimate that cattle grazing in the study area from 2001–2017 reduced average annual burn probability over the entire study area 45%, from 9.9% to 5.4% (Table 3; Fig. A4), when compared to a no-grazing scenario. Within lands where grazing is present, this equates to an average annual burn probability decrease of 88%, from 9.2% to 1.1%, directly attributable to grazing.

Estimating change in average annual burn probability under different intensities and extents of cattle grazing

Full extent of study area

Overall average annual burn probability decreased with increases in grazing intensity for full extent (Table 4, Fig. 5) and priority (Table 5; Fig. A5) scenarios. Increasing the intensity of grazing throughout the study area from no grazing to 0.41 AUY per grazed ha decreased the overall average annual burn probability 88%, from 9.9% to 1.2%, whereas increasing it from no grazing to 1.05 AUY per grazed ha decreased the overall average annual burn probability 98%, from 9.9% to 0.2% (Fig. 6a). Increasing grazing intensity from 0.41 to 1.05 AUY per grazed ha resulted in a decrease of 81%, from 1.2% to 0.23% (Fig. 6b). Applying even a nominal amount of grazing resulted in a decrease in overall average annual burn probability of 70%, from 9.9% to 3.0%, when comparing the 0.02 AUY per grazed ha full extent scenario to the no-grazing scenario. Sample points in the north-central part of the study area experienced the greatest decreases in average annual burn probability in relation to the application of grazing (Fig. 6).

Priority treatment areas

When comparing the baseline and no-grazing scenarios, the absence of grazing resulted in an average annual burn probability increase of 721%, from 1.1% to 9.4%, for low priority areas and 39%, from 7.8% to 10.8%, for high priority areas. Adding grazing to high priority areas, i.e., the priority scenarios, decreased their average annual burn probability in relation to the baseline scenario for all stocking rates (Fig. 7; Table A6). A decrease of 82%, from 7.6% to 1.4%, occurred in high priority areas when production level grazing, i.e., 0.41 AUY per grazed ha, was applied compared to the same areas under the baseline scenario. Increasing the intensity of grazing in high priority landscapes from 0.41 to 1.05 AUY per grazed ha decreased the average annual burn probability in those areas 29%, from 1.7% to 1.3%.

Scenario efficiency and key treatment units

Overall scenario efficiency was the greatest for the 1.05 AUY per grazed ha priority scenario, in which high priority areas grazed at 1.05 AUY per grazed ha were added to the baseline scenario (Fig. 8; Table A7). Priority scenario 0.73 was next, followed by full extent scenario 1.05, full extent scenario 0.74, then priority scenario 0.41.

In a more applied example, Fig. 9 shows the results of cross-referencing the grazing efficiencies and priority landscape scores to identify key treatment units for the use of cattle grazing. Seven units scored highly in both criteria, totaling 81,407 ha or 201,162 acres. These key units cluster in the center of the study area, where the majority of lands are designated wildland urban influence zone. Wildland urban influence zone is defined as wildland vegetation up to 2.4 km from WUI or intermix zones (CAL FIRE 2019).

Under baseline conditions, 37% of points within these seven key treatment units were assumed grazed. If grazing were extended to all points in these treatment units at the same stocking rate, i.e., 0.41 AUY per grazed ha, burn probability would decrease 74%, from 11.3% to 2.9%. These treatment units contained 14% of the sample points, but 26% of points with high priority landscape scores, i.e., three or greater.

DISCUSSION

Anthropogenic changes to California’s vegetation, climate, and landscapes have all contributed to the catastrophic wildfire threat the state faces today (Abatzoglou and Williams 2016). Although there are many fuel-treatment options, implementation across large areas, varied vegetation types, and changing terrain render large-scale treatment logistically impossible or economically infeasible, particularly when no product, e.g., timber, is harvested from the treatment (Thompson and Anderson 2015, Hunter and Taylor 2022). Here, we examined a possible synergy between cattle producers and communities in the WUI by quantifying the current and potential impacts of using cattle grazing as a fuel treatment in an urban-wildland study area.

We estimated the reduction in average annual burn probability attributable to cattle grazing for Napa and Sonoma Counties from 2001–2017. The decrease in average annual burn probability suggests cattle grazing created areas of lower burn probability on the landscape and reinforces recent work that demonstrates that cattle grazing plays an important role in reducing fire hazard (Rouet-Leduc et al. 2021, Ratcliff et al. 2022, Siegel et al. 2022).

Increasing the extent and intensity of grazing both independently and in combination reduced overall average annual burn probability, with the greatest extents and highest stocking rates garnering the lowest overall average burn probabilities (Table 6). Even at low intensities of grazing overall average annual burn probability decreased, with a drop of 6.92 percentage points (from 9.9% to 3.0%) from the no-grazing scenario to the application of 0.02 AUY per grazed ha over the full extent of the study area. A similar response was observed for Central Coast ecosystems, with shifts from no grazing to 0.12 AUY per grazed ha decreasing burn probability (Siegel et al. 2022). This phenomenon may not be solely due to livestock fuel manipulation, but instead may relate to inherent traits of grazing lands or to risk-reduction practices attributable to the management of grazing lands, like maintained infrastructure, well-distributed water resources, range improvement practices like the removal of invasive species or treatment of brush, and the benefits of having someone physically present, and frequently living, on the landscape and therefore able to address fires and potential fire hazards.

Scenarios in which grazing was extended to high priority landscapes resulted in decreases to average annual burn probability in high hazard, high population areas with relatively little additional acreage grazed. The most efficient scenarios assessed produced decreases of 26% and 27% respectively, 1.4% and 1.5% percentage points, in average annual burn probability from grazing high priority areas at stocking rates of 0.74 and 1.05 AUY per grazed ha. Whereas 41% of areas designated high priority landscapes for fuels treatment did not fall within grazing lands, a significant portion of these are adjacent to or share overlapping boundaries with them (CAL FIRE 2018). An extension of cattle grazing into priority areas should be considered by fuel managers seeking long term mitigation strategies, especially in combination with high intensity, short term treatments like targeted goat or sheep grazing (Taylor 2006).

Identifying key long term treatment areas is crucial for the adoption of cattle grazing into fuels management plans. As the installation and upkeep of fencing and infrastructure for cattle can be expensive, deciding where to invest in adopting cattle grazing as a fuel-management strategy requires careful consideration. Potential concerns like cattle grazing impacts on water quality, habitat, recreation, and other resources should be carefully considered, but contextualized with the reduction in burn probability provided by grazing and the impacts of a fire on the same resources. We used the findings of our study to exemplify how a land manager could determine where cattle grazing would be most effective in protecting communities by cross-referencing reduction in average annual burn probability due to grazing and priority landscape scores for a set of land units.

The number of cattle grazing in California has been declining for decades (Cameron et al. 2014, Brown et al. 2018, Saitone 2020, Swette and Lambin 2021, Shapero et al. 2022). Quantifying the benefits of grazing may be crucial to ensuring it remains extensive enough to maintain current fuel loads on rangelands, and to retain grazing as an option for treating additional areas in the future (Ratcliff et al. 2022). Ranching in California is an economically marginal activity, with 70% of ranchers making less than $10,000 profit annually (Wetzel et al. 2012). Recently, long term droughts and rising demand-based land prices have increased costs (Cameron et al. 2014, Macon et al. 2016). Whereas payment for fuel treatment by sheep and goats is a common practice, cattle ranchers usually pay grazing fees to lease federal or private lands, which they rely on to provide enough forage for their herds (Huntsinger and Bartolome 2014, Rimbey et al. 2015, Saitone 2020). The availability of these key government-controlled resources is also in decline, with authorization of grazing on federal lands decreasing by 36% in California from 2010–2015, to some extent due to the invasion of woody vegetation and policy-driven land use change (Oles et al. 2017). With little margin for error and highly volatile market prices for beef, diversification of income is a strategy employed by ranchers to maintain their operations (Brown et al. 2018, Saitone 2020). If the services cattle provide in protecting homes and precluding wildfire can be quantified, payment for fire hazard reduction or discounted grazing leases may provide additional revenue streams, continuing the availability of grazing to manage vegetation in California’s highly altered ecosystems (Germano et al. 2012, Barry and Huntsinger 2021, Buckley Biggs et al. 2021, Barry 2022).

We utilized econometric methods to estimate the cattle-fire relationship over time and space, which allowed for the incorporation of long term, large-scale trends in the regression analyses (Starrs et al. 2018, Siegel et al. 2022). For these analyses, data had to be available and consistent throughout the study period and exhibit adequate spatial and temporal variation. The grazing survey data was aggregated to an annual stocking rate at the property level, which eliminated the ability to account for seasonal and spatial variations in stocking rate within operations, both of which influence the impacts of grazing on vegetation use. The grazing scenarios and baseline estimate could similarly be improved if more temporally and spatially explicit data existed. A more complete fire dataset including smaller fires and data from additional rangeland properties could improve model performance.

The climate dataset chosen was a compromise between spatial and temporal resolution. Seasonal variation in climate variables has great influence on both vegetation and fire behavior. Terraclimate provided data for a suite of climate variables at monthly temporal resolution, but a spatial resolution of approximately 4.5 km (Abatzoglou et al. 2018). With the majority of the dataset at 30-m resolution, sampling at 200 m likely resulted in underestimation of standard error values for the regression coefficients. Increasing the distribution of sample points was not feasible because of the limited grazing and fire perimeter data. These factors may in part be responsible for the generally higher estimates for annual burn probabilities in comparison to analyses derived from other products and fire behavior models (Short et al. 2020, Siegel et al. 2022).

Whereas additional factors relevant to implementing cattle grazing should always be considered when making recommendations, e.g., fence locations, access to water, distance to roads, impacts to sensitive areas, management capacity, and alternative operation goals, this methodology provides the groundwork for providing decision support for two main objectives: determining the worth of services that cattle grazing provides to communities in terms of fire risk reduction, and assessing where, how much, and how intensely cattle grazing should be applied for fire risk reduction.

CONCLUSION

Grazing for fuel reduction has the potential to offer a low cost, landscape-level fuel treatment (Germano et al. 2012) and is already doing so on millions of acres of working landscapes of California. With fire hazard increasing as fuels go untreated, every tool available should be considered in addressing potential fire risk at scale, especially when many cattle operations already conveniently neighbor many communities in the WUI. With the advent of new data sources and methodologies, research into cattle grazing and fuel reduction is capturing the interest of the scientific community. What remains to be seen is how this translates to the formal adoption of cattle grazing into larger fuel-management strategies, and if it results in helping sustain grazing lands and grazing practitioners.

LITERATURE CITED

Abadie, A., S. Athey, G. W. Imbens, and J. M. Wooldridge. 2023. When should you adjust standard errors for clustering? Quarterly Journal of Economics 138:1-35. https://doi.org/10.1093/qje/qjac038

Abatzoglou, J. T., S. Z. Dobrowski, S. A. Parks, and K. C. Hegewisch. 2018. TerraClimate, a high-resolution global dataset of monthly climate and climatic water balance from 1958–2015. Scientific Data 5:170191. https://doi.org/10.1038/sdata.2017.191

Abatzoglou, J. T., and A. P. Williams. 2016. Impact of anthropogenic climate change on wildfire across western U.S. forests. Proceedings of the National Academy of Sciences 113:11770-11775. https://doi.org/10.1073/pnas.1607171113

Allred, B. W., B. T. Bestelmeyer, C. S. Boyd, C. Brown, K. W. Davies, M. C. Duniway, L. M. Ellsworth, T. A. Erickson, S. D. Fuhlendorf, T. V. Griffiths, et al. 2021. Improving Landsat predictions of rangeland fractional cover with multitask learning and uncertainty. Methods in Ecology and Evolution 12:841-849. https://doi.org/10.1111/2041-210X.13564

Arif, S., and M. A. MacNeil. 2022. Predictive models aren’t for causal inference. Ecology Letters 25:1741-1745. https://doi.org/10.1111/ele.14033

Bailey, R. G. 1995. Description of the ecoregions of the United States. U.S. Department of Agriculture, Forest Service, Washington, D.C., USA.

Bailey, D. W., J. C. Mosley, R. E. Estell, A. F. Cibils, M. Horney, J. R. Hendrickson, J. W. Walker, K. L. Launchbaugh, and E. A. Burritt. 2019. Synthesis paper: targeted livestock grazing: prescription for healthy rangelands. Rangeland Ecology & Management 72:865-877. https://doi.org/10.1016/j.rama.2019.06.003

Barry, S., and L. Huntsinger. 2021. Rangeland land-sharing, livestock grazing’s role in the conservation of imperiled species. Sustainability 13:4466. https://doi.org/10.3390/su13084466

Barry, S. J. 2022. Sustain livestock ranching to sustain habitat: land sharing at risk on San Francisco Bay Area exacted conservation easements. Ecology and Society 27:33. https://doi.org/10.5751/ES-13459-270333

Bell, C. E., J. M. Ditomaso, and M. L. Brooks. 2009. Invasive plants and wildfires in Southern California. Publication 8397, Division of Agriculture and Natural Resources, University of California, Davis, California, USA. https://doi.org/10.3733/ucanr.8397

Brown, E., J. Laird, and K. Pimlott. 2018. California’s forest and rangelands: 2017 assessment. State of California Department of Forestry and Fire Protection, Sacramento, California, USA.

Bowman, D. M. J. S., C. A. Kolden, J. T. Abatzoglou, F. H. Johnston, G. R. van der Werf, and M. Flannigan. 2020. Vegetation fires in the Anthropocene. Nature Reviews Earth & Environment 1:500-515. https://doi.org/10.1038/s43017-020-0085-3

Buckley Biggs, N., J. Hafner, F. Mashiri, L. Huntsinger, and E. Lambin. 2021. Payments for ecosystem services within the hybrid governance model: evaluating policy alignment and complementarity on California rangelands. Ecology and Society 26:19. https://doi.org/10.5751/ES-12254-260119

California Department of Forestry and Fire Protection (CAL FIRE). 2018. Priority landscapes. CAL FIRE, Sacramento, California, USA. https://data.ca.gov/dataset/priority-landscapes

California Department of Forestry and Fire Protection (CAL FIRE). 2019. Wildland urban interface. CAL FIRE, Sacramento, California, USA. https://www.fire.ca.gov/what-we-do/fire-resource-assessment-program/gis-mapping-and-data-analytics

California Department of Forestry and Fire Protection (CAL FIRE). 2022. Fire perimeter data. CAL FIRE, Sacramento, California, USA. https://www.fire.ca.gov/what-we-do/fire-resource-assessment-program/fire-perimeters

California Department of Conservation Farmland Mapping and Monitoring Program (FMMP). 2020. California important farmlands. California Department of Conservation FMMP, Sacramento, California, USA. https://gis.conservation.ca.gov/portal/home/item.html?id=22da298849d147679551680593b9b035

Cameron, D. R., J. Marty, and R. F. Holland. 2014. Whither the rangeland? Protection and conversion in California’s rangeland ecosystems. PLoS ONE 9:e103468. https://doi.org/10.1371/journal.pone.0103468

Clark, P. E., B. A. Porter, M. Pellant, K. Dyer, and T. P. Norton. 2023. Evaluating the efficacy of targeted cattle grazing for fuel break creation and maintenance. Rangeland Ecology & Management 89:69-86. https://doi.org/10.1016/j.rama.2023.02.005

Cleland, R. G. 1941. The cattle on a thousand hills: Southern California, 1850–1870. Huntington Library, San Marino, California, USA.

Croissant, Y., and G. Millo. 2008. Panel data econometrics in R: the plm package. Journal of Statistical Software 27:1-43. https://doi.org/10.18637/jss.v027.i02

Dewitz, J. 2019. National Land Cover Database (NLCD) 2016 products. Version 2.0, July 2020. U.S. Geological Survey, Reston, Virginia, USA.

Diamond, J. M., C. A. Call, and N. Devoe. 2009. Effects of targeted cattle grazing on fire behavior of cheatgrass-dominated rangeland in the northern Great Basin, USA. International Journal of Wildland Fire 18:944-950. https://doi.org/10.1071/WF08075

Ford, L. D., and G. F. Hayes. 2007. Northern coastal scrub and coastal prairie. Pages 180-207 in M. Barbour, T. Keeler-Wolf, and A. A. Schoenherr, editors. Terrestrial vegetation of California. University of California Press, Berkeley, California, USA. https://doi.org/10.1525/california/9780520249554.003.0007

Germano, D. J., G. B. Rathbun, and L. R. Saslaw. 2012. Effects of grazing and invasive grasses on desert vertebrates in California. Journal of Wildlife Management 76:670-682. https://doi.org/10.1002/jwmg.316

Grossinger, R., J. Collins, E. Beller, and S. Gardner. 2008. The historical ecology of Napa Valley: an introduction. San Francisco Estuary Institute, Richmond, California, USA. https://www.sfei.org/sites/default/files/biblio_files/HE_Napa_broadsheet_v11_lowres.pdf

Hammer, R. B., V. C. Radeloff, J. S. Fried, and S. I. Stewart. 2007. Wildland-urban interface housing growth during the 1990s in California, Oregon, and Washington. International Journal of Wildland Fire 16:255-265. https://doi.org/10.1071/WF05077

Hernán, M. A., J. Hsu, and B. Healy. 2019. A second chance to get causal inference right: a classification of data science tasks. CHANCE 32:42-49. https://doi.org/10.1080/09332480.2019.1579578

Hunter, M. E., and M. H. Taylor. 2022. The economic value of fuel treatments: a review of the recent literature for fuel treatment planning. Forests 13:2042. https://doi.org/10.3390/f13122042

Huntsinger, L., and J. W. Bartolome. 2014. Cows? In California? Rangelands and livestock in the Golden State. Rangelands 36:4-10. https://doi.org/10.2111/Rangelands-D-14-00019.1

Ingram, R. 2021. Targeted grazing survey report to Range Management Advisory Committee (RMAC). California Woolgrowers Association Wildfire and Grazing Ad Hoc Committee, California Board of Forestry and Fire Protection, Sacramento, California, USA. https://bof.fire.ca.gov/media/cu4f2cof/5-cwga-2021-targeted-grazing-survey-report-to-rmac_ada.pdf

Johnson, M. 2022. climateR. R package version 0.1.0. R Foundation for Statistical Computing, Vienna, Austria.

Kramer, H. A., M. H. Mockrin, P. M. Alexandre, S. I. Stewart, and V. C. Radeloff. 2018. Where wildfires destroy buildings in the U.S. relative to the wildland-urban interface and national fire outreach programs. International Journal of Wildland Fire 27:329-341. https://doi.org/10.1071/WF17135

LaMotte, A. E. 2016. National Land Cover Database 2001 (NLCD01). Page Data Series, U.S. Geological Survey, Reston, Virginia, USA. https://doi.org/10.3133/ds383

Leonard, S., J. Kirkpatrick, and J. Marsden-Smedley. 2010. Variation in the effects of vertebrate grazing on fire potential between grassland structural types. Journal of Applied Ecology 47:876-883. https://doi.org/10.1111/j.1365-2664.2010.01840.x

Li, S., V. Dao, M. Kumar, P. Nguyen, and T. Banerjee. 2022. Mapping the wildland-urban interface in California using remote sensing data. Scientific Reports 12:5789. https://doi.org/10.1038/s41598-022-09707-7

Lovreglio, R., O. Meddour-Sahar, and V. Leone. 2014. Goat grazing as a wildfire prevention tool: a basic review. iForest - Biogeosciences and Forestry 7:260. https://doi.org/10.3832/ifor1112-007

Macon, D. K., S. Barry, T. Becchetti, J. S. Davy, M. P. Doran, J. A. Finzel, H. George, J. M. Harper, L. Huntsinger, R. S. Ingram, et al. 2016. Coping with drought on California rangelands. Rangelands 38:222-228. https://doi.org/10.1016/j.rala.2016.06.005

McBride, J., and H. F. Heady. 1968. Invasion of grassland by Baccharis pilularis DC. Journal of Range Management 21:106-108. https://doi.org/10.2307/3896366

McBride, J. R., and J. Kent. 2019. The failure of planning to address the urban interface and intermix fire-hazard problems in the San Francisco Bay Area. International Journal of Wildland Fire 28:1-3. https://doi.org/10.1071/WF18107

Nader, G., Z. Henkin, E. Smith, R. Ingram, and N. Narvaez. 2007. Planned herbivory in the management of wildfire fuels: grazing is most effective at treating smaller diameter live fuels that can greatly impact the rate of spread of a fire along with the flame height. Rangelands 29:18-24. https://doi.org/10.2111/1551-501X(2007)29[18:PHITMO]2.0.CO;2

Oles, K. M., D. A. Weixelman, D. F. Lile, K. W. Tate, L. K. Snell, and L. M. Roche. 2017. Riparian meadow response to modern conservation grazing management. Environmental Management 60:383-395. https://doi.org/10.1007/s00267-017-0897-1

Parker, V. T., R. B. Pratt, and J. E. Keeley. 2016. Chaparral. Pages 479-508 in H. Mooney and E. Zavaleta, editors. Ecosystems of California. University of California Press, Oakland, California, USA.

R Core Team. 2023. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/

Radeloff, V. C., D. P. Helmers, H. A. Kramer, M. H. Mockrin, P. M. Alexandre, A. Bar-Massada, V. Butsic, T. J. Hawbaker, S. Martinuzzi, A. D. Syphard, and S. I. Stewart. 2018. Rapid growth of the U.S. wildland-urban interface raises wildfire risk. Proceedings of the National Academy of Sciences 115:3314-3319. https://doi.org/10.1073/pnas.1718850115

Ratcliff, F., S. Barry, D. Rao, R. Peterson, T. Becchetti, E. Kebreab, K. Motamed, M. Jung, and F. Mitloehner. 2023. Cattle grazing moderates greenhouse gas and particulate matter emissions from California grassland wildfires. Sustainability 15:13539. https://doi.org/10.3390/su151813539

Ratcliff, F., D. Rao, S. Barry, S. Dewees, L. Macaulay, R. Larsen, M. Shapero, R. Peterson, M. Moritz, and L. Forero. 2022. Cattle grazing reduces fuel and leads to more manageable fire behavior. California Agriculture 76:60-69. https://doi.org/10.3733/ca.2022a0011

Reiner, R. J. 2007. Fire in California grasslands. Pages 207-217 in M. Stromberg, editor. California grasslands: ecology and management. University of California Press, Oakland, California, USA. https://doi.org/10.1525/california/9780520252202.003.0018

Rimbey, N. R., J. A. Tanaka, and L. A. Torell. 2015. Economic considerations of livestock grazing on public lands in the United States of America. Animal Frontiers 5:32-35.

Rouet-Leduc, J., G. Pe’er, F. Moreira, A. Bonn, W. Helmer, S. A. A. Shahsavan Zadeh, A. Zizka, and F. van der Plas. 2021. Effects of large herbivores on fire regimes and wildfire mitigation. Journal of Applied Ecology 58:2690-2702. https://doi.org/10.1111/1365-2664.13972

Russell, W. H., and J. R. McBride. 2003. Landscape scale vegetation-type conversion and fire hazard in the San Francisco Bay Area open spaces. Landscape and Urban Planning 64:201-208. https://doi.org/10.1016/S0169-2046(02)00233-5

Saitone, T. L. 2020. Livestock and rangeland in California. Pages 207-224 in P. L. Martin, R. E. Goodhue, and B. D. Wright, editors. California agriculture: dimensions and issues, second edition. Giannini Foundation of Agricultural Economics, University of California, Davis, California, USA.

Schleicher, J., C. A. Peres, T. Amano, W. Llactayo, and N. Leader-Williams. 2017. Conservation performance of different conservation governance regimes in the Peruvian Amazon. Scientific Reports 7:11318. https://doi.org/10.1038/s41598-017-10736-w

Shapero, M., K. Siegel, J. A. Gallo, J. Brice, and V. Butsic. 2022. Land cover conversion and land use change combine to reduce grazing. Journal of Land Use Science 17:339-350. https://doi.org/10.1080/1747423X.2022.2086311

Short, K. C., M. A. Finney, K. C. Vogler, J. H. Scott, J. W. Gilbertson-Day, and I. C. Grenfell. 2020. Spatial datasets of probabilistic wildfire risk components for the United States (270m), second edition. Forest Service Research Data Archive, Fort Collins, Colorado, USA.

Siegel, K. J., L. Macaulay, M. Shapero, T. Becchetti, S. Larson, F. E. Mashiri, L. Waks, L. Larsen, and V. Butsic. 2022. Impacts of livestock grazing on the probability of burning in wildfires vary by region and vegetation type in California. Journal of Environmental Management 322:116092. https://doi.org/10.1016/j.jenvman.2022.116092

Society for Range Management. 2003. Glossary of terms used in range management, second printing. Society for Range Management, Denver, Colorado, USA.

Starrs, C. F., V. Butsic, C. Stephens, and W. Stewart. 2018. The impact of land ownership, firefighting, and reserve status on fire probability in California. Environmental Research Letters 13:034025. https://doi.org/10.1088/1748-9326/aaaad1

StataCorp. 2021. Stata Statistical Software, Release 17. College Station, Texas, USA.

Steel, Z. L., H. D. Safford, and J. H. Viers. 2015. The fire frequency-severity relationship and the legacy of fire suppression in California forests. Ecosphere 6:1-23. https://doi.org/10.1890/ES14-00224.1

Stephens, S. L., J. M. Kane, and J. D. Stuart. 2018. North coast bioregion. Pages 149-170 in J. W. Van Wagtendonk, editor. Fire in California’s ecosystems. Second edition. University of California Press, Berkeley, California, USA.

Stephens, S. L., and N. G. Sugihara. 2006. Fire management and policy since European settlement. Pages 431-443 in N. G. Sugihara, editor. Fire in California’s ecosystems. University of California Press, Berkeley, California, USA. https://doi.org/10.1525/california/9780520246058.003.0018

Sugihara, N. G., T. Keeler-Wolf, and M. G. Barbour. 2018. Introduction: fire and California vegetation. Pages 1-8 in J. W. V. Wagtendonk, N. G. Sugihara, S. L. Stephens, A. E. Thode, K. E. Shaffer, and J. A. Fites-Kaufman, editors. Fire in California’s ecosystems. University of California Press, Berkeley, California, USA.

Swette, B., and E. F. Lambin. 2021. Institutional changes drive land-use transitions on rangelands: the case of grazing on public lands in the American West. Global Environmental Change 66:102220. https://doi.org/10.1016/j.gloenvcha.2020.102220

Syphard, A. D., A. B. Massada, V. Butsic, and J. E. Keeley. 2013. Land-use planning and wildfire: development policies influence future probability of housing loss. PLoS ONE 8:e71708. https://doi.org/10.1371/journal.pone.0071708

Syphard, A. D., H. Rustigian-Romsos, and J. E. Keeley. 2021. Multiple-scale relationships between vegetation, the wildland-urban interface, and structure loss to wildfire in California. Fire 4:12. https://doi.org/10.3390/fire4010012

Syphard, A. D., R. M. Scheller, B. C. Ward, W. D. Spencer, and J. R. Strittholt. 2011. Simulating landscape-scale effects of fuels treatments in the Sierra Nevada, California, USA. International Journal of Wildland Fire 20:364-383. https://doi.org/10.1071/WF09125

Taylor, Jr., C. A. 2006. Targeted grazing to manage fire risk. Pages 107-112 in K. Launchbaugh and J. Walker, editors. Targeted grazing: a natural approach to vegetation management and landscape enhancement. American Sheep Industry Association, Centennial, Colorado, USA.

Thompson, M., and N. Anderson. 2015. Modeling fuel treatment impacts on fire suppression cost savings: a review. California Agriculture 69:164-170. https://doi.org/10.3733/ca.v069n03p164

U.S. Census Bureau. 2018. 2018 TIGER/Line geodatabases: roads national geodatabase. U.S. Census Bureau, Washington D. C., USA. https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-geodatabase-file.2018.html

U.S. Geological Survey (USGS). 2013. LANDFIRE: national elevation dataset. USGS, Reston, Virginia, USA.

U.S. Geological Survey (USGS). 2018. Watershed boundary dataset. USGS, Reston, Virginia, USA. https://www.usgs.gov/national-hydrography/access-national-hydrography-products

Wetzel, W., I. Lacher, D. Swezey, S. Moffitt, and D. Manning. 2012. Analysis reveals potential rangeland impacts if Williamson Act eliminated. California Agriculture 66:131-136. https://doi.org/10.3733/ca.v066n04p131