Better biodiversity accounting is needed to prevent bioperversity and maximize co‐benefits from savanna burning

Strategies for mitigating climate change through altered land management practices can provide win–win outcomes for the environment and the economy. Emissions trading for greenhouse gas (GHG) abatement in Australia's remote, fire‐prone, and sparsely populated tropical savannas provides a financial incentive for intensive fire management that aims to reduce fire frequency, severity, and extent, and it supports important social, economic, and land management opportunities for remote communities, conservation agencies, and pastoralists. These programs now cover >20% of Australia's 1.9 million km2 tropical savanna biome, encompassing areas of globally significant biodiversity value. A common assertion is that by reducing the frequency, severity, and extent of fires for GHG abatement, these programs provide biodiversity co‐benefits. However, such biodiversity benefits have been assumed rather than demonstrated. Much better accounting of how biodiversity is responding to changed fire management is required to ensure that there are no unintended outcomes for biodiversity (bioperversity), and that biodiversity co‐benefits are maximized. Such accounting could underpin the earning of formal biodiversity credits from improved fire management, and will go a long way to understanding and improving the biodiversity outcomes of savanna fire management.


INTRODUCTION
The development of a carbon economy is driving land management changes that potentially provide win-win outcomes for environments and economies (Bradshaw et al., 2013). Tree plantings and avoided deforestation are two well-known an accountable activity under the Kyoto Protocol, has been identified as a major GHG abatement opportunity globally (Lipsett-Moore et al., 2018).
Australia is the only country to recognize savanna burning in its national emissions accounts, and savanna fires contribute 4% of Australia's annual GHG emissions (Cook & Meyer, 2009). The Australian Government's Emissions Reduction Fund (ERF) provides a carbon-crediting mechanism for engaging industry to offset GHG emissions and provides the financial basis for savanna fire management projects. Such projects represent important livelihood and land management opportunities, especially for remote Aboriginal communities, which own and manage much of Australia's tropical savanna (Cook, Jackson, & Williams, 2012;Russell-Smith et al., 2013), but also pastoralists (Skroblin, Legge, Webb, & Hunt, 2014) and conservation agencies (Russell-Smith, Evans, Edwards, & Simms, 2017).
Landscape fire management has been critical to the management of natural resources, and access to them, by Aboriginal people in northern Australia for millennia, and fire management continue to play a fundamental role in the connection to and stewardship of land by Aboriginal communities (Cook et al., 2012). However, European colonization resulted in the dispossession and movement of Aboriginal people off their traditional lands, and fire became largely unmanaged across the vast tropical savannas of northern Australia. As a result, in the second half of last century, these landscapes became dominated by frequent, large, and highintensity wildfires, mostly occurring in the late dry season (LDS; Yates, Edwards, & Russell-Smith, 2008; Figure 1). These fires threaten fire-sensitive habitats embedded in the savanna matrix, as well as a range of relatively fire-sensitive taxa (Table 1). A widespread aim of conservation management is therefore to re-establish strategic fire management in order to reduce the extent of higher intensity fires occurring late in the dry season (Andersen et al., 2005).
Prescribed fire management early in the dry season (EDS) is the basis of savanna burning for GHG abatement, and so it is often argued that GHG abatement is strongly aligned with biodiversity conservation (Russell-Smith et al., 2013, 2015. However, the notion that EDS burning benefits biodiversity is based on assumption rather than robust evidence (Bowman & Legge, 2016), and it has received little scrutiny. Such scrutiny is important, because elsewhere potential trade-offs between emissions offset schemes and biodiversity are explicitly recognized (e.g., Lindenmayer et al., 2012), and frameworks for addressing trade-offs developed (e.g., Phelps et al., 2012). Given that savanna burning projects are rewarded only for annual GHG abatement, the potential exists for perverse biodiversity outcomes (bioperversity) if there is a lack of congruence between the outcomes of fire management used to reduce GHG emissions and the requirements of fire-sensitive biota (Abreu et al., 2017; Andersen, F I G U R E 1 Australia's northern tropical savannas showing the frequency of (a) late dry season fires (wildfires), (b) all fires from 2000 to 2018, and (c) current savanna burning emissions abatement projects Note. Indigenous, Indigenous-owned and Indigenous-managed projects (including Indigenous Protected Areas); Government, Government conservation agency; non-Government organization, non-Government conservation organization; pastoral, pastoral properties. Fire history data were obtained from http://www.firenorth.org.au/nafi3/ and savanna burning project shapefiles were obtained from http://www.cleanener gyregulator.gov.au/ERF/project-and-contracts-registers/projectregister/project-mapping-files. Woinarski, & Parr, 2012;Archibald, 2011;Richards et al., 2012).
Here, we examine the asserted benefits of savanna burning projects for biodiversity in northern Australia, and the potential for bioperversity. Our aims are two-fold: (a) to provide a critical analysis of the assumption of biodiversity benefit and (b) to offer policy makers and fire managers recommendations for maximizing biodiversity co-benefits (or minimizing potential biodiversity detriment). We focus on the direct effects of savanna burning on biodiversity, rather than long-term, indirect amelioration of climate change that may accrue from managed reductions of emissions. Note. Superscript numbers indicate reference source (reference list is provided in Supporting Information). a Environment Protection and Biodiversity Conservation Act 1999, which provides a legal framework to protect and manage nationally and internationally important flora, fauna, and ecological communities in Australia.

SAVANNA BURNING IN NORTHERN AUSTRALIA
The key aims of savanna burning are to shift the timing of fire from LDS to EDS, and/or to reduce overall fire frequency and extent (Cook & Meyer, 2009;Russell-Smith et al., 2013;Yates et al., 2015). This is achieved by prescribed burning during the EDS, when relatively moist vegetation, low winds, and lower temperatures produce fires of lower intensity and size (Figures 2 and 3). Such fires result in lower GHG emissions by reducing biomass consumed within the fire footprint (including dead organic matter such as hollow logs and trees; Cook, Meyer, Muepu, & Liedloff, 2016), and by limiting the spread of high intensity fires ( Figure 3) that will inevitably occur later in the dry season.
The ERF is a legislated offset scheme that allows land managers to earn Australian Carbon Credit Units (ACCUs) by reducing GHG emissions. These ACCUs are a financial commodity representing 1 ton of CO 2 or carbon dioxide equivalent (CO 2 −e ) prevented from release into the atmosphere and can be sold to third parties wishing to offset their emissions. For the purposes of the ERF, fires occurring after August 1 (the mid-point of dry season, according to emission reduction criteria) are considered "late," whereas those before this are "early" (Figure 2). The first project to use this approach was the West Arnhem Land Fire Abatement project ( Figure 1c), which, under contract to a multinational energy corporation, has offset over 100,000 tons CO 2 −e per year since 2006 (Russell-Smith et al., 2013;Ansell & Evans, in press).

F I G U R E 2 Rainfall in northern Australia is highly seasonal
with >90 % of the average annual rainfall (dashed white line) falling between November/December and March/April, during the monsoonal wet season. This rainfall drives vegetation growth, including that of grasses. The intervening period is characterized by little to no rain, and grasses cure and become more flammable as the dry season progresses. Current savanna burning methodology directs that prescribed burning takes place during the early dry season (before August 1). The end of this dry period is characterized by hot temperatures and frequent "dry" storm activity (lightning with little rain) that are conducive for uncontrolled fire conditions until the wet season rains begin. Fire size and intensity is indicated by the size of the flame. Rainfall data were obtained from the Bureau of Meteorology and are long-term averages of major towns in northern Australia Across northern Australia, there are now more than 80 savanna burning projects, which include Indigenous, private, and Government-managed protected areas, as well as pastoral leases. Collectively they cover 380,000 km 2 (ca. 20%) of the savanna (Figures 1c and 4), and include some of Australia's most structurally intact and biodiverse landscapes .

POTENTIAL FOR BIOPERVERSIT Y FROM SAVANNA FIRE MANAGEMENT
Like other emissions offset programs, savanna fire management may provide collateral benefits to biodiversity, but alternatively may result in bioperversity if trade-offs between emissions reduction and biodiversity conservation are not considered (Lindenmayer et al., 2012;Phelps et al., 2012). Such trade-offs have received little attention (Abreu et al., 2017;Archibald, 2011;Richards et al., 2012). Given (a) the rapid up-take of savanna burning (Figure 4), (b) interest from some groups in adding biodiversity credits to fire projects (Fitzsimmons et al., 2012), and (c) the potential to apply savanna fire management for emissions control globally (Lipsett-Moore et al., 2018), it is important that potential trade-offs are considered, so that any negative impacts on biodiversity can be avoided.
The concern about biodiversity impacts of any management activity in Australia's tropical savannas is pertinent F I G U R E 3 Wildfires that occur in north Australia's late dry season are extensive (a) and generally leave very little to no unburnt vegetation within the fire scar (b). These wildfires are detrimental to savanna biodiversity and make relatively high contributions to greenhouse gas emissions. Fires that occur earlier in the dry season and in the late wet season are usually smaller (c) and leave pockets of unburnt vegetation, including an intact canopy (d) that act as refugia for savanna biodiversity. These earlier fires release less greenhouse gas (GHG) emissions than those that occur later in the year because the region is home to many threatened species currently undergoing marked declines (Table 1). Although the extent to which fire is implicated in these declines remains poorly resolved, current evidence suggests that biodiversity responses are more nuanced than the "early fire is good" versus "late fire is bad" dichotomy (Table 1). Below we identify a range of potential conflicts between savanna burning for GHG emissions abatement and biodiversity conservation (Table 2).

Fire timing
Savanna burning considers seasonality in terms of fire intensity and fuel consumption but does not consider impacts of timing of fire on biodiversity. The effects of fire on photosynthesis rates and export of carbohydrates to roots or lignotuber and thus plant growth, flowering, and seed production vary markedly with the timing of fire in savannas (Beringer et al., 2015;Prior, Eamus, & Bowman, 2004). Savanna fires have the greatest impacts on plants during the early dry season, T A B L E 2 Summary of potential negative outcomes for biodiversity (bioperversity) from savanna burning greenhouse gas emissions abatement programs with potential solutions because carbohydrate/nutrient reserves are depleted, and new leaves are sensitive to heat damage (Bowman & Prior, 2005). The effects of fire on plant physiology and phenology have implications for animals (Andersen et al., 1998) as fire consumes vegetation, and the earlier the fire, the longer it is that plant resources remain unavailable before replenishment in the wet season (Radford & Andersen, 2012). Additionally, some species including the threatened partridge pigeon (Geophaps smithii), which are sedentary and nest on the ground in the EDS, are at risk from fires occurring then (Fraser et al., 2003).

Limited evidence of biodiversity responses to fire
Most studies in Australian savannas linking faunal responses to fire regimes have design limitations (Andersen et al., 2012; Griffiths & Brook, 2014). For instance, the same variables driving fire mosaics (e.g., rockiness, soil type, local topography, and proximity to water) also influence species' occurrence and populations. Any association with fire regimes therefore cannot be assumed to be causal. There are few studies where fire can be confidently identified as a causal factor in population trends of threatened fauna (Williams, Woinarski, & Andersen, 2003), and these are often inconsistent. For instance, modeling of mark-recapture data from the Kapalga fire experiment in the Northern Territory (NT) indicated that an increase in fire frequency above once every 5 years increased extinction risk for several small mammals (Dasyurus hallucatus, Trichosurus vulpecula, Isoodon macrourus, and Melomys burtoni; Griffiths, Garnett, & Brook, 2015). In contrast, a fire experiment on Melville Island (elsewhere in the NT) indicated that fire frequency did not influence the abundance of T. vulpecula, and fire had no effect on mammal diversity (Davies et al., 2018). These differences presumably reflect geographic variation among study systems, and caution against generalizations about how small mammals respond to fire.

Older-aged vegetation as a liability
Australian savannas are characterized by frequent fire, mostly recurring every 1-4 years (Figures 1 and 5), and consequently only a small proportion of the landscape remains long-unburnt ( Figure 5). Long-term and large-scale exclusion of fire in Australian savannas is almost impossible (Russell-Smith et al., 2013). However, exclusion of fire for more modest periods (e.g., 3-10 years) is realistic in some areas and may result in notable improvements in quality of savanna habitats for threatened species (Woinarski & Legge, 2013). Intensive management around fire-sensitive habitats embedded in the dominant savanna matrix (e.g., rainforest patches; Table 1) may provide long-term protection of these distinctive nonsavanna environments from fire. However, in the dominant savanna communities longer unburnt areas are often viewed as risks in GHG-abatement programs, because fuel loads typically increase for 3-10 years postfire (Cook & Meyer, 2009;Murphy & Russell-Smith, 2010). Accordingly, some Note. Indigenous, Indigenous-owned and Indigenous-managed projects (including Indigenous Protected Areas); Government, Government conservation agency; non-Government organization, non-Government conservation organization; pastoral, pastoral properties. Data were obtained from the Australian Government's Clean Energy Regulator, available at http://www.cleanenergyregulator.gov.au/ERF/project-andcontracts-registers/project-register/project-mapping-files.
GHG-abatement programs deliberately burn such areas to reduce their extent. Such targeted burning is directly at odds with the widely recognized need to "increase" the area of longer unburnt savanna to conserve fire-sensitive biota (Table 1).

Focus on annual rather than long-term fire patterns
GHG abatement projects report on, and are funded according to, performance on annual fire patterns, rather than longer term fire histories. Therefore, taxa that require older aged vegetation or longer fire-free intervals (Table 1) (a and b), but no such shift in overall fire frequency (c and d) and consequently, no increase in the area that remains longer unburnt (e and f). Fire histories were generated using the project reporting tools available at http://www.ntinfonet.org.au/ infonet2/ and timing (Perry, Vanderduys, & Kutt, 2016). Evidence that repeated LDS fires are detrimental to obligate seeding plants and sedentary fauna with small home ranges (e.g., Yates et al., 2008) is not evidence that a high incidence of EDS fires is favorable. For fire sensitive taxa, fires (whether EDS or LDS) could still be harmful if they occur frequently or are too extensive (Table 1).
Increasing the area of unburnt country is widely seen as a key goal for conservation management (Andersen et al., 2005), but to date savanna burning projects have not achieved this because the economic imperative is to shift the season of burning within an annual fire cycle (Figure 5; see also Evans & Russell-Smith, 2019; Russell-Smith et al., 2015). This is not a directly negative consequence of savanna burning, but rather an example of how savanna burning could be suboptimal for biodiversity that requires longer fire-free intervals (see below).
The emissions calculation methodology was updated in 2018 to include carbon sequestered in dead organic matter in addition to avoided GHG emissions (Cook et al., 2016). This changes the focus to longer term fire management, as it places more emphasis on reducing fire frequency and increasing the proportion of the landscape with longer unburnt patches. This is a welcome advance that has potential benefits for biodiversity, such as greater survival of old hollow trees and logs. Nonetheless, it is important that such benefits be demonstrated rather than assumed. Moreover, there is no requirement for existing projects to switch to this new methodology.

Unrealistic distinction between "early" and "late" fires
The contrast between "low-intensity and patchy" EDS fires and "high-intensity and extensive" LDS fires is a generalization; all fire types can occur at any time during the dry season. Furthermore, the "hard" dry season midpoint (August 1) delineating EDS and LDS fires is insensitive to regional and inter-annual variation in the timing and extent of wet season rainfall (Perry et al., 2017). This is particularly the case in low rainfall areas where temporal rainfall patterns vary greatly and there is no relationship between weather patterns and GHG cutoff dates (Perry et al., 2017). Regional and inter-annual variation in rainfall affects grass growth, and hence fuel loads, and curing rates then determine when fires will take, their intensity, rate of spread, and therefore their extent. Consequently, burning to abate GHG emissions can adhere to the methodology yet fail to produce small, patchy, low-intensity fires required by fire-sensitive species (Table 1).

Conflict between market economy and biodiversity outcomes
Fire management (relying heavily on prescribed burning with ignitions dropped from aircraft) for biodiversity conservation is expensive (Russell-Smith et al., 2017). For example, if ignited fires only travel short distances before self-extinguishing, then multiple and repeated ignitions are required. The cheapest option within a market economy is to ignite fires within the regulated EDS period when fires will burn extensively with the fewest ignitions and minimized flight times. Thus, ACCUs can be maximized by extensive EDS burning, rather than creating many small, low-intensity fires that are favored by some taxa (Table 1).

Interactions between prescribed fire and other threatening processes
Fire can interact with other threatening processes in a way that is not considered by savanna burning. For instance, EDS burning can exacerbate grazing pressure by introduced herbivores (Table 1; see also Legge et al., 2019). If this is repeated, it can lead to decreases in landscape condition, loss of understory vegetation, and loss of structural cover for threatened taxa (Table 1), creating conditions favorable to invasive predators due to increased hunting success (Table 1). Failure to consider the interactions of fire with other threatening processes could amplify negative impacts of these threats (Table 1).

Sacrificial areas and fire breaks
A commonly used method in prescribed burning is to repeatedly burn topographic features such as water courses and valley systems to create barriers to protect other areas (e.g., Murphy, Cochrane, & Russell-Smith, 2015;Price, Edwards, & Russell-Smith, 2007). This approach may help to reduce the risk of LDS fires in nearby habitats, but may be detrimental to intrinsic biodiversity values, such as riparian vegetation and its distinctive fauna assemblages (Table 1). Furthermore, such repeated burning can exacerbate the effects of grazing and trampling by large herbivores.

Maximizing biodiversity co-benefits from savanna burning
Although savanna burning may lead to some beneficial biodiversity outcomes, it could be perceived as obviating the need for dedicated fire management for specific biodiversity benefit, and thus have suboptimal outcomes for species requiring a finer-scale burning approach (Table 1).
Monitoring of threatened small mammals in the Northern Kimberley (nearly entirely managed for GHG abatement, Figure 1c) shows a correlation between species richness and abundance, and longer unburnt vegetation (Figures 6a  and 6b). Here, savanna burning has increased EDS fires and reduced the area burnt by LDS fires (Figure 6c), but the total fire extent has slightly increased ( Figure 6c) and there has been no increase in the extent of longer unburnt vegetation (Figures 5e and 5f), suggesting that outcomes that most benefit biodiversity are not being achieved.
If prescribed burning interacts negatively with other threatening processes, this could also produce suboptimal outcomes. For instance, EDS fires can be detrimental if the resulting burnt areas are too large, because they expose small mammals to enhanced predation by feral cats (Table 1). Similarly, frequent EDS burning may lead to a grass-fire cycle of flammable native annual and exotic perennial grasses (Table 1). If frequent EDS burning promotes the spread of invasive grasses (e.g., Andropogon gayanus; Table 1), then org.au/nafi3/) was used to determine vegetation age at each plot, which is 0.25 ha and contains a standard set of 24 traps (various types) and is operated for 120 trap-nights per year. Average area burnt in two assessment periods: baseline (no fire management, [2000][2001][2002][2003][2004][2005][2006][2007][2008] and the last 10 years (fire-managed, 2009-2018) shows that fire management has increased the extent of early dry season fires and reduced the extent of late dry season fires but has not increased the extent of unburnt area (c). Bars are standard errors this would reduce the size of the project area, as areas dominated by invasive grasses cannot be included in savanna burning projects.

Traditional and contemporary approaches to savanna fire management
Although savanna burning (and EDS fire management generally) is intended to partly redress the loss of pre-European Aboriginal fire management (Russell-Smith et al., 2013), there are differences between the pre-European fire regimes and contemporary fire management for GHG abatement. Under current savanna burning methodology, there is an economic incentive and contractual obligations to apply fire in a particular way (Fache & Mozio, 2015;Perry et al., 2018), and financial penalties for fires that occur after August 1 across the entire savanna ( Figure 2). Traditional burning was not motivated by money, western conservation ethic, or a binary two-season approach (Petty, deKoninck, & Orlove, 2015). Rather it occurred throughout the entire year to fulfil cultural obligations, to facilitate passage through the landscape, and to attract game animals, and implementation varied from region to region (Preece, 2002). The application of fire was gradual as people moved around the landscape on foot, lighting many small fires as vegetation cured and became flammable. However, as the number of Aboriginal people living on and managing the landscape has markedly decreased, and most people reside in regional centers or larger towns, aerial prescribed burning has largely replaced walking, and fire is now applied to the landscape in a very short period of time ( Figure 2). Contemporary fires applied for GHG abatement therefore do not replicate past fire regimes (Petty et al., 2015).

THE NEED FOR BETTER BIODIVERSIT Y ACCOUNTING IN SAVANNA BURNING
We are not suggesting that savanna burning projects are having a detrimental impact on biodiversity. Rather, we are arguing that there is "potential" for this to be occurring, and that it is not possible to know if this is being realized without a better understanding of how the spatial and temporal arrangement of fires influences biodiversity, and interactions with other threatening processes (Driscoll et al., 2010;Parr & Andersen, 2006), or directly monitoring biodiversity outcomes. Such monitoring, however, is not a requirement of savanna burning projects. Prescribed burning is no different to any other management intervention; its effectiveness and impacts across a range of values need to be adequately reviewed (Legge, 2015).
Currently, monitoring that evaluates savanna burning is simple and inexpensive: freely available fire scar imagery is analyzed to determine the annual extent of EDS and LDS burning relative to a target based on a nominated level of improvement from the fire regime existing before the establishment of the program (e.g., Figures 5a, 5b, 6c, and 6d). Biodiversity monitoring is more complex because biodiversity responses may take years to be realized, will require on-ground biological surveys, and biodiversity may have varied and nuanced responses among different species, habitats, and regions. Assessment of the impacts of savanna burning on biodiversity will also have stronger inference if based on a Before-After-Control-Impact design.
Partnerships among organizations can assist in the design, implementation, and interpretation of biodiversity monitoring programs (e.g., Austin et al., 2017;Gillespie, Stevens, Mahney, Legge, & Low Choy, 2015). Such programs can be designed to report on trends in key biodiversity features (e.g., Table 1) or biodiversity surrogates or indicators (e.g., extent of longer unburnt savanna (Figures 5e and 5f) and firesensitive nonsavanna habitats). However, there is currently no ready funding sources for such monitoring. Given the increasing number of projects ( Figure 4) with multiple outcomes, those intending to incorporate biodiversity monitoring may need further support; or existing programs may need to be modified to ensure adequate provision for biodiversity monitoring.
It is important to recognize that there is no "one-sizefits-all" approach (Perry et al., 2016). Bioperversity may be avoided by incorporating the requirements of threatened and fire-sensitive taxa (Table 1) into the design and implementation of fire management programs. If potential trade-offs are identified, their broader implications can be transparently measured. For instance, if frequent EDS burning is good for abating GHG emissions and community livelihoods (e.g., financial rewards), but detrimental to threatened taxa (Table 1), then decisions about which interests hold precedence, or options for compromises, can be developed.

CONCLUSION
Savanna burning has proven to be a successful tool for incentivizing improved fire management in northern Australia, while providing important social, economic, and broader environmental benefits (Ansell & Evans, in press). It also has the potential to provide biodiversity benefits (Evans & Russell-Smith, 2019). However, this potential has not been demonstrated, and has received little scrutiny. As a result, proponents contracted to these schemes may inadvertently subvert biodiversity values, or at least fail to optimize biodiversity benefits. The management of fire for biodiversity conservation needs to explicitly consider the responses of taxa that may be disadvantaged by any or all fire regimes (Table 1). Here, we have identified a range of potential conflicts (and solutions; Table 2) between savanna burning for GHG abatement and biodiversity outcomes. If these are addressed during project planning stages, then biodiversity co-benefits can be maximized (Table 2).
Finally, proponents should not make claims of biodiversity co-benefits unless these are demonstrable, and where they are, proponents should be able to seek premium funding. For a market-based economy to affect change in GHG emissions while supporting biodiversity, a more flexible framework is required; and, given the typically slow rate of biodiversity recovery, market payments should relate to long-term performance. The rapid uptake of savanna burning (Figure 4) indicates that if similar incentives were introduced for fire management that maximizes biodiversity benefits, then transformational change could be achieved. A successful biodiversity accounting system will require funding for biodiversity monitoring that complements the accounting of GHG emissions, and also payments that respond in the shortand long-term to indication of biodiversity benefits.