Climate change must be factored into savanna carbon- management projects to avoid maladaptation: the case of worsening air pollution in western Top End of the Northern Territory, Australia

Introduction

An urgent global challenge is arresting anthropogenic climate change. This demands decarbonisation of energy and industrial processes and careful management of terrestrial carbon stocks (Keenan and Williams 2018). Landscape fires are a particularly important component in carbon management because of their potential to either increase or decrease greenhouse gas (GHG) emissions (Bowman et al. 2023). A prime example of this concerns tropical savannas, a highly fire-prone vegetation type that accounts for approximately 65% of the global carbon emissions from landscape fires (Van Der Werf et al. 2017). It has been argued that stewardship of savanna fire regimes globally presents an opportunity to reduce GHG emissions by reducing fire intensity and area burned (Lipsett-Moore et al. 2018). This proposal is based on the development of carbon-abatement projects in northern Australian tropical savannas (Edwards et al. 2021). These savanna carbon projects are now regulated under the Australian Government’s carbon offset scheme (Carbon Credits (Carbon Farming Initiative) Act 2011 (Cth)) and earn Australian carbon-credit units (ACCUs) under a statutory savanna-burning carbon-offset method. Each ACCU issued under the scheme is supposed to represent real and additional abatement of GHG emissions equivalent to 1 tonne of carbon dioxide equivalent (CO_2e). The ACCUs are a tradable financial instrument, with the primary source of demand now being large polluting facilities that can use the ACCUs to meet statutory obligations to keep their net emissions below prescribed emission limits.

In brief, the premise of the savanna carbon projects is that intentionally setting fires in the early dry season produces less GHG emissions because most fires are of lower intensity, consume less biomass per hectare, and are less extensive and more spatially patchy, reflecting higher fuel moisture when compared to late dry season wildfires that were prevalent before the projects began (Edwards et al. 2021). Operationally, the Australian approach rewards landowners and land managers who burn before 1 August, the date used to define the start of the late dry season (Australian Government 2018a, 2018b; Department of Climate Change 2023). The savanna carbon projects were first established in the Northern Territory in 2006 on Aboriginal land in western Arnhem Land and have been steadily expanded throughout this region (Fig. 1).

Fig. 1.

Savanna burning projects in the top part (latitudes north of 16°S) of the Northern Territory, Australia. (a) The geographic location of the Top End of the Northern Territory. (b) Spatial extent of the projects is shown. (c) The number of Australian carbon-credit units (ACCU) for year (since 2012), broken down by project, where a ACCU represents 1 tonne of carbon dioxide equivalent (CO_2e) emissions stored or avoided by a project. We selected only savanna-burning projects that had not been revoked, and were located within the top half of the Northern Territory (Australian Government Clean Energy Regulator 2023). Date of registration each project is shown in parentheses.

As of 1 October 2023, there were 79 registered savanna- burning projects across Australia. Since the ACCU scheme commenced, savanna burning projects have received 12.5 million ACCUs, just over 9% of total ACCU issuances, making them the fourth largest project type under the ACCU scheme by credit issuances (Australian Government Clean Energy Regulator 2023). The Northern Territory currently has 24 registered projects that cover ~48% of the northern part of the Northern Territory, above the latitude of 16°S (Fig. 1) (Australian Government Clean Energy Regulator 2023). These projects have received a total of 6.7 million ACCUs to October 2023, worth more than AU$200 million at prevailing market prices.

There is substantial public health concern that the tropical city of Darwin has the worst air quality of any Australian capital city because of wildfire smoke pollution (Jones et al. 2022; Tham et al. 2022). Particulate matter, especially the smaller size class particles of less than 2.5 μm in diameter (PM_2.5), in wildfire smoke are known to cause significant ill health and death, with no safe lower threshold, albeit air quality standards are set to limit exposures above 25 μg m⁻³, measured over a 24 h period (Vardoulakis et al. 2020). Greenhouse gases and particulate matter in wildfire smoke are positively correlated (Reisen et al. 2018), so the recent declining air quality in Darwin despite all the concurrent activity in the Top End to reduce carbon emissions, is counterintuitive. It would be expected that air quality in Darwin should be benefiting from the savanna burning schemes in the Top End of the Northern Territory (Fig. 1). Jones et al. (2022) investigated Darwin’s worsening air quality, finding a statistical association with area burned up to 500 km from Darwin in the early, but not late, dry season. These authors suggested that the worsening air quality in Darwin may be linked to the savanna carbon projects.

Here, we extend the work of Jones et al. (2022) to better investigate the seeming paradox of GHG savanna burning being correlated with worsening air quality in Darwin. Our approach is multipronged, using available data to provide a robust spatial–temporal context of Darwin air quality and changes in landscape burning. Specifically, we reconstruct PM_2.5 in Darwin air shed in the early, late, and whole dry season by using airport visibility records, building on the approach of Bowman et al. (2007a), and determine the correlation of Darwin PM_2.5 with region-wide air quality by using remote sensing-based estimates of air pollution (Van Donkelaar et al. 2021). We then evaluate trends in the area burned and the fire intensity in the early, late, and whole dry season, and model monthly changes in fire intensity and fuel moisture derived from gridded meteorological data. We discuss our findings in terms of their public health significance and the implications for effective savanna burning GHG management. We also briefly discuss the implications of our research for potential maladaptive (Schipper 2020, 2022) effects of landscape carbon abatement programs that are being widely promoted in savanna landscapes across northern Australia and elsewhere in the world. All analyses presented were done using the R statistical software (v.4.2.2, https://www.r-project.org/).

Methods

Results

Air pollution

The historically reconstructed monthly PM_2.5 data in Darwin are closely correlated with both the instrumental record (r = 0.96) and the aerosol optical depth (AOD) derived estimates (r = 0.72) for the entire dry season (Fig. 2a), as well as the early dry season (r = 0.77) (Fig. 2b) and late dry season (r = 0.71) (Fig. 2c). For the historical record, there is no trend for whole dry season since 1960, but since the commencement of strategic early dry-season fire management activities in 2006 and the registration of the first ACCU projects in 2012 (Fig. 2a–c), there has been a slight increase in PM_2.5 pollution in the early dry season and a corresponding decrease in late dry-season PM_2.5 pollution.

Fig. 2.

Trends in PM_2.5 in the Darwin airshed on the basis of instrumental, aerosol optical depth (AOD), and historical reconstruction for (a) whole dry season (May to October), (b) early dry season (May to July), and (c) late dry season (August to October). The trendline and 95th percentile confidence bounds (grey shading) are indicated, the commencement of strategic early dry-season fire-management activities in 2006 is indicated by a dashed red line, and the registration of the first ACCU projects in 2012 is indicated by a dashed green line.

Dry-season particulate pollution is correlated with the pollution in the surrounding hinterland areas (Fig. 3a). In the early dry season, there was the strongest positive correlation (P < 0.01) between the mean monthly values of AOD-derived PM_2.5 for Darwin and those for all 0.05° grid cells across the Top End (Fig. 3b). The distribution of grid cells with a significant correlation with Darwin’s PM_2.5 was more geographically restricted to the region immediately surrounding Darwin in the late dry season (Fig. 3c).

Fig. 3.

Statistically significant (P < 0.01) Pearson correlations between mean monthly PM_2.5 derived from aerosol optical depth (AOD) for Darwin and regional mean monthly AOD-derived PM_2.5 for 0.05° × 0.05° grid cells for (a) whole dry season, (b) early dry season, and (c) late dry season.

Area burned and fire intensity

Within the available MODIS satellite record, the savanna-burning projects have reduced the the area burned in the dry season (Fig. 4a), with the clearest decline in late dry season and a corresponding, albeit smaller, increase in the early dry season (Fig. 4b, c). Notably, these trends are not apparent in areas that are outside the savanna carbon projects, which were subject to less burning both prior to and after the commencement of strategic early dry-season fire management activities (Fig. 4). Nonetheless, the mean fire intensity (FRP ha⁻¹) was higher within carbon burning projects for the whole dry season, and the early and late dry seasons (Fig. 5a–c). Generalised linear modelling confirmed these differences, showing comparable fire intensity before the assumed establishment of the ACCU projects in 2012, with a marked increase thereafter, particularly on the project areas (Fig. 5d). Fig. 6 shows how fire intensity increases by month, and how the intensity curves are higher post-establishment of the ACCU projects compared to pre-establishment, with no differences between project and non-project areas.

Fig. 4.

Trends in percentage area burned within and outside carbon burning project areas in the Top End of the Northern Territory (10.5 million ha vs 12.6 million ha respectively). Burned area (ha) for (a) whole of dry-season, (b) burned area in early dry-season, and (c) burned area in late dry-season year, for project and non-project areas. The commencement of strategic early dry-season fire management activities in 2006 is indicated by a dashed red line, the registration of the first ACCU projects in 2012 is indicated by a dashed green line, and grey shading denotes the 95% confidence interval of the trend lines.

Fig. 5.

Yearly trends in mean fire intensity (FRP ha⁻¹) in the Top End of the Northern Territory. Fire intensity (a) in whole of dry-season, (b) in early dry-season, (c) in late dry-season year for project and non-project areas, where the commencement of strategic early dry-season fire-management activities in 2006 is indicated by a dashed red line and grey shading denotes the 95% confidence interval of the trend lines. (d) Mean (and ±standard error) monthly fire intensity (FRP ha⁻¹) for pre- and post establishment of ACCU projects (2012) between project and non-project areas on the basis of generalised linear model: fire intensity ~ pre/post × project area/non-project area + month with family = Gaussian(link = ‘log’). (e) Mean (and ± standard error) monthly fuel moisture index (FMI) for pre- and post-registration of ACCU projects (2012) between project and non-project areas on the basis of generalised linear models: FMI ~ pre/post × project area/non-project area + month.

Fig. 6.

Monthly trends in fire intensity (FRP ha⁻¹) and fuel moisture index (FMI) during the dry season (May to October) in the Top End of the Northern Territory. Fire intensity (a) before and (b) after 2012; fuel moisture index (c) before and (d) after 2012.

The divergence in the fire intensity on savanna carbon-project and non-project areas is apparent in Fig. 6a, which shows no difference prior to the registration of the ACCU projects (pragmatically assumed to be 2012) and a marked increase after their registration, with highest intensities on project areas. The increase in fire intensity means that fire intensities in August after the ACCU projects were first registered in 2012 were equivalent to the fire intensities in October before projects were first registered in 2012 (Fig. 6a, b). These monthly patterns are apparent, albeit in the opposite direction, for estimated fuel moisture (Fig. 6c, d), with a statistically significant inverse correlation between monthly fire intensity and monthly fuel moisture before and after the first registration of the ACCU projects in 2012 for savanna carbon project areas (−0.69 and −0.77 respectively) and for non-project areas (−0.55 and −0.64 respectively).

Discussion

This research was primarily motivated by a desire to understand the worsening air quality in Darwin, particularly in the early dry season, over the past two decades, coincident with the landscape-wide prescribed burning in the early dry season designed to reduce GHG emissions that commenced in 2006. Poor air quality in Darwin has manifest adverse impacts on human health demonstrated by numerous studies (Johnston et al. 2002, 2007; Hanigan et al. 2008; Campbell et al. 2022).

The savanna-burning carbon projects in the Top End of the Northern Territory have shifted burning to the early dry season, with a marked reduction in area burned in the late dry season (Fig. 4). However, this change is not evident on areas that are not managed for carbon abatement (Fig. 4). While we have shown that the proportion of the area burned in the late dry season has declined since the early 2000s, it is not clear that this period is representative of pre-project late dry-season fire activity. According to the NOAA Fire History Mapping dataset (Landgate 2023), used by the Australian Government to estimate savanna fire emissions in the national greenhouse accounts (Bowman et al. 2023), the beginning of the 21st century was characterised by high late dry-season fire activity in Northern Territory savannas that was significantly greater than the 1990 average. Such variability is likely to affect assumptions about the effectiveness of savanna burning projects in reducing dry-season fire extent, including our finding of a decline in area burned since the early 2000s (Fig. 4). Analysis of longer fire history mapping would help evaluate whether savanna burning projects have reduced dry-season fire extent and potentially explain the negligible change in the smoke pollution in Darwin since 1960 (Fig. 2).

A previous study by Jones et al. (2022) found that early dry-season burning in a 500 km radius from Darwin affected air pollution in Darwin between 2004 and 2019, and suggested a link with savanna-burning projects. Our study has extended the research of Jones et al. (2022) by providing a historical context to understand trends in particulate pollution in Darwin in the early, late, and whole dry season since 1960. We expected that air quality in Darwin would have improved following the implementation of the savanna-burning projects. However, we found that Darwin PM pollution in the early, but not late, dry season has worsened since the implementation of the savanna-burning projects (Fig. 2). Our geospatial analysis confirmed that Darwin air pollution seasonal patterns are a regional rather than local phenomenon, by demonstrating a strong correlation between Darwin particulate pollution with estimates of air pollution across the Top End (latitudes north of 16°S) of the Northern Territory (Fig. 3). The cause for the strong correlation is most likely to relate to the consistent prevailing south-eastern winds in the early dry season that would drive smoke towards Darwin (Jones et al. 2022).

Our analysis also identified that areas managed for carbon abatement have higher fire intensity compared to those that are not managed for carbon. Interestingly, this difference was marginal and statistically non-significant prior to the first registration of the ACCU projects in 2012, but has subsequently increased in magnitude and has become significantly different after their registration (Fig. 5d, e). The cause of the difference following the registration of the ACCU projects in 2012 remains unclear, but may reflect different patterns of burning. For instance, because financial benefit is derived from early dry-season fires, it is possible that the fixed date to define the end of the early dry season incentivises managers to burn as late as possible to increase the extent of ‘early’ dry-season fires and thereby reduce the risk of fires occurring in the late dry season. Frequent burning is known to increase the abundance of flammable grasses in Northern Territory savannas (Bowman et al. 2007b, 2014), creating fuel beds that burn intensely (Prior et al. 2017). Hence, repeated burning associated with carbon projects may have changed vegetation characteristics and fuel load, leading to more intense fires, including initiating a grass-fire cycle (D’Antonio and Vitousek 1992).

Our research also found a credible linkage between fire intensity and fuel dryness that can also contribute to worsening air quality in the early dry season (Fig. 6). There is a global pattern of declining fuel moisture caused by hotter and drier conditions associated with anthropogenic climate change (Ellis et al. 2022). Drier fuels are causing larger and more intense fires nationally and globally (Canadell et al. 2021; Ellis et al. 2022). A trend toward drier fuels highlights the vulnerability to climate change of fire management interventions designed to reduce GHG pollution (Bowman et al. 2023). The current savanna burning carbon method assumes fixed emissions per unit area, rigidly defined by early or late dry-season periods on the basis of historical observational data from different vegetation types (Meyer et al. 2012; Australian Government 2018a, 2018b; Department of Climate Change 2023). Our analysis has shown that the assumption of an early and late dry-season dichotomy in emissions owing to differing fuel consumption is likely to be flawed, given the strong decline in dry-season monthly fuel moisture that influences fire intensity.

From the data produced in this study, it is not possible to reach definitive conclusions on whether, and the extent to which, the savanna burning projects in the Northern Territory have reduced GHG emissions. The study was intended to explore the causes of observed worsening air quality in Darwin, not to evaluate the efficacy of savanna burning projects in abating emissions. Nonetheless, the negligible overall trend in Darwin PM_2.5 pollution since the 1960s, albeit with a shift from late to early dry season, suggests that the emission reductions generated by the projects may be small and that projects may have been over-credited (i.e. the number of carbon credits may significantly exceed the actual emission reductions).

Further research is warranted to better understand the effects of changing dry-season fuel moisture on GHG emissions associated with changing fuel loads, flaming versus smouldering combustion and fuel consumption, particularly in coarse woody debris (Volkova et al. 2019). Such research should specifically test the finding that there is negligible difference in emissions factors between early and the late dry season in Australia’s savannas, which was reported by Meyer et al. (2012), and which is relied on in the savanna burning carbon method to estimate abatement (Australian Government 2018a, 2018b; Department of Climate Change 2023). Also required is an improved understanding of the proportion of available fuel that is burned at a landscape scale, as a function of fuel moisture (e.g. the patchiness of fires, the consumption of different fuel components and associated emissions) (Desservettaz et al. 2017). Such further research necessarily demands a mix of further field, laboratory, and fine-resolution remote-sensing studies. In this context, it is important to note that current quantitative assessments of GHG benefits from savanna burning projects hinge on broadscale remote-sensing assessment with limited field validation (Edwards et al. 2021).

The current savanna carbon projects appear to be maladaptive to climate change, being rigid in their application and unable to accommodate natural variability in dry-season onset and length, as well as the effects of climate change on fuel moisture and fire intensity. The legislative focus on changing the season of burning is likely to have diverted attention away from determining whether GHG emission have been reduced by the savanna burning programs. Furthermore, the projects may have other maladaptive aspects such as causing unintended environmental and social adverse effects (Schipper 2020, 2022), including worsening early dry-season air pollution and human health impacts (Jones et al. 2022), loss of biodiversity (Corey et al. 2020) and cultural disruptions (Petty et al. 2015). To be more adaptive, the savanna burning projects should have greater flexibility in the definition of the end of the dry season, for example, by measuring changes in fuel moisture. An additional maladaptive aspect of the current savanna carbon projects is that they are not restoring traditional Aboriginal patch burning. For example, Bowman et al. (2022b) found that the commencement of strategic early dry-season fire management activities in central Arnhem Land in 2006 was associated with a population decline of the comparatively fire-sensitive savanna conifer Callitris intratropica. Arguably, a more adaptative approach to reducing GHG emissions from savannas would be to increase the amount of long unburned vegetation through careful patch burning more closely aligned with traditional Aboriginal practices, an approach that would most likely have benefits for biodiversity (Trauernicht et al. 2015).

More broadly, our study has highlighted the need to carefully evaluate the likely positive and negative social and environmental effects of applying extending savanna burning carbon abatement projects to savanna landscapes outside Australia (Lipsett-Moore et al. 2018; Laris 2021). Key issues that must be addressed include the feasibility of defining early and late dry season for burning (Laris et al. 2017), understanding how particulate and GHG emission factors vary among seasons and among fuel types (Laris et al. 2023), and integrating Indigenous knowledge and burning practices into project design (Laris et al. 2017). Proponents of savanna burning carbon abatement projects must demonstrate that there are viable ways of ensuring that GHG abatement is not over-credited and of mitigating their potentially maladaptive effects, including adverse ecological impacts and environmental harms such as smoke pollution that we have demonstrated in this study (Jones et al. 2022).

Conclusions

We have shown that the current approach to GHG mitigation using the early dry-season savanna burning method has caused increased smoke pollution in the Northern Territory capital Darwin, especially in the early dry season, when south-easterly winds transport smoke from savannas that surround Darwin. The cause of the worsening smoke pollution is most likely a combination of climate change making fuels drier in the early dry season and, hence, causing more intense fires, as well as possible management effects associated with the implementation of savanna burning projects. We suggest the ACCU scheme savanna burning carbon method is maladaptive, given it relies on a fixed definition of the end of the early dry season and an assumption that fires are less intense and polluting at this time. To be more effective, and truly adaptive to climate change, savanna carbon schemes must be flexible and responsive to environmental changes, particularly those caused by climate change such as fuel moisture and fire intensity.