The potential of using magnetic susceptibility to identify past wildfires in Australia

Introduction

As the world heads into a largely uncertain climate future driven by anthropogenic climate change, it has never been more important to fully understand the potential risks this poses. Wildfires in particular are extremely hazardous to human health, not just through the immediate impacts of the fires themselves, but also owing to the ongoing health effects caused by smoke inhalation. It has been estimated that for the period 2021–2030, healthcare costs in Australia related to conditions caused by smoke from wildfires will amount to AUD 110 million and may cause as many as 2418 deaths (Ademi et al. 2023). However, the data used to estimate wildfire frequency in Australia are woefully inadequate, as they rely on records of observed wildfires. The specific records only extend back to 1926 in New South Wales (Montoya 2014). Approximately a century of fairly stable climate gives a severely incomplete picture of climate and wildfire interactions, meaning Australia’s wildfire risk may not be fully understood. There are multiple sources that can be used to obtain longer climate records (e.g. Gergis et al. 2016) but getting longer fire records may be more difficult. Fossil charcoal records are traditionally used, but with its tendency for long-distance transport (MacDonald et al. 1991; Tinner et al. 1998; Woodward and Haines 2020) and differential dispersal depending on fuel source (Vachula and Rehn 2022), charcoal only goes part of the way to complete the wildfire risk picture. Charcoal analysis itself is also time consuming and destructive to the analysed sediments, as the charcoal needs to be isolated for identification, categorisation and quantification (Patterson et al. 1987). Owing to the chemical interactions in soil caused by the heat from fires, magnetic susceptibility may help complete the picture of wildfire frequency in Australia. The full calculation for future wildfire risk requires measures of, amongst other variables, wind speed, soil moisture and curing value (i.e. proportion of dead material; Sauvage et al. 2024), which are not easily obtained from palaeorecords. Therefore, providing information on Australian fire frequency and to which degree it is influenced by the prevailing climate is all that may be gleaned from extending the fire records. This will still be of great value, as it can, for instance, provide insight into the success of Indigenous fire management practices (Mariani et al. 2024).

Magnetic susceptibility (MS) is the measure of how much a material becomes magnetised when exposed to a magnetic field. Specifically, enhanced MS in soils has been found to be caused by iron oxides converting from haematite to magnetite or maghemite, which are much more strongly magnetic forms of iron oxide (Le Borgne 1955, 1960 in Tite 1972). For this enhancement to be possible, iron not only needs to be present in the soil (Tite 1972), but it must be present in forms that are readily convertible, such as haematite. If not, any peaks in MS may be due to in-wash of allochthonous materials rather than any in situ conversions (Rummery 1983). The conversion of haematite to maghemite takes place in stages, the first stage being the conversion of haematite to magnetite through reduction. Magnetite can then be re-oxidised to maghemite by exposing it to air. The initiation of this process may be caused by a number of factors, for example through simply wetting and then drying of the soil. Here, haematite is reduced to magnetite when organic material is broken down anaerobically during a wet phase. Magnetite is then re-oxidated to maghemite when the soils dry out and become exposed to air in a subsequent dry phase (Le Borgne 1955, 1960 in Tite 1972; Fitzpatrick et al. 2014). The enhancement of MS in soils may thereby be completely uneventful, simply being caused by average climate cycles, without the need for extreme droughts or flooding. It also follows that a soil rich in convertible forms of iron in an area with wet and dry periods has its MS gradually enhanced over time (Singer et al. 1996). In wetter regions, it is possible for magnetotactic bacteria to form biogenic magnetite (Fassbinder et al. 1990; Maher 1998), or the ferrimagnetic iron sulfide greigite (Stanjek et al. 1994), thereby enhancing the MS of the soil or sediment.

A more sudden and significant increase in MS is caused by wildfires as the consumption of oxygen by the fires leads to a reducing atmosphere in the top layers of the soil, which causes the reduction of haematite to magnetite. The subsequent re-oxidation to maghemite then occurs when the soils cool down after the fire and are re-exposed to air (Fig. 1). Wildfires have been found to be a particularly effective way of enhancing the MS of soils, with values up to 2–3 orders of magnitude greater in burned versus unburned soils (Rummery et al. 1979). The degree of enhancement is temperature-dependent, and MS can therefore be used as a proxy for soil burn severity, or how the fire affected the underlying soil. This may or may not be related to fire intensity, or the amount of energy released by the fire (Parsons et al. 2010). As the enhancement only affects the top section of the soil profile, possibly only the top 0–2 cm in some areas (Jordanova et al. 2019), this signature will be present in discrete bands in the soils. This makes it a potentially very valuable wildfire proxy that can be used to recreate the fire history of a site extending back thousands of years (Gedye et al. 2000).

Fig. 1.

Conceptual diagram of how soils become magnetically enhanced by fires.

MS has been used to examine and track the impact of wildfires all over the world for decades, being particularly popular in Europe (e.g. Rummery 1983; Gedye et al. 2000; Oldfield and Crowther 2007; Blundell et al. 2009; Jordanova et al. 2019) and the US (e.g. Millspaugh and Whitlock 1995; Clement et al. 2011), though in the latter, it is more commonly used as a proxy for erosion (e.g. Dunnette et al. 2014; Glover et al. 2020).

Gedye et al. (2000) used multiple measurements and quotients of MS, along with pollen and charcoal analyses, to create a fire record for Lago di Origlio in the Swiss Alps. They found a strong correlation between their proxies indicating several fire events over the past 4000 years, the top half of their record. Before this, MS does not correlate with the fire record and is generally low, which was interpreted as being as a result of erosion in the catchment prior to human settlement, leading to insufficient amounts of convertible forms of iron. However, even in the top part of the record, only half of the fire events identified by charcoal left a magnetic trace. The authors theorised that this could be due to the location of the fires, as only microscopic charcoal was used in the study, meaning it could be from outside the catchment, unlike the soils. It could also be that the magnetically enhanced soil was not deposited in the lake, or the fires may not have caused a sufficient degree of soil burn severity to leave a signal, or they could have been canopy fires, and therefore untraceable in the MS record. Interestingly, some fire events in this record were first identified using a specific MS quotient that is extra sensitive to magnetic grain size changes, then confirmed with more detailed pollen and charcoal analyses where the initial analyses showed no such event. This study shows the value in using multiple fire proxies to gain a complete fire record.

A valuable US study is that of Clement et al. (2011), who analysed a series of short cores from a recently burned area of the Everglades in Florida. They compared samples from completely burned and unburned sites, as well as sites that had been moderately affected by the fires. They found strong magnetic enhancement at sites that had been completely burned, no enhancement at unburned sites and a range of results from sites that had been burned to an intermediate degree. This highlights the potential to use MS as a proxy for soil burn severity in certain areas of the US, with it being more useful as a proxy for erosion in others (Glover et al. 2020). It also shows the importance of the soils being heated to a sufficient degree for there to be a definitive fire signal. This is a key limitation of using MS as a fire proxy, and it will be fully explored in later sections.

This review determines the potential of using MS as a proxy for wildfire in Australia by examining how it has been used in the past globally and where it is likely to be successful in Australia. Data from international studies examining the impact of various factors are presented and interpreted in the Australian context, owing to the lack of Australian studies pertaining to wildfire. The comprehensive soil sample study of England and Wales by Blundell et al. (2009) and that of Australia by Hu et al. (2020) are used repeatedly owing to their thorough investigations of potential factors influencing MS measurements.

Previous applications of magnetic susceptibility in fire contexts

Magnetic susceptibility has been used extensively as a proxy for all manner of events and variables, all of which are important to keep in mind, as these are the types of factors that will influence MS readings. When examining ancient rocks, studies of magnetic enhancement have been used to locate iron ore deposits (e.g. Chamalaun and Porath 1967; Tompkins and Cowan 2001), a major export for Australia.

When examining ancient rocks as well as more recent sediments, MS has been used as a proxy for the in-wash of inorganic materials (Kaiho et al. 2013; Püspöki et al. 2023). As stated previously, enhanced MS in soils naturally poor in convertible forms of iron is an indicator that there has been in-wash of soils either rich in the convertible forms or that already have enhanced MS, leading to its use as a proxy of in-wash from soil erosion (Oldfield 1988; Thompson and Oldfield 1986 in Maher and Taylor 1988). Some studies have considered fire-induced erosion events using MS, either through in-wash in basins (Millspaugh and Whitlock 1995; Long et al. 1998; Carcaillet et al. 2006) or blown to distant locations (Kletetschka and Banerjee 1995; Ravi et al. 2019). Others have been able to trace the source of in-washed sediments, link multiple sediment cores through their MS signals (Thompson et al. 1975; Bloemendal et al. 1988) or distinguish between industrial and natural sources of dust (Oldfield et al. 1985). The gradual enhancement of MS caused by repeated wet and dry phases, and the requirement that water be present for a gradual enhancement to occur, has led to some researchers using MS to reconstruct mean annual precipitation (Balsam et al. 2011; Alekseev et al. 2023). However, it has been found that the relationship between MS and rainfall fails below and above certain thresholds and these studies are therefore limited in scope (Hu et al. 2020).

The great enhancement in soil MS caused by fire has been used with some success in archaeological studies to identify the location of hearth fires within a site (Barbetti 1986; Marwick 2005; Lowe et al. 2016) as well as a variety of repeated anthropogenic fires on the same site, either in the shape of hearth fires or managed landscape burns (Bellomo 1993; de Sousa et al. 2023). Archaeological studies thereby highlight the importance of soil being heated to a sufficient degree for MS readings to be significantly enhanced. Through the use of MS, it has been found that wildfires have the capacity to destroy the archaeological heritage signals in hearth stones by re-heating these stones (Sanchez-Roda et al. 2022). For archaeological studies, it is important to separate wildfire MS signals from anthropogenic MS signals. One major difference between these types of fires is the concentrated nature of hearth fires, with repeated intense fires in a limited area burning for a long period of time, as opposed to wildfires, which burn over a larger area for a shorter period of time and less intensely. For hearth fires, therefore, MS will be significantly enhanced in a very small area, as compared with wildfires where MS is likely to be less enhanced over a wide area. This means that MS can be used to pinpoint where the hearths were located and, importantly, when the site was used for human occupation, as has been done successfully in northern Australia (Lowe et al. 2016). However, an extensive site magnetic survey may not be necessary to determine periods of intense human occupation, as these leave clear signals in the MS record, which can be seen in Fig. 2, with data from both Australia and overseas. The data used for this figure are from different levels of archaeological sites classified according to the number of artefacts discovered at each level. Most, but not all, of these levels were marked as containing active hearth fires; others were just marked according to the number and type of artefacts (de Sousa et al. 2023et al.) or simply marked as archaeological material in the ‘General’ category (Oldfield and Crowther 2007).

Fig. 2.

Magnetic susceptibility at various occupation levels in archaeological sites. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Oldfield and Crowther (2007), Lowe and Wallis (2020), de Sousa et al. (2023).

It is also possible to widen the scope and create a record of past wildfires for a specific location based on MS (e.g. Rummery et al. 1979; Gedye et al. 2000; Leys et al. 2016). By incorporating grain size analysis, researchers can even distinguish between allochthonous and autochthonous sediments, as soils directly impacted by fire tend to be finer-grained, with fire directly causing these grain size changes through the production of ferrimagnetic mineral assemblages to a greater extent than weathering and soil formation processes (Kletetschka and Banerjee 1995; Oldfield and Crowther 2007). This is one way to differentiate between local and regional wildfire events (Rummery 1983), but similar studies can also be performed by examining MS and charcoal records for the same sites. Simultaneous peaks in charcoal concentrations and MS indicate a local fire, but a peak in charcoal with no simultaneous peak in MS likely means there was a fire outside of the catchment, or there was a local fire that did not lead to a significant erosion event (Millspaugh and Whitlock 1995), or did not heat the soil enough to leave an identifiable signal.

MS has some benefits over charcoal as a wildfire proxy, the prime one being its ability to identify wildfires causing high soil burn severity, affecting ground surface characteristics such as hydrophobia, causing increased runoff (Parsons et al. 2010). These fires may be particularly destructive and dangerous and therefore of inherent interest. Canopy fires may lead to high charcoal concentrations in the sediments but only slight MS enhancements owing to the soils not reaching the high temperatures required for the strongest signals (Rummery et al. 1979), as compared with ground-level burns causing a high soil burn severity. These may leave similar charcoal concentrations in the sediments but lead to a much stronger MS enhancement. MS enhanced soils are also less likely to be transported great distances, as opposed to charcoal, particularly from canopy fires, as charcoal is light enough to be transported by wind as well as water and other common erosion mechanisms (Oldfield and Crowther 2007). Microscopic charcoal has been found to be able to travel up to 100 km (MacDonald et al. 1991; Tinner et al. 1998), and very large pieces of charcoal (≤5 cm long) have been found at least 50 km from their source (Woodward and Haines 2020). However, MS also has its drawbacks, for instance that it may be recording an erosion signal rather than a wildfire signal, with an influx of inorganic sediments leading to a peak in MS unrelated to any wildfires. Certain criteria also need to be fulfilled for there to be any MS enhancement at all. These include the presence of convertible iron (Rummery 1983), sufficient rainfall and high enough temperatures (Tite and Linington 1975) and the presence of the right amount of organic matter (Ketterings et al. 2000; Blundell et al. 2009). These are all explored further below.

However, as mentioned in the introduction, one of the most important factors is the heating of the soil, as temperatures need to reach at least 100–200°C for magnetic enhancement to occur, with the most significant enhancement occurring in soils that have been heated higher than 500°C (Gedye et al. 2000). As grass fires have less substantial fuel and therefore burn more quickly and at a lower temperature, they have been found to not produce a significant magnetic enhancement of soils, and only particularly intense fires with plenty of woody fuel are likely to produce a measurable effect (Roman et al. 2013). The importance of woody fuel may be linked to the importance of fire duration, with a study by Hartford and Frandsen (1992) suggesting that duration may be more important than maximum fire temperature as regards heating the soil. In this study of controlled burns in the US, the authors found that even a fire lasting more than 16 h only heated the soil to 400°C, not high enough for the most significant magnetic enhancement to occur. In fact, from Fig. 3, it appears as though only samples heated to over 600°C produce a significant effect. In this figure, there is a dip in MS values at 700–800°C, which is due to thermal transformation of magnetic minerals at these temperatures, leading to paramagnetism. At temperatures above this, MS is significantly enhanced again, owing to the pyrite in the loess samples used here changing to pyrrhotite as well as the goethite changing into haematite and magnetite (Zhao et al. 2022).

Fig. 3.

Magnetic susceptibility at various levels of fire. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Tite and Linington (1975), Longworth et al. (1979), Ketterings et al. (2000), Blake et al. (2006), Oldfield and Crowther (2007), Roman et al. (2013), Eldiabani et al. (2014), Jordanova et al. (2018, 2019), Zhao et al. (2022).

In the Australian context, Bradstock and Auld (1995) measured soil temperatures at 0–10 cm depth during controlled fires at two sites close to Sydney and did not record temperatures over 150°C, with most readings being below 100°C. As seen in Fig. 3, such temperatures may not be enough to cause a significant increase in MS. However, these experiments, along with the US experiments outlined above, were necessarily carried out under controlled conditions, in weather unlikely to lead to natural fires. Naturally ignited wildfires will burn hotter under different climatic conditions (Luke and MacArthur 1978) and may lead to more elevated soil temperatures. In addition, Bradstock and Auld (1995) used a limited study region, but many factors are likely to influence how effectively the soil is heated during a wildfire in different regions. These factors include moisture content (Hartford and Frandsen 1992), soil type and porosity as well as vegetation. In Fig. 3, some samples are simply labelled ‘burned’ or ‘unburned’. This is owing to these samples having been collected in the field after a fire event without any determination being made as to the severity of the fire. Even so, the burned samples still show a magnetic enhancement compared with the unburned samples, apart from the subsurface samples, which would not have been heated sufficiently. The broad ‘high severity’ category also shows a significant enhancement over the other burn categories. In fact, the samples classified as coming from areas burned at a ‘low’ or ‘moderate severity’ may be indistinguishable from unburned samples, showing the importance of sufficiently heated soil. In the studies used here, the fire severity categories are based on the terms laid out by Keeley (2009), where fire severity ‘refers to the loss of organic matter aboveground and belowground.’ (Keeley 2009, p. 4) The samples in the ‘Burned surface’ category have a wide spread of MS readings, with the lowest readings at, or very near, 0. This is likely due to some of these samples being obtained from soils that were burned at what was termed ‘low intensity’, and therefore not heated sufficiently. This spread in readings is similar to that reported by Clement et al. (2011) from their ‘moderately burned’ category. The samples represented in Fig. 3 come from around the world, with no Australian samples represented.

Some studies using MS as a proxy for wildfires have been carried out in Australia previously (e.g. Blake et al. 2006), but to date, there has been no wide-ranging exploratory study to determine its true potential to be used on this heavily fire-affected continent. In the next sections, the key site factors needed to use MS as a wildfire proxy are examined along with any regions of Australia that may fulfil these criteria. Being able to utilise an additional wildfire proxy in Australia would enable researchers to extend the current fire records as well as provide context as to the soil burn severity of these fires. This will improve our understanding of wildfire patterns and frequencies in Australia, thereby allowing us to better understand the emerging wildfire situation.

Key factors needed for a definitive fire signal

There are three main factors that influence the potential of a soil to produce an MS fire signal: an iron-bearing parent material (Fig. 4; Rummery 1983), a beneficial climate and sufficient soil heating. The climate should be warm and damp enough to ensure that the iron is present in the soil in a convertible form, such as haematite, through the action of hydrolysis and the release of iron from the parent material (Dearing et al. 1996, 2001; Blundell et al. 2009). Australia is rich in iron-bearing geology, haematite in particular, which is why Australia is the world’s leading producer of iron ore (Summerfield 2020), and why an iron-bearing parent material is unlikely to be a limiting factor for MS studies anywhere in the country. Despite this, parent material has been found to be the main factor controlling magnetic properties of Australian soils, owing to the different levels of weathering present (Hu et al. 2020). It is also of note that sand and sandstone, common parent materials in Australia, produce some of the lowest magnetic susceptibility readings in Fig. 4. This is likely due to their porosity causing magnetic minerals to be leached out of the soil (Hu et al. 2020). Importantly, on iron-rich soils, it is vital to be able to distinguish between normal background levels of magnetic iron oxides, and those formed through the conversion of iron-bearing minerals into their magnetically enhanced counterparts. The existence of such resources as an Australian soil magnetic database (Hu et al. 2020) is therefore of key importance when it comes to teasing out primary (background) and secondary (converted) magnetic enhancements of the soils.

Fig. 4.

Magnetic susceptibility for different parent rocks and minerals. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Scheffer et al. (1959), Hedley (1970), Prasad and Ghildyal (1975), Tite and Linington (1975), Mullins (1977), Longworth et al. (1979), Blundell et al. (2009), Hu et al. (2020), Shirzaditabar and Heck (2021).

It has been shown that a mean annual precipitation (MAP) of at least 500 mm is vital for the necessary conversions to take place, and that they occur at peak efficiency at ~1000–1500 mm year⁻¹ but decline above ~3000 mm year⁻¹ (Fig. 5; Singer and Fine 1989; Maher and Thompson 1995; Singer et al. 1996), though these figures vary by study (Maher 2011). Mean annual temperature (MAT) may be a minor factor, as it affects evapotranspiration and therefore effective moisture, or the proportion of MAP that remains in the soil (Han et al. 1996). A MAT under 6°C may also limit the enhancement of MS, as colder temperatures tend to slow down chemical reactions (Singer and Fine 1989). Temperature will not be a limiting factor in Australia, as only a few high-altitude areas have a MAT of less than 6°C. Precipitation, however, is likely to be a major limiting factor, with most of the continent having a MAP below 600 mm (Fig. 6), and effective moisture being much lower in the arid interior. It is important to note that the degree to which MAT affects MS readings is much lower than for MAP, as is evident in Fig. 5.

Fig. 5.

Magnetic susceptibility by (a) Mean Annual Precipitation and (b) Mean Annual Temperature. Red dots represent surface samples, orange subsurface and black unspecified. Data from: (a) Lukshin et al. (1968), Vadyunina and Babanin (1972), Vadyunina and Smirnov (1978), Oldfield et al. (1979), Maher (1984), Tite and Linington (1986), Alekseev et al. (1989), Maxted (1989), Singer and Fine (1989), Fassbinder (1991), Alekseev unpub. (1994 in Maher and Thompson 1995), Maher et al. (1994), Tipping and Peters (1995), Han et al. (1996), Blundell et al. (2009), Balsam et al. (2011), Alekseev et al. (2023), Maher unpub. (in Maher and Thompson 1995), (b) Han et al. (1996), Blundell et al. (2009), Alekseev et al. (2023).

Fig. 6.

Map showing mean annual precipitation for Australia: solid line representing 1000 mm year⁻¹, dashed line representing 600 mm year⁻¹ (Australian Government Bureau of Meteorology 2020) together with locations of sites described in the text. Black diamonds represent sites where magnetic susceptibility can or has been used to reconstruct fire events, red diamonds sites where magnetic susceptibility has been found not to be able to reconstruct fire events, blue diamonds sites where magnetic susceptibility responds to local signals or has unclear results. Sites numbered as follows: (1) Tiam Point (Rowe 2007), (2) Walala (Prebble et al. 2005), (3) East Trinity (Grogan et al. 2003), (4) Bromfield Swamp (Burrows et al. 2016), (5) Gledswood Shelter (Lowe et al. 2016), (6) Marillana (Marwick 2005), (7) Fern Gully Lagoon (Kemp et al. 2020, 2021), (8) Little Llangothlin Lagoon (Gale et al. 2004; Woodward et al. 2011, 2014), (9) Macquarie Marshes (Graves et al. 2019), (10) Myall Lake (Skilbeck et al. 2005), (11) Tuggerah Lake (Mooney and Maltby 2006), (12) Nattai River catchment (Blake et al. 2006), (13) Two Mile Lake (Itzstein-Davey 2004; Gouramanis et al. 2016), (14) Callemondah Billabongs (Tibby et al. 2003; Gell et al. 2005), (15) Tea Tree Swamp (Gell et al. 1993), (16) Caledonia Fen (Kershaw et al. 2007), (17) Bolin Billabong (Leahy et al. 2005), (18) Lake Keilambete (Mooney 1997), (19) Lake Johnston (Anker et al. 2001), (20) Macquarie Harbour (Augustinus et al. 2010), (21) Lake Osborne (Fletcher et al. 2014), (22) Paroo (Pearson et al. 2001), (23) Lake Euramoo (Haberle 2005), 24) Cape Barren Island (Adeleye et al. 2021).

In Fig. 7, we can see the importance of soil type on MS readings, as well as the importance of sampling depth, also evident in Fig. 5. The samples here are from around the world, with no Australian samples, and the magnetic enhancement is likely to be due to MAP and moisture content to a certain extent, with none of the samples categorised as recently burned. The low readings for gley and peat-based soil types (Fig. 7) are likely due to poor drainage leading to reducing conditions, thereby somewhat inhibiting MS enhancement (Table 1; Blundell et al. 2009).

Fig. 7.

Magnetic susceptibility by soil type. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Lukshin et al. (1968), Vadyunina and Babanin (1972), Vadyunina and Smirnov (1978), Alekseev et al. (1989), Alekseev unpub. (1994 in Maher and Thompson 1995), Maher et al. (1994), Dearing et al. (1996), Maher (1998), Singer and Fine (1989), Hannam and Dearing (2008), Blundell et al. (2009), Jordanova et al. (2016).

Multiple studies have shown organic matter content to be a limiting factor for the enhancement of MS in soils (e.g. Ketterings et al. 2000), with a minimum content of 2% found to be necessary (Vodyanitskiy 1982 in Singer and Fine 1989). However, Blundell et al. (2009) showed that this is only true up to a threshold level of ~6%, though Fig. 8 shows significant enhancement is possible through soils burning with an organic carbon content as high as 20%. Organic matter is made up of ~50% carbon, so that an organic carbon content of 20% is equivalent to an organic matter content of ~40%. It has been suggested that an organic matter content higher than 6% indicates an area of low drainage, low MAT and/or acidic conditions that prevent not only the efficient breakdown of organic matter but also the magnetic enhancement of soils (Blundell et al. 2009; Table 1). This is due to the reductive dissolution that takes place in the anaerobic conditions caused by prolonged waterlogging. This dissolution leads to significant reductions in magnetic susceptibility readings (Mullins 1977; Maher 1986; Thompson and Oldfield 1986; Dearing et al. 1995) and the effects may be noticeable on sub-decadal timescales, as demonstrated by field and laboratory studies (Hannam 1999; Dearing et al. 2001). It has therefore been suggested that a well-drained soil is of key importance for the enhancement of MS (Table 1; Blundell et al. 2009). However, a soil that is too well drained may be subject to significant leaching of iron particles or even chelation in acidic conditions (Dearing et al. 1985, 1995, 1996; Maher 1986). The negative effects of waterlogging may be mitigated in some gley soils through the formation of ferrimagnetic compounds by magnetotactic bacteria (Fassbinder et al. 1990; Stanjek et al. 1994; Maher 1998). In addition, even lake sediments may not provide the right conditions for dissolution to take place to a significant degree, as palaeofire records have been developed from lake sediments thousands of years old using MS (Gedye et al. 2000). But it is important to note that magnetotactic bacteria may be particularly prevalent in lacustrine systems (Snowball 1994), therefore causing a magnetic enhancement unrelated to fire.

Fig. 8.

Magnetic susceptibility by (a) soil organic carbon content with red dots representing burned samples, black unburned; and (b) pH. Data from: (a) Dearing et al. (1995), Oldfield and Crowther (2007), Blundell et al. (2009), Lowe and Wallis (2020), (b) Dearing et al. (1995), Blundell et al. (2009).

Rummery et al. (1979) found that the intensity of the fire is of great importance when it comes to enhancement of MS, as discussed in the previous section. Significant enhancement occurs when the soil is heated to more than 400°C, but the most significant enhancement occurs when the soil is heated to more than 550°C (Fig. 3). Depending on the duration of the fire and amount of ground-level fuel burned, the temperature reached in the soil will vary significantly, possibly over short distances, and may not reach 400°C (Rasmussen et al. 1986). In particular, grass fires may not cause the soil to reach the required temperatures (Roman et al. 2013). High fuel load is therefore of key importance to achieve the most significantly enhanced MS in soils through wildfires that result in the highest soil burn severity (Fitzpatrick et al. 2014). It has also been found that plants with a C₄ carbon fixation pathway result in a stronger MS signal when burned compared with plants with the more common C₃ pathway (Fig. 9; Table 2; Lu et al. 2000; Lu and Liu 2001). The vastly different MS readings from different plant ashes are likely also why the samples in Table 2 simply labelled ‘ash’ present such different readings. The comparatively high reading from the sample marked ‘white ash’ may at least in part be due to white ash resulting from complete combustion of fuel and therefore likely to indicate an intense fire (Hudak et al. 2013). It is important to note that the readings in Table 2 all come directly from plant ash and charcoal and as such do not necessarily reflect the MS readings of the affected soils.

Fig. 9.

Magnetic susceptibility by vegetation type, burned to charcoal (‘burned’) or burned to ash (‘ash’). Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Lu et al. (2000), Lu and Liu (2001).

Another potentially important factor is land use, though this causes a smaller enhancement than other factors previously discussed, with soil samples from a 20-year-old secondary forest clear-cut for agriculture producing the highest readings by far in Fig. 10 (Ketterings et al. 2000), though the numbers involved are much lower than those for soil type, for instance (Fig. 7). From Fig. 10, it also appears that grain size has a greater impact on MS readings than land use, which can also be seen in Fig. 11. However, this impact is not felt in soils heated above 600°C, owing to the texture changes observed in soils at these temperatures, which lead to an increase in sand and a decrease in silt and clay (Ketterings et al. 2000). For Australian soils, a coarse grain size appears to lead to a wider range of readings (Fig. 11), possibly linked to the aforementioned porosity leading to leaching, but only when precipitation is sufficiently high (Hu et al. 2020). Lastly, it is important to consider the post-fire environment, as this may have a significant impact on the measurable MS signal. Important factors include bioturbation, ploughing, dissolution and erosion (Blake et al. 2006). Erosion, in particular, may work to obscure the fire signal, both by eroding the burned soils and by depositing burned or naturally enhanced soils from elsewhere, which, depending on the depositional environment, may have travelled some distance (Bracken et al. 2015). However, as previously mentioned, by combining MS measurements and grain size analysis, it is at least possible to distinguish which sediments were enhanced as a result of fire, as these have a significantly finer grain size than sediments enhanced through weathering and soil formation alone (Oldfield and Crowther 2007). As highlighted in Gedye et al. (2000), using multiple parameters and quotients of MS and grain size is important to provide a complete fire record. These authors used six different parameters, one of which, MS measured at low frequency (χ_lf), is used in the figures presented in the present review, using the standardised units 10⁻⁶ m³ kg⁻¹, as the exponent varies between studies. One of the other parameters used by Gedye et al. (2000) is χ_ARM, or anhysteretic remanent susceptibility, which is sensitive to fine-grained magnetic material. Using this parameter, or a quotient derived from it, in addition to the more common measure used here (χ_lf), may aid in determining the origin of the MS signal. The other measurements used by Gedye et al. (2000) are: χ_fd, ‘Soft’, χ_ARM/SIRM and χ_fd/χ_ARM. χ_fd is frequency-dependent magnetic susceptibility measured as a percentage. This measures the presence of ultrafine superparamagnetic ferrimagnetic grains. ‘Soft’ indicates the presence of ferrimagnetic minerals and is measured as a percentage. χ_ARM/SIRM is the quotient of ARM and Saturation Isothermal Remanent Magnetisation (SIRM) and is used to distinguish between magnetic grain sizes. It is measured in 10⁻³ A m⁻¹. χ_fd/χ_ARM is a dimensionless quotient that is also used to distinguish between magnetic grain sizes. All these measurements are very useful in determining the source of the MS enhancement, as bacterial magnetosomes, for instance, only produce ultrafine magnetic grains, though these can also be produced in topsoils through normal pedogenic processes or through fire (Gedye et al. 2000). Χ_lf is used in the present review as it is the most commonly reported parameter. Older studies sometimes use other measures, such as emu (electromagnetic unit) (g^–1 × 10⁶) (Tite and Mullins 1971) or 10⁻⁶ G Oe (Gauss-Oersted) cm³ g⁻¹ (Rummery 1983). These are not presented here owing to the need to convert the units prior to comparison.

Fig. 10.

Magnetic susceptibility by land use, where ‘slashed’ indicates complete vegetation clearing. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Dearing et al. (1995, 1996), Ketterings et al. (2000), Blundell et al. (2009), Hu et al. (2020). res., residential.

Fig. 11.

Magnetic susceptibility by mechanical grain size and temperature. Each box represents the main spread of values, from the 25th to 75th percentiles, the central line in each box represents the median value, the dotted lines represent the minimum and maximum values in the data set, excluding outliers, the individual points represent statistical outliers. Data from: Longworth et al. (1979), Crockford and Olley (1998), Ketterings et al. (2000), Crockford and Willett (2001), Blundell et al. (2009).

Bioturbation is an important post-fire factor, as studies have shown that even severe wildfires may not negatively affect levels of bioturbation caused by ant mounds in Australian environments. This is thought to be in part due to their nesting habits (Richards et al. 2011a), with ant mounding having been found to increase the amount of highly erodible material by a factor of 2–30 times (Humphreys and Mitchell 1983). Ants are not the only cause of bioturbation in Australia, surface scraping by lyrebirds and mammals being another major category, which may in fact exceed that of ant mounding in some areas. Burrows of wombats and cicadas as well as termite mounds and earthworms have been found to have a less significant impact (Richards et al. 2011b). However, studies have also suggested that such burrows and mounds may in fact help to reduce the amount of soil loss during post-fire rains. The mounds can help in this regard by providing obstacles to the flow of water, and the ants’ tunnels and burrows made by other animals may help the water by-pass topsoil that may have been made water repellent by fire (Shakesby et al. 2007). Soils in Australian eucalypt forests tend to be water-repellent even without the presence of fire, but repellency has been found to increase after wildfires. Interestingly, soil water repellency has also been found to be eliminated in Australian soils heated to 260–340°C (Doerr et al. 2004). As discussed above, wildfires in Australia may not cause soils to heat to such a degree (Bradstock and Auld 1995), leaving this water repellent top layer in place. With bioturbation likely to vary greatly between locations, it is important to examine its impact on each chosen study site by, for instance, observing how distinct the soil layers are. This is not a new concern in palaeoscience, as bioturbation or roots penetrating to great depths can cause chronological issues through the introduction of younger materials (Matthews 1985).

Potential areas for use in Australia

As seen in Fig. 6, much of Australia is too dry for MS to be a viable option to reflect palaeofire records, which is probably why MS was found not to respond to wildfires at the Macquarie Marshes (Graves et al. 2019) or Two Mile Lake in Western Australia (Itzstein-Davey 2004; Gouramanis et al. 2016), and why it did not appear to respond to hearth fires at a rock shelter in the Pilbara region (Marwick 2005). Despite their name, the Macquarie Marshes only receive a MAP of 442 mm, not enough to supply a sufficient concentration of convertible iron oxides (Site 9 in Fig. 6; Singer and Fine 1989). Nearer the coast (with a higher MAP), at the Nattai River catchment, MS does appear to correspond well with wildfire events (Site 12 in Fig. 6; Blake et al. 2006), as it does at Lake Osborne in southwestern Tasmania, just slightly lagging behind peaks in charcoal (Site 21 in Fig. 6; Fletcher et al. 2014). It has also been shown that wildfires lead to the formation of maghemite in soils in northern Queensland (Site 3 in Fig. 6; Grogan et al. 2003), and a record of Aboriginal campfires has been constructed using MS at a rock shelter (Lowe et al. 2016), also in northern Queensland (Site 5 in Fig. 6). All these sites receive enough rainfall (i.e. >500 mm year⁻¹) for MS to be enhanced in the soils.

As MS has never been fully explored as a palaeofire proxy in Australia, very few authors have reported its use as such. However, where MS and charcoal have been published for the same site, implications can be made as to the potential of MS as a palaeofire proxy. It appears to have this potential at Tiam Point in northern Queensland (Site 1 in Fig. 6; Rowe 2007), Fern Gully Lagoon on North Stradbroke Island in Queensland (Site 7 in Fig. 6; Kemp et al. 2020, 2021) and Tea Tree Swamp in Victoria (Site 15 in Fig. 6; Gell et al. 1993). Other sites look promising but lack full charcoal records, meaning a definitive wildfire signal cannot be ascertained without further investigation. These sites include Myall and Tuggerah Lakes on the New South Wales–Queensland border (Sites 10–11 in Fig. 6; Skilbeck et al. 2005; Mooney and Maltby 2006), Callemondah (Site 14 in Fig. 6; Tibby et al. 2003; Gell et al. 2005) and Bolin Billabongs in Victoria (Site 17 in Fig. 6; Leahy et al. 2005), and Little Llangothlin Lagoon in New South Wales (Site 8 in Fig. 6; Gale et al. 2004; Woodward et al. 2011, 2014). At other sites, the MS record seems to correlate best with changes in organic matter content (e.g. Pearson et al. 2001; Haberle 2005; Adeleye et al. 2021), charcoal influx (Anker et al. 2001) or pollution from nearby mines (Augustinus et al. 2010). Certain sites in northern Queensland may be responding to changes in local hydrology (e.g. Prebble et al. 2005; Burrows et al. 2016), but for Caledonia Fen (Site 16 in Fig. 6; Kershaw et al. 2007) and Lake Keilambete (Site 18 in Fig. 6; Mooney 1997) in Victoria, it is unclear what the MS record is responding to. For Caledonia Fen, this could be due to the persistently wet conditions at this site leading to reductive dissolution of any ferrimagnetic compounds that may form (e.g. Thompson and Oldfield 1986; Blundell et al. 2009). The exact conditions that make this site so valuable for pollen studies may make it unsuitable for use of MS to reconstruct palaeofires. For certain periods, the MS record at Lake Keilambete appears to respond to changes in lake level, but for others, it may be responding to erosion events. The site is a highly saline volcanic crater lake (Mooney 1997), and it may be that owing to the chemical interactions in such a system, it does not produce a clear MS record. It is also possible that the soils were not heated to a sufficient degree at these Victorian sites, thus leaving an unclear MS signal. This can be conclusively determined by performing MS studies at similar sites, or by performing new MS studies at the same sites using different measurements.

Conclusion

The use of magnetic susceptibility as a proxy for past wildfires has had mixed success in Australia in the past, but there is clear potential in many regions, as well as a proven capacity overseas, both for modern and ancient sediments. Large soil sample studies both in Australia and overseas have highlighted the importance of parent material and the right level of precipitation on magnetic readings. For these background readings to then be significantly enhanced by fire, the soil needs to be heated to a high temperature, which may not occur during most Australian fire events. This could be due to a lack of woody fuel, a short fire duration or the prevalence of canopy fires over ground-level fires.

In order to clarify what site and fire parameters are of most importance on this iron-rich, dry and fire-prone continent, a comprehensive field study is recommended. This study should take into account all site characteristics mentioned above and needs to include sites of various drainage capacities, depositional environments, dominant vegetation and organic matter contents, as well as sites with varying levels of inorganic influx in order to quantify the significance of these parameters. Targeting different levels of soil burn severity will also be of the utmost importance to determine the capacity of wildfires in Australia to heat the soil to a sufficient degree for a clear MS signal. Site parameters influencing post-depositional factors also need to be considered, with erosion and bioturbation being top of the list. It is clear that the use of MS as a wildfire proxy in Australia will be restricted to high precipitation regions around the coast, but it will nonetheless be of vital importance when determining Australia’s true wildfire frequency and how this has been influenced by prevailing climate conditions in the past. Using MS in combination with charcoal may be the ideal methodology, as these two techniques complement each other in some ways. Charcoal can be used for a general fire signal, with MS used to identify which events caused a particularly high soil burn severity at a local level, and possibly to determine which specific areas were most impacted. To what extent this can be done successfully in the Australian context has yet to be determined.