1 Introduction

There is growing consensus that anthropogenic climate change is causing longer and more intense fire seasons (Balch et al., 2022; Ellis et al., 2022; Jain et al., 2022), with a corresponding increase in the number of wildfires exhibiting extreme behaviour with adverse socio-ecological and -economic impacts (Bowman et al., 2017; Duane et al., 2021). Climate projections suggest this trend is likely to continue throughout the 21st century (Flannigan et al., 2013; Wotton et al., 2017; Abatzoglou et al., 2019; Abatzoglou et al., 2021). Western North American forests, for example, have shown a rapid increase in annual burnt area in the last half century (Westerling, 2016; Williams et al., 2019) – albeit this trend is partially driven by a long history of deleterious land-use practices such as fire suppression (Abatzoglou et al., 2018). Current climatological and ecological modelling suggests the occurrence of catastrophic megafire events in the USA, such as those seen in the 2020 fire season (Higuera and Abatzoglou, 2020), will continue to increase across affected forests due to anthropogenic climate change (Bowman et al., 2020). The potential severity of these events, however, may be reduced by fuel management intervention that restores pre-colonial fire regimes (Abatzoglou et al., 2021; Hessburg et al., 2021). A similar increase in extreme fires is also evident in southeastern Australia (Sharples et al., 2016; Abram et al., 2021), where trends in both drought and extreme fire weather combined with the loss of Indigenous fire management are believed to have contributed to the catastrophic 2019 – 2020 fire season dubbed the Australian Black Summer (Boer et al., 2020; Nolan et al., 2020; Canadell et al., 2021; Collins et al., 2021; van Oldenborgh et al., 2021; Mariani et al., 2022).
The fuel moisture content of live and dead fine fuels is inherently linked to the behavioural and ignition determinants of wildfire driving these trends across multiple spatial and temporal scales (Bradstock, 2010; Murphy et al., 2013). Dead fine fuel moisture content (DFFMC) has specifically been identified as a key switch determining the spread of vegetation fire. Dead surface fuels such as leaves, bark, twigs, and grass are capable of rapidly equilibrating with atmospheric humidity in under 10 hours, and consequently represent the most ignitable component of vegetation. As the ignition of these fuels can then provide the energy to ignite both larger dead fuel components and live fuels, DFFMC is therefore a key determinant of overall wildfire occurrence, behaviour, and greater pyrogeographic patterns. To track DFFMC and associated potential fire danger, DFFMC can be measured directly in the field (e.g, Bowman et al., 2020) or estimated using meteorological indices such as vapour pressure deficit (VPD) or the Canadian Fire Weather Index system (FWI: Van Wagner, 1987). Accordingly, derived or measured DFFMC functions as a prominent proxy for an ecoregion’s potential to ignite and sustain wildfire, accurately reflecting the surface litter most prone to ignition and fire spread (Murphy et al., 2013; Flannigan et al., 2016; Kelley et al., 2019). DFFMC is also of fundamental importance in shaping spatiotemporal patterns of landscape fires from local to global scales. For example, moisture differences are understood to control the spread of fires across different vegetation types such as savannas and rainforests (Little et al., 2012).
Prior research has identified specific values of DFFMC as marking the upper and lower bounds of wildfire potential (Fernandes et al., 2008; Wotton, 2008; Slijepcevic et al., 2015; Flannigan et al., 2016; Nolan et al., 2016; Boer et al., 2017; Filkov et al., 2019; Clarke et al., 2022). For instance, FWI- or VPD-derived DFFMC values of between 8 and 12% have been linked to uncontrollable wildfire (Slijepcevic et al., 2015; Flannigan et al., 2016; Nolan et al., 2016; Boer et al., 2017; Filkov et al., 2019), while values of up to 30% have been found to represent the upper limits of vegetation fire (Fernandes et al., 2008; Nolan et al., 2016) under current atmospheric oxygen concentrations due to carbon fibre saturation (Luke and McArthur, 1978; Scott and Glasspool, 2006).
Global and regional studies frequently estimate DFFMC within the Canadian FWI system due to its ease of calculation and interpretation (Field, 2020); these studies have often applied single value thresholds to assess patterns and likely trends in global fire potential. For instance, an established 10% threshold (Wotton, 2008; Flannigan et al., 2016) was used to infer that the proportion of fire seasons falling below this critical threshold had significantly increased between 1979 and 2019 for most ecoregions worldwide (Ellis et al., 2022). Such a shift is likely to affect established fire regimes, particularly for more productive or wet ecoregions.
There are, however, inherent problems with applying single-value thresholds like this to compare different vegetation types and climate domains. Forest ecosystems have evolved with fire differently in response to naturally occurring environmental and biological constraints like soil fertility, climate, and the local biota. It is thus unlikely that a single DFFMC value could possibly reflect the critical point of flammability across grossly different ecoregions. Furthermore, fire danger indices are developed for specific, regional forest types, with the FWI having been developed based on a mature Pinus banksianaLamb. and Pinus contorta landscape in southern Canada (Van Wagner, 1987; Wotton, 2008). Despite the FWI’s established global applicability, there is no reason to assume that the vegetation in these forests is globally representative, and its use may consequently be inappropriate for assessing fire potential or true fuel moisture in different vegetation types (e.g., Aguado et al., 2007; Wotton and Beverly, 2007; Schunk et al., 2017). This raises a question about how fire weather indices can best be used to retain local relevance by identifying ecoregion-specific thresholds controlling the switch from a non-flammable to a flammable state (e.g, Clarke et al., 2022).
Building on these studies, we seek to answer the question of whether a universal fuel moisture threshold exists as a control on fire, while advancing our understanding of how fuel moisture acts as a switch for landscape fire both regionally and globally. Using a comprehensive dataset of over 700 hierarchically defined ecoregions, we identify the ecoregion flammability thresholds (EFTs) as the FWI-derived DFFMC value most associated with rapid step changes in remotely sensed fire activity records. Our methods (Figure 1) bypass the issues inherent with applying fire weather indices globally by localising the relationship between DFFMC and fire to each individual ecoregion with associated satellite fire records. We then use a combination of nonmetric multidimensional scaling and inferential statistical modelling to verify the veracity of EFTs as a biological, localised mechanism and examine both the biogeographic and climate characteristics constraining our identified EFTs as critical fuel-fire switches.