Introduction
Globally, the distribution, seasonality, frequency and intensity of fires has changed in recent decades due to anthropogenic global change drivers including climate change, land-use change (with fire often used to clear vegetation to facilitate land-use change) and, in some cases, invasion by more flammable species (McLauchlan et al., 2020, Kelly et al., 2020). These changes are predicted to accelerate over the next few decades (Sheehan et al., 2019; Enright et al., 2015; Aragão et al., 2008). There is particular concern regarding the impacts on fire-sensitive tropical ecosystems, many of which are being rapidly lost and degraded (Alroy, 2017; Busch & Ferretti-Gallon, 2017), making the tropics the epicentre of current and future extinction risk (Edwards et al., 2019). Given these changing fire regimes, it is crucial to understand how fire influences biodiversity, and the rate of recovery following fire events (Kelly et al., 2020). This need is widely recognised, for example, by the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES, 2019) and the UNFCC REDD+ program (UNFCC, 2019).
The impacts of fire on biodiversity are, however, complex and incompletely understood (Gill et al., 2013; McLauchlan et al., 2020; Tingley et al., 2016), with positive and negative impacts reported (Kelly et al., 2020; Giorgis et al., 2021). Some tropical biomes, such as woody savannas and grasslands, are frequently exposed to fire, and several species that are characteristic of these biomes require fires to persist (Simon & Pennington, 2012). In such biomes, fire positively influences the diversity of photophilic floras and faunas (Pausas & Keeley, 2019), with a landscape mosaic of vegetation patches that vary in the time since they were burnt typically maximising biodiversity (Driscoll et al., 2010). Long-term suppression of fire in these systems typically generates more homogenous vegetation patches that support fewer species (Giorgis et al., 2021; Abreu et al., 2017), promotes woody species and gradual shifts from grasslands to woody savannas, and then shrublands and forests (Probert et al., 2019).
In other biomes, such as tropical moist forests, fire is historically extremely rare and most plant species are highly sensitive to fire (Cochrane & Schulze, 1999, Giorgis et al., 2021). Consequently, recent increases in the number of fires are a primary driver of tropical moist forest degradation and biodiversity loss (Barlow et al., 2019; Lewis et al., 2015), including tree (Galvao de Melo & Durigan, 2010, Cochrane & Schulze, 1999) and forb communities (Gordijn et al., 2018). Increased exposure to fire can also eventually convert moist tropical forest ecosystems into open habitat and savannas (Flores & Holmgren, 2021).
Biodiversity will gradually recover following a fire event, and should increasingly resemble the pre-fire community as time increases (Machida et al., 2021). Frequent fire events can, however, prevent full recovery by driving fire sensitive species to regional extinction (Gallagher et al., 2021), and species recovery following fire can be much slower in fire sensitive biomes than those that traditionally experience fire (Nelson et al., 2014).  Understanding of how biodiversity recovers following fire events is, however, still insufficiently developed. In part, this is because many studies of biodiversity responses to fire focus exclusively on species richness, even though fires have strong impacts on community composition and generate considerable turnover, i.e. beta diversity (Gordijn et al., 2018; Durigan et al., 2020, Peterson & Reich, 2008). The influence of landscape context on biodiversity recovery following fire is also insufficiently understood. Recovery rates are likely to be faster within relatively intact ecosystems (i.e. effectively protected from anthropogenic stressors) in which a greater abundance of natural vegetation increases the availability of propagules that can recolonise burnt sites. Well managed protected areas may thus facilitate faster recovery from fires, although tropical protected areas vary greatly in their effectiveness, including at reducing fire risk (Laurance et al., 2012).
Most studies assessing fire impacts on plant biodiversity focus on single study locations. Meta-analyses are scarce but have assessed the relative fire sensitivity of native and exotic plant species (Jauni et al., 2015; Alba et al., 2014; Aslan & Dickson, 2020). Here, we build upon a systematic compilation of data from published studies of tropical and sub-tropical plant community responses to fire. We work on plants as they comprise a wide range of life forms and life history strategies and provide the habitat structure and resources that are exploited by other taxonomic groups. We assess post-fire recovery of plant species richness and composition following fire events. Specifically, we test whether species richness and beta diversity (i.e., species turnover) between burnt and unburnt plots respond differently to time since fire and fire type (prescribed burns versus non-prescribed burns). We also assess if protected area status (protected vs unprotected) moderates’ responses of species richness and species turnover to fire events. Our analyses take biome identity into account and distinguishes between prescribed and non-prescribed burns. We do so as prescribed burns are often used in management programmes to reduce the amount of flammable material and thus the size and intensity of subsequent fires. The practice has, however, been criticised (Ryan et al., 2013), with some studies suggesting that prescribed burns can alter plant communities in a manner similar to non-prescribed burns (Ffolliott et al., 2012; Pastro et al., 2011).