The existing records are distributed across the main subregions of the Neotropics, ranging from the moist tropical forests to the semiarid xeric vegetation (Figs. 1 and 2). While the Andes are particularly well sampled (Fig. 2a), areas such as central Amazonia, arid and semiarid regions (e.g., northeastern Brazil), and central regions of the Neotropics located between 20 and 40° S (Fig. 2a) are underrepresented.

In terms of chronological representation, site density increases towards the Holocene (Fig. 2b). Across the Neotropics, considering time bins of 400-year intervals, the LGM (21–19 ka) is supported by an average of 30 pollen and 20 charcoal records, followed by the deglacial period (19–11.7 ka; 62 and 35, respectively), and the Holocene (<11.7 ka; 152 and 82; Fig. 2b). Therefore, the Pleistocene and particularly millennial-scale events such as Heinrich stadial 1 (HS1, 18–14.8 ka), Antarctic Cold Reversal/Bølling–Allerød (ACR/BA) (14.8–12.9 ka), and the Younger Dryas (YD) (12.9–11.7 ka) are represented by fewer samples and consequently lower spatial representativity.

Arboreal pollen and charcoal influx z-scores are positively correlated in TrAn (r=0.41), SESA (r=0.63), and ExTrAn (r=0.78), and negatively correlated in NNeo (r=-0.56) (p<0.001) and CEB (r=-0.34), while no clear correlation is observed for Amazonia (r=0.18) (Fig. A2).

Despite being represented by a relatively small number of records, during the LGM (21–19 ka), positive AP z-scores are recorded in the Northern Hemisphere subregion NNeo (Fig. 4a), in opposition to mostly negative AP z-scores in Southern Hemisphere subregions (Fig. 4b–g). Meanwhile, charcoal influx z-scores are predominantly negative for all subregions (Fig. 4k).

By the onset of the deglacial period (19–11.7 ka), AP increases in TrAn, NEB, SESA, and ExTrAn (Fig. 4c, d, f, g), contrasting with AP decrease in NNeo (Fig. 4a). No clear long-term trend is detected for Amazonia and CEB. Charcoal influx z-scores in the Southern Hemisphere exhibit negative values during the deglaciation (Fig. 4i–m), followed by a stepwise increase towards interglacial levels at 14 to 13 ka. On the other hand, the Northern Hemisphere (NNeo) presents a decrease in charcoal influx z-scores at 13 ka, towards minimum EH values (Fig. 4h).

The Holocene period shows distinct regional patterns in AP z-scores and charcoal influx:

In the early Holocene (EH; 11.7–8.2 ka), AP z-scores reach high values for NNeo and TrAn (Fig. 4a, c), low values for NEB, CEB, and Amazonia (Fig. 4b, e), and maintain a continuous increase in ExTrAn and SESA (Fig. 4f, g). This period also records higher charcoal influx z-scores in Amazonia, CEB, SESA, and ExTrAn, and lower values in the NNeo.

In the mid Holocene (MH; 8.2–4.2 ka) AP z-scores values increase in Amazonia, NEB, CEB, SESA, and ExTrAn, opposed to decreasing values in NNeo. The charcoal influx reaches maximum values in the TrAn and increases in the NNeo, coinciding with decreasing AP trends. On the other hand, charcoal influx decreases slightly in the CEB and ExTrAn and remains relatively constant in Amazonia and SESA.

During the late Holocene (LH; 4.2–0 ka), NNeo, TrAn, and NEB show a drop in AP z-scores reaching the lowest Holocene levels. In NNeo, these shifts coincide with some increase in charcoal influx. Conversely, during the same period, AP z-scores attain maximum values in Amazonia, CEB, SESA, and ExTrAn, while charcoal influx decreases in CEB and increases in ExTrAn.

The comparison between z-scores and the proportional relative scaling method from McMichael et al. (2021) shows that despite changes in the amplitude of variability, all regions hold mostly coherent patterns of changes and a similar temporal structure. In addition, z-scores appear more stable and better suited to capturing long-term trends in fire activity (see Supplement, Figs. S1 and S2).

During the LGM, high levels of tree cover and weak fire regimes (Fig. 5d, e) are consistent with estimates of 4–5 °C drop in mean annual temperatures (Correa-Metrio et al., 2012) and wet conditions (Hodell et al., 2008; Deplazes et al., 2013; Fig. 5a, c). Throughout the deglaciation, a marked decrease in tree cover and intensification of the fire regime relate to predominant drier phases promoted by southward displacements of the ITCZ associated with millennial-scale events (HS1 and the YD) (Fig. 5b, c) (Deplazes et al., 2013; Haug et al., 2001). The wetter interval linked to BA/ACR (Deplazes et al., 2013; Fig. 5b) is not clearly detected in our analyses. Additionally, human occupation in the NNeo began during the late Pleistocene (Ardelean et al., 2020), although in smaller populations (Fig. 5d), and likely started contributing to environmental changes. These combined changes importantly affected megafaunal populations in NNeo, yet the implications of their decline and subsequent extinction for fire and vegetation dynamics remain unclear (Dávila et al., 2019; Rozas-Davila et al., 2021). Nevertheless, late Pleistocene tree cover and fire dynamics suggest predominant hydroclimate control (Fig. 5b–f) and an overall negative correlation between tree cover and fire activity (Fig. A2a) suggests moisture-limited conditions for fire.

At the onset of the Holocene, during the EH, a marked shift towards expansion of tree cover and minimum fire activity (Figs. 5e, f; A2a) were induced by increasing wetter conditions (Haug et al., 2001; Hodell et al., 2008). During this period, human populations began engaging in agricultural activities, domesticating maize and squash by ca. 9 ka (Piperno et al., 2009), likely impacting local fire regimes. However, the gradual long-term tree cover decrease and fire activity increase synchronous with a progressive transition to drier conditions (Haug et al., 2001), still suggest a main climatic driver from the EH until ca. 4 ka (Fig. 5b, e, f). From ca. 4 ka onwards, the expanding anthropogenic pressure in Central America became a clear driver of vegetation and fire changes (Fig. 5d) (Harvey et al., 2019; Leyden, 2002). The LH drop in tree cover decouples from the more gradual climate-driven decrease observed since the onset of the Holocene (Figs. 5a, e; A2a). This latter shift in AP is coeval with the demographic expansion of Mesoamerican populations (Fig. 5d). Fire activity also stays relatively high during this period (Fig. 5f). At ca. 1.2 ka, the increase in tree cover and decrease of fire activity was possibly related to the rapid decrease in populations likely associated with the collapse of the Maya civilization (Gill et al., 2007; Haug et al., 2003).

The gigantic extent of Amazonia and its equatorial positioning results in heterogeneous meridional and zonal environmental patterns for the region. Limited number of records in the LGM and deglacial period results in poor spatial and temporal constraints of tree cover and fire activity variability, thus preventing a generalization and detailed assessment of the observed patterns (Fig. 6f). Additionally, caution is needed when interpreting vegetation changes from fluvial and floodplain records, as edaphic factors, rather than climate, may control the observed patterns.

During the LGM, the region featured reduced tree cover, mainly at the ecotones and eastern areas, and weak fire activity (Fig. 6e, f), although its western and core regions remained mostly forested (Akabane et al., 2024; Colinvaux et al., 1996; Haberle and Maslin, 1999; Urrego et al., 2005). This pattern of low tree cover was likely a response to 4–6 °C colder temperatures (Bush et al., 2001; Colinvaux et al., 1996; Stute et al., 1995), low CO_2atm (Fig. 6a) (Bereiter et al., 2015), and reduced rainfall in eastern Amazonia (Fig. 6b, c) (Häggi et al., 2017; Wang et al., 2017). Reduced fire activity, despite low tree cover, may also have been a consequence of colder conditions and reduced convective activity, which would result in fewer lightning ignitions, and absence of widespread human impacts.

During the deglacial period, our data indicate oscillating tree cover patterns during HS1 and highest deglacial AP values by 13–12 ka, coinciding with minimum fire activity (Fig. 6e, f), an intensified SASM (Cheng et al., 2013; Mosblech et al., 2012) and increased CO_2atm (Fig. 6a, b). These pronounced oscillations in AP and charcoal z-scores may arise from contrasting meridional changes in precipitation patterns associated with millennial-scale events. For instance, during HS1 and the YD, while drier conditions expanded over northern Amazonia (Akabane et al., 2024; Deplazes et al., 2013; Zular et al., 2019), southern areas experienced wetter conditions (Campos et al., 2019; Mosblech et al., 2012; Novello et al., 2017).

The subsequent EH decline in AP was likely driven by a weakening of the SASM and the stepwise increase in fire activity (Fig. 6e, f), which was mainly concentrated to the eastern and southern areas (Gosling et al., 2021) and may have played an important role in reducing tree cover over ecotones. Moreover, the end of the Pleistocene marks the beginning of human occupation of the Amazon basin, which increased in the EH (Fig. 6d), when they were already engaging in resource management practices (Neves et al., 2021). The impacts of megafaunal extinction also initiated long-term changes in both nutrient distribution and species turnover (Doughty et al., 2013, 2016b). However, its correlation with overall tree cover and fire regime changes remains elusive for the region.

During the MH, a stepwise increase in tree cover is recorded at 7 ka, mostly reflecting forest expansion in northern Amazonia (Behling and Hooghiemstra, 2000) and coinciding with monsoon strengthening (Fig. 6b, e). Meanwhile, some increase in fire activity, amid forest expansion and progressively wetter conditions, may relate to expanding human populations in Amazonia (Cordeiro et al., 2014; Gosling et al., 2021; Riris and Arroyo-Kalin, 2019). The absence of a clear decrease in tree cover as consequence of anthropogenic activity is likely due to agroforestry practices that allowed for long fallow periods and forest recovery (Iriarte et al., 2020).

In the LH, gradual tree cover expansion is mainly driven by the southward forest expansion that formed the modern extent of the rainforest (Fontes et al., 2017; Mayle et al., 2000) associated with a further intensification of monsoon strength (Baker and Fritz, 2015) and moisture increase. Some increase in fire activity after ca. 3 ka has been potentially favored by dryness associated to intensifying ENSO activity (Kanner et al., 2013; Mark et al., 2022) in combination to expanding human populations (Fig. 6d). The decline in fire activity over the last 0.5 ka coincides with the indigenous demographic collapse following the European contact (Fig. 6d, f).

Climate changes in the Andes were heterogeneous and asynchronous (Bush and Flenley, 2007). During the LGM, predominating open vegetation (Fig. 7e) was likely controlled by 5–8 °C cooler-than-modern temperatures exerting the main limiting factor for tree cover development, even though moist conditions prevailed (Baker et al., 2001; Bush et al., 2004; Cheng et al., 2013; Paduano et al., 2003; Valencia et al., 2010) (Fig. 7a, b). Subsequently, TrAn saw an increase in arboreal taxa from 18 to 14.8 ka, when the region became warmer and significantly wetter (Fig. 7a) (Martin et al., 2018; Palacios et al., 2020). Despite uncertainties related to the limited number of records, a slight fire increase in the second part of HS1 may have resulted from higher biomass availability (Fig. 7d). From 14 to 12.9 ka, pervasive decrease in tree cover throughout the BA/ACR until the onset of the YD likely relate to drier conditions in the southern TrAn and overall glacier expansion (Jomelli et al., 2014), which is then followed by a tree cover recovery throughout the YD potentially associated to further warming and the return of wetter conditions (Fig. 7a, d). Positive correlation of increasing fire activity throughout the LGM and deglacial period along with increasing trends of tree cover (Fig. A2c) point to fuel-limited conditions for fire, although prevailing wet conditions during the late Pleistocene may have also played a key role in restraining fire (Fig. 7a, b).

During the EH, tree cover achieved its highest elevation during the Holocene (Fig. 7a). This tree cover increase was likely maintained by warmer temperatures and rising of the tree line, despite relatively drier conditions in response to a weaker SASM, as suggested by δ18O data from ice cores and speleothems (Cheng et al., 2013; Thompson et al., 1998; Vuille et al., 2003) (Fig. 7a). As in NNeo and Amazonia, evidence for human agricultural activities in TrAn began during this period (Pagán-Jiménez et al., 2016). Additionally, an overall decline of megafauna in tropical Andes resulted in sensitive ecological consequences associated with vegetation turnover, e.g., the encroachment of both palatable and woody taxa, as well as fuel build-up (Bush et al., 2022; Pym et al., 2023). Several interacting factors, such as warming, hydroclimate changes, initial expansion of human populations, and megafaunal decline, collectively contributed to intensifying the fire regime in the region (Bush et al., 2022; Pym et al., 2023) (Fig. 7f). Consequently, increased fire activity hampered further development of a continuous tree cover above the tree line resulting in a sharper vegetation boundary (Rehm and Feeley, 2015). This suggests a complex interplay of abiotic and biotic drivers shaping the observed pattern.

The MH is characterized by a stepwise intensification of the fire regime, with tree cover only showing a slight decreasing trend, as indicated by our regional-integrating AP composite (Fig. 7d, e). However, subdividing TrAn at 8° S in northern and southern sectors reveals distinct trends (Fig. A3): while fire activity increased across both regions, tree cover remained stable in the north but declines in the south. These trends are consistent with heterogeneous hydroclimatic patterns along Andes, with drier conditions recorded by lowered lake levels, mainly in the Altiplano (Baker et al., 2001; Bush and Flenley, 2007; Hillyer et al., 2009) and persistent wet conditions recorded in central sectors of the TrAn (Bustamante and Panizo, 2016; Cheng et al., 2013; Polissar et al., 2013) (Fig. 7a). Modern climate patterns indicate that anomalies in equatorial Pacific SST can produce different regional impacts in precipitation by affecting moisture-laden easterlies and the Bolivian High (Garreaud et al., 2003; Poveda et al., 2020; Vuille, 1999). For instance, reduced ENSO variability has been suggested to decrease moisture balance in parts of the Andes, particularly in the southern TrAn, parts of the northern TrAn and NNeo (Figs. A3; 7c) (Polissar et al., 2013). Accordingly, a potential correspondence between the zonal Pacific SST gradient anomaly and tree cover trends during the EH and MH may exist for southern TrAn but not for northern TrAn (Fig. A3). Furthermore, while δ18O in ice and speleothem cores and δD from lake sediments primarily reflect rainy season precipitation (Cheng et al., 2013; Fornace et al., 2014; Vuille et al., 2003), fluctuations in lake levels are influenced by annual precipitation (Theissen et al., 2008). This may alternatively suggest that despite a gradual increase in summer precipitation throughout the Holocene, the MH featured a decrease in annual precipitation over the Altiplano. Nevertheless, the maintenance of tree cover was also favored by the persistence of moist microclimates (Ledru et al., 2013; Nascimento et al., 2019). These conditions, as well as anthropogenic activities, would have favored a regional intensification of the fire regime and decline in tree cover. By ca. 6 ka, agropastoral systems based on maize agriculture and llama herding became widespread (Nascimento et al., 2020).

The LH marks a major decrease in tree cover and high fire activity in the whole TrAn (Fig. 7c), probably related to the expansion of human impacts in the region and increasing sedentism (Goldberg et al., 2016; Valencia et al., 2010). Anthropogenic activity likely maintained fire activity at similar rates to the previous MH dry phase (Fig. 7f) despite moisture increase, maintaining a lowered and sharper tree line (Schiferl et al., 2023). During the last 0.5 ka, tree cover expansion and decreased fire activity may point to the abandonment of sites as consequence of the European contact and demographic collapse in the region (Koch et al., 2019).

In NEB, the scarcity of records decreases the precision and accuracy of the observed trends (Fig. 8f) and prevents producing a composite curve using charcoal data. Therefore, we only discuss major features of tree cover and fire dynamics.

During the LGM, the region experienced low tree cover driven by prevailing dry conditions (Cruz et al., 2009; Dupont et al., 2001; Mulitza et al., 2017). The deglaciation period is marked by tree cover expansion, associated with phases of southward displaced ITCZ during HS1 and YD (Fig. 8a, b, d) (Bouimetarhan et al., 2018; Cruz et al., 2009; Dupont et al., 2001; Mendes et al., 2019; Venancio et al., 2020). These periods featured the onset of forest corridors connecting Amazon and Atlantic forests and decreased fire activity during YD (Bouimetarhan et al., 2018; Dupont et al., 2009; Ledru and de Araújo, 2023; De Oliveira et al., 1999), which is depicted as the interval with highest AP z-scores over the last 21 ka (Fig. 8d).

During the EH, decrease in tree cover is likely a consequence to both climate changes and increasing human impacts. This period features drier conditions relative to YD in northern NEB due to the northward displacement of the ITCZ (Fig. 8a) (Mendes et al., 2019; Prado et al., 2013b; Venancio et al., 2020) and the expansion of human populations in the region (Araujo et al., 2025) (Fig. 8c). In the subsequent MH period, increasing trends in tree cover (Fig. 8d) may reflect relatively wet conditions due to a weak Nordeste Low (Cruz et al., 2009; Prado et al., 2013a) (Fig. 8b) and southward shifted seasonal migration range of the ITCZ (Chiessi et al., 2021), in addition to decreased human populations in inland parts of NEB (Araujo et al., 2025) (Fig. 8c). The LH was marked by a decrease in tree cover and intensification of the fire regime, likely in response to the establishment of the modern semiarid conditions over most NEB (Chiessi et al., 2021; Cruz et al., 2009; Utida et al., 2020) and rapid increase in human populations (Fig. 8c) (Araujo et al., 2025).

The relationship between tree cover and fire in the NEB remains elusive due to the limited number of records (Fig. 8e). Charcoal records suggest a negative correlation between tree cover and fire (Fig. 8f). This observation, however, is apparently counterintuitive, given that Caatinga currently exhibits the opposite pattern (fuel-limited conditions), where the lack of fuel inhibits fire activity (Argibay et al., 2020) (Fig. 4a, c). We propose two potential explanations for this apparent discrepancy. First, the fire records are located at the margins of the semiarid region, thus capturing influences from adjacent tropical savannas and forests, rather than exclusively reflecting xeric vegetation-fire dynamics. Second, increasing human occupation during the LH (Fig. 8c) may have intensified burning activities, even as natural fire frequencies declined. While NEB contains the earliest evidence of human occupation of the Americas (Boëda et al., 2014), only the LH saw a proliferation of more sedentary, possibly agricultural, ceramic-producing societies (Oliveira, 2002).

During the LGM, tree cover and fire activity exhibit high trends; however, the scarcity of records hampers a regional generalization and underscores the need for additional data from this period (Fig. 9f). During the deglacial period, high tree cover and low fire activity trends were favored by periods of intensified rainfall in the region, i.e., HS1 and the YD (Fig. 9a, b, d, e) (Campos et al., 2019; Martins et al., 2023; Meier et al., 2022; Stríkis et al., 2015, 2018), which allowed the widespread migration of cold- and moist-adapted tree taxa through central Brazil (Pinaya et al., 2019). An increase in fire activity centered in ca. 16.5 ka may result from a short dry incursion during HS1 (Stríkis et al., 2015), which has yet to be confirmed, as the scarcity of data prevents detailed assessment of the observed pattern (Fig. 9e).

Our compilation suggests a prevailing decrease in tree cover and intensification in fire activity from the YD to the EH (Fig. 9d, e). These trends were likely a response to a weaker SASM/SACZ led by low austral summer insolation, which yielded drier conditions in the region (Cruz et al., 2005; Prado et al., 2013b; Wong et al., 2023). The stepwise increase in fire activity, peaking at ca. 10.5 ka, further contributed to the tree cover rapid decrease (Fig. 9d, e). Additionally, human activity probably also contributed to tree cover and fire trends during this period, as archaeological records show well-established occupations from ca. 13 ka onwards and expanding population in the EH (Fig. 9c) (Araujo et al., 2025; Strauss et al., 2020). Furthermore, the megafauna functional extinction during this period likely initiated long-term changes in vegetation composition (Raczka et al., 2018) and potential increase of fuel loads (as in Gill, 2014). This effect, however, seems to have been secondary, as a decrease in tree cover is opposite to the expected response by megafaunal extinction alone (Doughty et al., 2016a; Macias et al., 2014).

Throughout the MH and the LH, the progressive increase in rainfall driven by a progressive strengthening of the SACZ (Cruz et al., 2005; Meier et al., 2022) (Fig. 9b) favored the expansion of tree cover and the attenuation of fire activity (Fig. 9d, e). In the LH, despite a second wave of human impacts in the region after an occupation hiatus in the MH (Araujo et al., 2005, 2025), tree cover exhibits a continuous increase and fire activity a decreasing trend (Fig. 9c, d, e).

In CEB, long-term tree cover changes are negatively correlated with fire activity (Figs. 9d, e, A2d). This pattern points to a feedback mechanism in which herbaceous vegetation facilitates fire activity, while fire, in turn, contributes to the dominance of herbs by limiting tree encroachment. Conversely, moisture-driven development of woody formations leads to the suppression of fire, which further contributes to tree cover expansion (Fig. 9d, e).

During the LGM, the low arboreal pollen z-scores reflect the dominating open physiognomies (Fig. 10c) (Behling, 2002b; Gu et al., 2018), despite moist conditions sustained by a strong SASM/SACZ influence (Cruz et al., 2005). This predominating open vegetation was likely controlled by 3–7 °C lower temperatures (Behling, 2002a; Chiessi et al., 2015) and reduced CO_2atm levels (Fig. 10a, b) (Bereiter et al., 2015). Moreover, stronger Antarctic cold fronts may have shifted the woody savanna and forest boundaries further north, favoring the prevalence of open grasslands. The presence of a diverse megafauna potentially contributed by restricting both woody encroachment and fuel accumulation (Furquim et al., 2024; Macias et al., 2014; Prates and Perez, 2021; da Rosa et al., 2024). The combination of low woody biomass, higher impact of herbivory, cold and moist conditions restricted fire activity during this period (Fig. 10f). The long-term and gradual trend of tree cover expansion and fire intensification from the LGM to the EH suggests that scarcity of fuel was a limiting factor for fire activity during the Pleistocene (positive tree cover-fire correlation) (Figs. 10e,f, A2e). Alternatively, or concurrently, both trends may have been driven by progressive deglacial warming. Notably, our composite curves of tree cover and fire activity show no clear correlation with millennial-scale hydroclimate variability throughout most of the late Pleistocene, suggesting that moisture availability was not the primary driver, with temperature and CO_2atm exerting greater control. However, Campos et al. (2019) suggested that no generalized increase in precipitation occurred in SESA during HS1 and the YD. Furthermore, while human occupation began at ca. 14–13 ka in the region, impacts were likely still limited (Araujo et al., 2025; Araujo and Correa, 2016; Suárez, 2017).

During the EH, a peak in fire activity (Fig. 10f) was facilitated by the availability of biomass coupled with warmer temperatures and relatively drier conditions (Fig. 10b, c, e, f). The intensified fire regime, on the other hand, probably contributed to the retraction of forest formations during this period (Fig. 10e, f). Throughout the Holocene, a continuous increase in arboreal biomass decoupled from CO_2atm and temperature trends suggests a forest expansion mainly driven by increasing precipitation and the gradual suppression of fire activity (Fig. 10c, e, f). The extinction of megafauna by the end of the EH potentially contributed, at least in part, to the expansion of woody taxa (Behling et al., 2023; Macias et al., 2014). From 6 to 3.5 ka, the continuous expansion of tree cover is slowed down, coinciding with a peak in fire activity and expanding human populations in the region (Fig. 10d–f). In the last 2 ka, despite an increase in human populations, tree cover and fire activity do not exhibit a clear response.

The Pleistocene to Holocene transition likely represents a transition from temperature/CO_2atm-limited to moisture-limited conditions, when an inflexion in the correlation of tree cover and fire is observed (Fig. A2e). Over the 21-kyr analyzed period, tree cover and fire activity show a positive correlation with both increasing in the long-term but with a negative correlation on detrended data (r=-0.48, p<0.001, not shown) (Fig. 10c, d). This suggests different relationships at different timescales. In the long-term, either the strengthening of the fire regime was driven by increase in fuel availability from woody biomass, or both tree cover and fire responded to the same external forcing (i.e., deglacial warming). In the short-term, fire events likely acted as a limiting factor for tree cover development.

During the LGM, ExTrAn was characterized by open physiognomies and a weak fire regime (Fig. 11e, f). This was likely consequence of significantly cold conditions (Fig. 11a, b) (Massaferro et al., 2009), while both wet and dry conditions have also been suggested along ExTrAn (Montade et al., 2013; Moreno et al., 2018). After ca. 18 ka, tree cover increases along with warming temperatures (Kaiser et al., 2005) and rising CO_2atm (Fig. 11a, b), with an accelerated tree cover increase during HS1 coinciding with a warming phase and retreating glaciers in the region (Fig. 11b, e, f) (Barker et al., 2009; Kaiser et al., 2005; Moreno et al., 2015; Palacios et al., 2020). This warming period also coincides with evidences of human presence in the region (Dillehay et al., 2015). Fire activity features a stepwise increase from 13.8 to 12 ka, corresponding to the second deglacial warming phase (Fig. 11a, b, e), under decreasing, albeit still high, precipitation levels (Montade et al., 2019) and drier summers (Moreno, 2020). This threshold in the fire regime was likely induced by warming temperatures and availability of biomass. The positive correlation between woody biomass and fire activity supports a fuel-limited fire regime in the region, and/or suggests that observed increases in tree cover and fire activity were both driven by deglacial warming (Fig. A2f). Nevertheless, peak fire activity in the region is currently achieved at intermediate levels of both woody and herbaceous biomass (Holz et al., 2012), supporting the role of increasing woody vegetation in creating optimal conditions for fire. This is also suggested by the relevant contribution of wood charcoal in lakes from the region (Whitlock et al., 2006). Human populations only started to expand after the ACR and likely also contributed to the intensification of the fire regime (Fig. 11d) (Perez et al., 2016; Salemme and Miotti, 2008). At ca. 13–12 ka, Nothofagus emerges as the predominant taxon and marks the widespread expansion of temperate forests. This expansion exhibited latitudinal variability, with a gradual trend in southern areas and a steeper increase after 15 ka in northern regions (Nanavati et al., 2019). As a result of combined human pressure and environmental changes, the extinction of megaherbivores could have contributed to arboreal growth and fuel accumulation; however, the extent of this impact is still unclear (Abarzúa et al., 2020; Villavicencio et al., 2016).

During the EH, peak fire activity potentially contributed to the deacceleration of tree cover expansion (Fig. 11d, e), which were both favored by a decrease in moisture due to the weakened and poleward shifted Southern Westerly Winds (SWW) (Lamy et al., 2010; Moreno et al., 2021; Nehme et al., 2023) (Fig. 11c) coupled with relatively warmer conditions. The EH climate amelioration also allowed human populations to spread and colonize other localities in the region (Miotti and Salemme, 2003; Perez et al., 2016; Salemme and Miotti, 2008), expanding its contribution to increased fire activity. In the MH, the expansion of forests and decrease in fire activity (Fig. 11e, f) was potentially driven by increased moisture due to a strengthening and equatorward shift of the northern boundary of the SWW (Fig. 1c, d) (Razik et al., 2013; Villa-Martínez et al., 2003) and reduced ENSO (Mark et al., 2022; Rein et al., 2005). Unlike other regions of South America (e.g., TrAn, CEB, NEB), the MH marks a rapid growth in human populations in ExTrAn, which continued through the LH and may have benefited from expanding forests (Fig. 11e) and consequent increase in availability of resources. By 5.5 ka, the region experiences the highest tree cover, whereas fire activity attains peak values after 3.5 ka, concurrent with a major expansion of human populations (Perez et al., 2016) and intensification of ENSO variability (Mark et al., 2022; Rein et al., 2005; Whitlock et al., 2006).

The different changes in vegetation and fire activity observed across the Neotropics highlights the influence of competing and context-dependent drivers that operate on different spatial and temporal scales. In general, low levels of CO_2atm, such as during the LGM, reduce photosynthetic efficiency, mainly in C₃ plants, limiting tree biomass potential growth and favoring C₄ grasses (Boom et al., 2002; Levis et al., 1999; Maksic et al., 2022). A stronger impact of low CO_2atm is expected on vegetation in warm tropical regions (e.g., NNeo, Amazonia, CEB, NEB) compared to cold regions (e.g., ExTrAn, SESA, TrAn) (Sage and Coleman, 2001). Reduced CO_2atm is also suggested by modelling data to weaken the fire regime by altering availability and properties of biomass (Haas et al., 2023), while it can also increase the severity of fires by slowing post-burn tree recovery (Bond et al., 2003).

In sub- and extra-tropical latitudes (ExTrAn, SESA) and high-altitude sites (TrAn), temperatures 3–8 °C lower than present likely constrained biomass growth during LGM, even in locally moist areas (Fig. 12a) (Cruz et al., 2005; Massaferro et al., 2009; Moreno et al., 2018). While cooler temperatures can improve water-use efficiency, frost events can produce long-term plant mortality and shape forest-savanna boundaries (Hoffmann et al., 2019; Inouye, 2000). Weak fire regime over these regions likely resulted from limited biomass in addition to lower temperatures and high moisture levels (Fig. 12g).

In warmer tropical regions (NNeo, Amazonia, CEB, NEB) precipitation seems to have played a pivotal role in the control of vegetation and fire dynamics. In NNeo, high tree cover and low fire activity were sustained by moist conditions (Fig. 12a, d). Strengthened SASM and an east-west precipitation dipole (wet west Amazonia and dry central-east Amazonia and NEB) (Cheng et al., 2013; Cruz et al., 2009; Kukla et al., 2023; Wang et al., 2017), led to decreased tree cover in Amazonian ecotones and in NEB (Fig. 12a, d). In CEB, tree cover and fire patterns were heterogeneous (Fig. 12d) and influenced by both climatic and edaphic conditions. Nevertheless, colder and wetter conditions favored the increase in tree cover and migration of woody taxa across central Brazil (Pinaya et al., 2024). Fire activity remained mostly weak in tropical regions (Fig. 12g), likely constrained by low biomass availability due to reduced CO_2atm and dryness in NEB (fuel-limited conditions) or by persistent moisture in NNeo and western Amazonia (moist-limited conditions).

The deglaciation is marked by progressive warming (Shakun et al., 2012) and rising CO_2atm levels (Bereiter et al., 2015), along with substantial shifts in precipitation linked to millennial-scale events. In the Neotropics, HS1 (18–14.8 ka) and the YD (12.9–11.7 ka) were characterized by wetter conditions in south tropical latitudes (Campos et al., 2019; Meier et al., 2022; Mulitza et al., 2017) and drier conditions in north tropical latitudes (Deplazes et al., 2013; Zular et al., 2019), driven by southward ITCZ shifts. While our analyses can address the long-term trends shaped by these events, the low availability of high temporal resolution and continuous records with robust chronological control hinders a more detailed assessment of site-specific vegetation and fire anomalies over time.

Between 19 and 14.8 ka, Southern Hemisphere tropical regions (TrAn, NEB, SESA, ExTrAn) exhibit increasing tree cover and decreasing fire activity, mostly driven by rising temperatures and enhanced rainfall (Fig. 12e, h) (Campos et al., 2019; Shakun et al., 2012). In contrast, NNeo exhibits the opposite trends (Fig. 12e). In the second part of deglaciation, 14.8–11.7 ka, a generalized increase in tree cover and fire activity (Fig. 12f, i) coincides with further warming (Shakun et al., 2012) and CO_2atm rise (Bereiter et al., 2015). TrAn and the northern parts of ExTrAn indicate an increase in tree cover (Fig. 12f), likely related to the combination of warming and wet conditions (Baker and Fritz, 2015; Montade et al., 2019). Notably, vegetation and fire responses in NEB (wetter) and SESA (drier) may reflect the prevailing precipitation dipole pattern (Campos et al., 2022; Cruz et al., 2009; Wong et al., 2023), although lower temperatures could have still played a key role in limiting tree cover in SESA (Fig. 12c,f). The stepwise intensification in southern hemisphere fire during this period, ca. 14–13 ka, agrees with a worldwide shift in fire regime intensification (Daniau et al., 2012), indicating that warmer conditions, contingent to fuel availability, were critical for this shift.

The greater availability of continuous records spanning the Holocene allows for a more detailed spatial assessment of vegetation and fire dynamics (Fig. 13). The similarities of tree cover and fire activity changes with precipitation shifts suggest that moisture availability became a more important driver of Holocene vegetation and fire dynamics in the whole Neotropics.

Compared to the YD, the EH is characterized by negative tree cover trends in Amazonia, NEB, and CEB, slow tree cover increase in SESA and ExTrAn, where fire activity also increased, and opposite trends in NNeo (Fig. 13a, d, g). These changes are coherent with hydroclimate patterns related to a northward-displaced ITCZ and weaker SASM/SACZ resulting in drier conditions in most southern tropical regions (Cheng et al., 2013; Cruz et al., 2005), relatively drier but still humid northeastern Brazil (Cruz et al., 2009; Venancio et al., 2020), and wetter conditions in northern Neotropical latitudes (Haug et al., 2001). This period also featured a weaker, southward-displaced SWW, but relatively strengthened in its core (ca. 53° S) (Lamy et al., 2010), contributing to drier conditions in most of ExTrAn. Additionally, evidence of increasing human populations in Central and South America (Araujo et al., 2025; Goldberg et al., 2016; MesoRad, 2020) and plant domestication in NNeo, TrAn and Amazonia (Piperno, 2011) suggests that, although still limited, they played an active role as ecosystem engineers, influencing fire regimes and landscape dynamics.

In the MH, tree cover expanded over northern Amazonia, NEB, CEB, SESA, and ExTrAn (Fig. 13b,e). This pattern coincides with gradual precipitation changes driven by a southward expanded migration range of the ITCZ (Chiessi et al., 2021; Haug et al., 2001), a gradual intensification of the SASM/SACZ (Cheng et al., 2013; Prado et al., 2013b; Wong et al., 2023), and a weakening of the Nordeste Low (Cruz et al., 2009). The increase in fire activity over SESA and southwestern and western Amazonia during this period may have also resulted from increasing human activity over these regions (Araujo et al., 2025; Brugger et al., 2016; Lombardo et al., 2020), while in most of the southern Neotropics this period is marked by an occupation hiatus (Araujo et al., 2005, 2025). Although humans have been present in the Neotropics since the late Pleistocene (Goebel et al., 2008), their large-scale influence on tree cover and fire regimes became more pronounced by the end of MH, as consequence of a marked demographic expansion (Gill et al., 2007; Goldberg et al., 2016; Maezumi et al., 2018). This period also featured an intensification of the SWW and a northward migration of its northern boundary (Lamy et al., 2010; Razik et al., 2013) and a reduced strength/frequency of ENSO variability compared to the EH (Koutavas and Joanides, 2012; Polissar et al., 2013). This may have contributed to the ExTrAn moisture and arboreal vegetation increase and fire decrease, while in the Altiplano, centered around 12° S, tree cover reduced and fire intensified amid dry conditions (Figs. 13e, h; A3).

During the LH, tree cover expanded in southern Amazonia, SESA, and part of CEB (Fig. 13c, f). This pattern was facilitated by increased moisture in these regions due to the strengthening of the SASM/SACZ (Cheng et al., 2013; Cruz et al., 2005). While our study did not assess for vegetation compositional changes that might highlight human impacts in these regions (Flantua and Hooghiemstra, 2023), intensified fire activity, despite increased moisture, may point to human influence. In contrast, tree cover declined in NNeo and NEB, where fire activity intensified. These declines were likely a result of both intensified human activity and climatic shifts toward drier conditions. The LH period marks the onset of semi-arid conditions in NEB (Cruz et al., 2009; Chiessi et al., 2021) and precipitation reduction in NNeo (Haug et al., 2001). Meanwhile, in TrAn, tree cover also declined despite moist conditions related to a strengthened SASM. Notably, NNeo and TrAn became densely populated during the LH, and the environmental impacts of human activities specially in these areas likely outweighed those driven by climate alone. In ExTrAn, tree cover remained relatively stable and fire activity increased, despite the rising moisture (Fletcher and Moreno, 2012; Lamy et al., 2010), also suggesting the influence of human activities in this region. In the Neotropics, numerous pollen records indicate human activity, particularly in the last 2 ka (Flantua et al., 2016; Flantua and Hooghiemstra, 2023), in line with our observations.