The Amazon-Cerrado transition zone is a significant agricultural frontier experiencing rapid land-use change, primarily driven by cattle ranching and crop intensification. This conversion of native forests and savannas fundamentally alters fire regimes, with deforestation-related fires accounting for a substantial portion of burned area. Management practices in pasturelands also contribute significantly to fire activity. The distinct ecological dynamics of the Amazon and Cerrado biomes influence their responses to fire; in the Amazon, fire suppression in agricultural fields reduces escaped fires, while in the fire-adapted Cerrado, fire suppression can lead to woody encroachment and more intense fires. Current understanding of fire activity changes with the evolution of this agricultural frontier is limited, hindering the development of effective conservation strategies. This study addresses this gap by modeling fire probability as a function of time since land clearing to understand how frontier age and climate drivers influence fire activity across the Amazon-Cerrado transition.

Existing research indicates reduced fire activity due to agricultural intensification, but opposing patterns are expected in the Amazon and Cerrado. Studies highlight the unintended ecological consequences of agricultural intensification, including landscape fragmentation and the increased risk of fire ignition even with sparse ignitions. A 'zero-fire' policy can be detrimental in fire-dependent systems like the Cerrado, leading to fuel accumulation and intense fires during droughts. The Amazon-Cerrado transition also faces added pressure from climate change, with longer, drier dry seasons increasing fire risk. Climate projections suggest that land-use contraction alone may be insufficient to reduce future Amazon understory fires, emphasizing the need to consider interactions between land use and climate change mitigation. This study builds on this research by comprehensively analyzing lagged fire responses to different land-use transitions at the interface of these biomes.

This study used time series data from MapBiomas Project datasets (MapBiomas 6.0 and MapBiomas Fire 1.0) from 1986 to 2020 to analyze land use and burned areas. Data was regridded from 30m to 500m resolution for computational efficiency. The study modeled the proportion of burned area associated with different land-use transitions (forest to pasture, savanna to pasture, grassland to pasture, forest to cropland, savanna to cropland, grassland to cropland, and pasture to cropland) as a function of time since land clearing (frontier age). The time interval between land conversion and burned area served as a proxy for the relationship between frontier age and fire activity. The analysis also incorporated climate drivers, estimated vapor pressure deficit (VPD) and maximum cumulative water deficit (MCWD), to assess their combined influence on fire activity. Generalized Additive Models (GAMs) were used to model the relationship between fire probability and frontier age, capturing non-linear relationships. Segmented regression was used to identify shifts in fire activity associated with frontier aging. Spatial patterns of fire probability rate before and after conversion were analyzed using linear regression. Finally, the study examined the combined effect of frontier age, drought, and air dryness on fire probability by analyzing different subsets grouped by annual burned area extents.

Large-scale conversion of native vegetation to agriculture occurred during the study period, with forest replacement by pasture being the most prevalent transition. Land-use transitions significantly influenced burned area probability, with the majority of burned areas associated with forest conversion to pasture or croplands. Burned area tended to decrease after 2004, but recent increases were observed in some regions due to pasture replacement by cropland. Analysis of fire probability associated with frontier age revealed contrasting patterns between forest conversion to pasture and transitions to croplands. Forest to pasture conversion showed an initial increase in burned area before land clearing, which remained elevated for several years post-transition, especially in the Amazon (45% fire probability at conversion, elevated for at least 8 years after). In contrast, the transition to cropland resulted in generally lower fire probabilities at conversion and faster elimination of fire from the landscape. Spatial analysis showed contrasting trends in fire probability between the Amazon and Cerrado. The Amazon consistently exhibited higher fire probabilities at the time of transition, particularly for forest conversion to pasture. Transitions to cropland exhibited less noticeable shifts in fire activity, due to declining burning before conversion. Analysis of the 35-year period divided into 7-year intervals revealed an overall non-stationarity in fire spikes during forest to pasture conversion, with recent decreases in fire spike intensity. This suggests a broader trend reflecting changes in clearing rates, in addition to frontier aging. Transitions from pasture to cropland displayed lower magnitudes of fire probability spikes. A strong association was found between deforestation, drought, and air dryness with the occurrence of larger extents of burned area. In transitions from pasture to cropland, the influence of drought and air dryness on fire probability was negligible.

This study demonstrates the complex interplay between land-use transitions, frontier age, and climate drivers in shaping fire regimes across the Amazon-Cerrado transition. While agricultural intensification reduces fire probability, deforestation-related fires remain a significant concern, particularly in the Amazon. The significant fire activity preceding and following deforestation highlights the need to address pre-conversion fires. The faster elimination of fire in transitions to cropland, compared to pasture, underscores the role of agricultural intensification in fire suppression, although the ecological long-term consequences are yet to be determined. The study’s results emphasize the need to incorporate frontier age into fire models to better understand and manage fire risk. The findings reinforce the critical importance of preventing deforestation and protecting native vegetation to preserve these important ecosystems.

This study reveals that time since land-use transition (frontier age) is a key driver of fire activity in the Amazon-Cerrado region. While agricultural intensification reduces fire probability, deforestation remains a major contributor to fire risk, particularly when coupled with drought. Incorporating frontier age into fire models is crucial, and preventing deforestation is paramount to preserving the region’s forests.

The accuracy of the MapBiomas Fire 1.0 dataset might introduce uncertainty, particularly regarding understory forest fires and cropland fires. The study focuses primarily on the Amazon-Cerrado transition, and findings might not be directly generalizable to other regions. The analysis relies on observational data and does not explicitly model the mechanistic processes driving fire ignitions.