ABSTRACT : The forest - savanna transition is the most widespread ecotone in the tropical regions and with important ecological, climatic, and biogeochemical implications at local to global scales. However, the processes and mechanisms that control this transition vary among regions and remain not fully understood in all of them. In general, this transition is influenced by interactions between vegetation and environmental factors such as climate, soil properties, fire, and herbivory. However, the importance of these effects can vary substantially across continents, which can result in different responses to environmental change. For this reason, more regional studies are needed to describe and understand the factors and interactions that control forest - savanna transition in different regions. Using remotely-sensed data, we examined the relationship between the tropical forest-savanna transition and several environmental factors in northern South America, in the Llanos ecoregion. We used several vegetation structure metrics, as well as multiple precipitation statistics, soil properties, and a fire regime descriptor. In addition, we developed a statistical analysis on the interactive effects of soil silt content, fire frequency as well as three dry season precipitation variability components (season length, wet day frequency, and precipitation intensity) on the forest - savanna transition, using tree cover, canopy cover and PAVDmax as indicator variables that differentiate forest from savanna. Our results show that savannas in the Llanos region occur in mean annual precipitation (MAP) levels in which forest would be predicted based on previously proposed thresholds to other savanna regions. Our results also highlight that the MAP range in which both forest and savanna can occur in our study area correspond, almost exclusively, to forest in other South American regions and globally. Although both forest and savanna can also occur in a large interval of intermediate values of dry season precipitation variability (PV) components, forest dominates in areas with higher precipitation frequency and intensity than savanna. Savanna tends to occur in pixels where fires are present, while fires are absent in forest. However, a large proportion of pixels classified as savanna pixels have no fires in the analysis period, even those that occur in the same climatic or edaphic space of forest. Finally, our analysis shows that fire frequency and dry season precipitation are the most important variables to predict the forest-savanna transition. This highlights the role of fire regime and water availability in determining the limits between forest and the second largest area of savanna in South America. Further, our results support the importance of refining our understanding of the factors, relationships, and mechanisms that control forest-savanna transition at regional scales, as a requirement to assess the effects of environmental change on tropical forest and savanna distribution.