Tree mode of death provides insights as to why soil and topography explain only about 25% of the spatial variation in tree mortality in Central Amazonia, and permit predictions about what types of mortality are most probable under climate change. We studied tree mortality by mode of death in 72 1-ha permanent plots spanning 64 km2 of tropical moist forest in Reserva Ducke, Manaus, Amazonas,
... [Show full abstract] Brazil. Plots were re-censused twice (2003–2005 and 2005–2008). Tree mode of death was assigned for trees ⩾4 cm dbh as standing, uprooted or snapped. We also recorded whether trees died alone or were pushed over by treefalls. Standing death was predominant, representing 54% of deaths of trees with dbh ⩾ 10 cm, followed by snapping (26%) and uprooting (14%). Trees that fell alone represented 25% of deaths, while 16% were pushed over. Most small fallen dead trees (4 ⩽ dbh < 30 cm) were pushed over by other trees, while most large dead trees (dbh ⩾ 30 cm) died alone. Standing mortality was weakly related to soil and topography, but 20% of variation in uprooted mortality and 11% in snapped mortality of trees with dbh ⩾ 10 cm was explained by soil and topography. The variation in mortality explained for small trees (18% for uprooted mortality and 13% for snapped mortality) was higher than for large trees (14% for mortality by snapping only). In spite of little variation in mortality associated directly with soil and slope, analyses assessing the effect of topographic categories (plateaux, slope, and valley) on tree mortality detected higher differences, even though causal factors remain unidentified because topographic position may encompass both topographic and soil properties. There was an increase from the first to the second census interval in the effects associated with soil and topography on tree mortality by uprooting and snapping, and this was likely due to storms, which led to a disproportional increase in tree mortality for these tree modes of death. Presently, uprooting and snapping mortality are not dominant and the use of soil and topographic variables for modeling of tree mortality is therefore limited. However, under predicted climate-change scenarios of higher frequency of extreme storms, soil and topography may become more useful to improve estimates of tree mortality and biomass losses over large areas in Amazonia.