Tree populations in mesic (>650 mm precipitation/yr) savannas of the world have strong demographic bottlenecks to the transition of subadult trees to the canopy layer. Although such bottlenecks are a major determinant of savanna physiognomy, the factors that allow subadults to traverse the bottleneck are little studied. In a landscape‐scale field experiment in a northern Australia savanna, we determined the survival and growth of 1506 permanently marked juveniles (<150 cm tall) and saplings (150–599 cm tall) of canopy species in response to season of fire (early dry season, late dry season, wet season, and unburned), and understory type (herbaceous forbs vs. sorghum [native annual grass]) that differ in seasonal growth patterns and competitive regimes. Trees were assessed before fires and at the end of the following growing season, without repeat fires. We used Akaike‐information‐criterion‐based model selection and multi‐model inference for data analyses. Initial height was an important explanatory variable for all responses except genet mortality wherein fire season was important for juveniles and understory type for saplings. Fire season was important to height growth of large juveniles and small saplings (enhanced the year following dry‐season fires). Fire season × understory interactions were important for height growth of small juveniles and for the proportion of juveniles transitioning to saplings. Changes in stem numbers were affected by all explanatory variables. All fires topkilled most juveniles (fewer in early dry‐season fires in herbaceous understory), but genet death was rare. Late dry‐season fires topkilled most saplings; they failed to regain previous height and some died the following year. Given no further fires, persistent large juveniles can grow to sapling size within a year; whereas sapling success is severely hampered by late‐dry‐season fires, especially in grassy understory. Differences in seasonal phenological patterns of both understory vegetation and trees that vary with size and life history stage are among suggested explanatory mechanisms. Weighted averaged model coefficients for all responses to the explanatory variables are provided for use in population dynamics models. A conceptual framework links landscape‐scale variables to tree attributes and responses, with implications for population ecology and community assembly.