1. Woody encroachment in southern African savanna has been partly attributed to rising atmospheric CO2 fertilising the growth of C3 trees but less so that of competing C4 grasses. However, growth conditions (resource availability, competition, rooting space, and herbivory) must be suitable for the effects of elevated CO2 (eCO2) to be realised.

2. This research investigated the interactions between the positive effect of eCO2 on tree seedling growth and limitations imposed by drought, disturbance, and competition with C4 grasses. Seedlings of the prolific encroacher C3 tree Vachellia karroo were grown at ambient (400 ppm) or eCO2 (800 ppm) in Open-Top Chambers and exposed to a variety of stresses and disturbances typical of savanna systems. Photosynthetic, growth and allocation responses to eCO2 and other treatments were determined.

3. Unsurprisingly, we show strong growth and water-saving responses of V. karroo seedlings to eCO2 when in the absence of competition and herbivory. However, the addition of either competition or simulated herbivory in the first season of growth moderated this, while neither drought nor shading diminished the eCO2 effect relative to similarly treated plants grown at ambient [CO2].

4. Synthesis. We demonstrate that eCO2-induced C3 stimulation in encroaching savanna species such as V. karroo will be inconsistent across time and space. This research does not detract from the suggestion that increasing atmospheric CO2 is implicated in woody encroachment, but rather that eCO2 benefits to C3 tree seedlings are only realised when growth conditions are suitable. Inconsistencies in eCO2 response will translate into spatial and temporal variation in seedling responses to eCO2 and CO2-driven woody encroachment, explaining some of the variability observed in woody encroachment across geographic regions and disturbance gradients.

Data for all experiments are combined into one dataset. This means that it will appear that there are missing values but this is just because of joining different datasets, so look for data of the interaction desired. Please feel free to contact me if you would like data that was not included in this manuscipt but is likely to have been collected.