Life & Death: Classifying Large Tree Survival Probabilities

Large adult trees (> 5m in height) are important savanna landscape components, providing many ecosystem services including food, habitats, and shade for many species, as well as nutrient recycling. They are indeed iconic in their own rights! However, just like any other living organism, large trees do die, and this can occur at any point throughout their lifespan. In order to become a large adult tree, the individual tree is required to survive several environmental threats at various life stages, including seed predation, seedling herbivory, megaherbivore trampling, intense fires, competition with other tree- and grass- species, as well as potentially harsh climatic conditions. Once established as an adult tree though, the threats to its survival do not end. But which factors contribute to adult tree loss in our savanna systems, and how do we tie them together?

For the past 18 years, Elephants Alive have been monitoring the survival rates of false marula (Lannea schweinfurthii), knobthorn (Senegalia nigrescens), and marula (Sclerocarya birrea) trees across the Associated Private Nature Reserves (APNR). Started in 2004 by Dr. Michelle Henley, these surveys, which comprise over 3,000 trees, have allowed us to track individual tree loss through various conditions. Why is this important? Too often we view savannas as static, where the same factors are at play throughout the system. We thus standardise the cause and effect across the system. However, savanna systems are far more dynamic and complex than what was once thought, with environmental pressures working in ebbs and flows. So, what does this mean for our individual tree? Well, if we measure tree survival in relation to factors such as elephant impact, elephant density, fire history, mean annual rainfall, and the presence or absence of tree protection methods (such as wire-netting), we can start to piece together the fuller story.?

Using a special algorithm, combined with years of survey data, we can now make probability predictions! For example, we can now predict that wire-netting, if properly applied, can double a tree’s survival chances, versus trees with no wire-netting. That sounds great on its own. But wait - add an intense fire to the equation and the wire-netted tree’s survival chances will decrease six-fold! We can also see that for trees with severe elephant impact, false marulas have a significantly higher chance of surviving for another five years or longer in compassion to knobthorns and marulas. That means we can now statistically predict a tree’s survival chances for the foreseeable future once it has received a specific level of elephant impact. And lastly, when modelling elephant density as a factor versus tree mortality, we find that the relationship is complex. Numbers alone do not equate to treefall, rather, it is how elephants make use of the landscape on a temporal and spatial scale, which ultimately affects tree survival probabilities. For example, an individual marula tree in a low rainfall region, surrounded by mopaneveld, is statistically more likely to receive elephant impact in comparison to marula trees in high rainfall, marula/knobthorn dominated vegetation zones. We can thus see how a tree’s survival chances can vary across the APNR as the presence of (and variation within) various factors strengthen or weaken.???

These results highlight the importance of long-term surveys for both analysing and predicting tree survival probabilities. When various factors and their interactions with elephant impact can be analysed through algorithms on large datasets over multiple years, we gain a stronger understanding of which combinations are strong predictors of tree loss. These results can then aid management in the decision-making process going forward.?

These results form a part of the PhD degree of Robin Cook at the University of the Witwatersrand, under the supervision of Dr. Michelle Henley and Prof. Ed Witkowski. Full results will be made available upon the completion of the degree.?