Evaluating the ecogeographical effects of earth's largest terrestrial herbivore

Savannas cover a fifth of Earth's land surface, are home to over a half billion people, and disproportionately affect interannual variability of the global carbon cycle. In Africa, these open, grassy and sparsely wooded habitats support pastoralist cultures, the world's largest array of megafauna and thriving tourist economies. However, savannas and their uses are under threat: woody plant encroachment linked to increases in atmospheric CO2 concentrations is reducing grassy cover required by both domestic livestock and wildlife, while also encroaching on the open views of wildlife critical to tourism. Yet, understanding of the determinants of savanna woody cover (bushes and trees) is limited. To this end, a growing number of site-specific studies have found that tree mortality rates in protected areas are principally controlled by African savanna elephants (Loxodonta africana). However, it is not known whether these impacts significantly affect the total woody cover of the larger landscapes and region. This dissertation focuses on quantifying the relationship between elephant densities and savanna woody cover in protected areas across elephants' Eastern African range. Research questions are addressed in three self-contained chapters. Chapter 1 tests multiple approaches to mapping savanna woody cover fractions across 12 protected areas (PAs) using Landsat imagery, and presents a novel approach to reference data generation. The results show a machine learning approach, Random Forests, produces the most accurate maps and demonstrates that accurate maps do not require more than a single annual image - which is advantageous given the general image scarcity in these areas. In Chapter 2, the most accurate mapping approach from Chapter 1 is used to produce over 30 years of savanna woody cover data. These data are then used to assess whether there is a relationship between woody cover and elephant densities across the 12 PAs, as well as for specific landscapes within the PAs. Results point toward climate, principally wet season precipitation, being the major determinant of woody cover across the PAs (R2 = 0.38, p = 0.03), though elephants were linked to increased woody cover on hill slopes far from permanent water bodies (R2 = 0.41, p = 0.03). In addition, areas near water contain consistently low levels of woody cover unexplained by any of the variables considered. Last, Chapter 3 presents a meta-analysis of the literature comparing the relative impacts of elephants and fire on woody cover. The majority of studies (80.3%) find elephants to be the primary disturbance, with the relative dominance of the two disturbances linked to climate. The coolest and wettest savannas are more likely to be dominated by fire, while elephants are most likely to dominate across a comparatively broad set of environmental conditions. Overall, the evidence presented here suggests (1) both overall woody cover and the relative impacts of elephants and fire are principally regulated by climate; (2) elephants, perhaps through the dispersal of seeds and nutrients, increase woody cover on hillslopes far from permanent water bodies; (3) areas near water are in a long-term state of low woody cover, potentially driven by disturbances, and (4) given the dominant role of elephants as a disturbance in the majority of sites and climates, conservationists should consider increasing elephant populations as a means of mitigating the woody encroachment threatening Africa's savannas, wildlife and pastoralist cultures.