Elephant population studies have become important especially because of the long standing perception that high elephant densities have negative impact on vegetation and other wildlife species. Thus, in areas of high elephant density, managers attempt to re-distribute them or keep their numbers low through provision of water, translocation or culling. These approaches are thought to keep the population within the limits that can be sustained by the ecosystem, termed “the ecological carrying capacity”, a management option hinged on equilibrium theory. Equilibrium systems are considered stable, with resources and the animals that depend on them being at balance with each other. This stability is rarely the case in tropical savannas where the rule appears to be “a flux of nature” rather than “a balance of nature”. Tropical savannas, where over half of the African elephant live, are prone to constant environmental fluctuations, especially prolonged droughts, and hence there is a growing understanding that populations of wildlife species and their communities are rarely at equilibrium. Therefore, it is critical to understand how the constant environmental flux in this system affects wildlife populations and the implication for their management. In this thesis, the central focus is to investigate the role of drought occurrences on elephant population dynamics in tropical savannas. To address this question, it is important to have a good understanding of the historical changes of elephant population in relation to drought events and the ecology of elephant in semi-arid savannas - their distribution and density, their movements and behaviour. For the historical data, I analysed the best existing long-term data in Africa of wild elephant population that has been consistently monitored for over 40 years where life histories of over 3000 wild individual elephant are known, at Amboseli National Park in Kenya. In addition, I also analysed geo-referenced elephant mortality data collected daily for 10 years from Tsavo Conservation Area. Further, I analysed 2 years data from 8 GPS collared African elephant to investigate their movement response to seasonal water and forage distribution in Tsavo Ecosystem. First, I investigated the temporal effects of drought duration (number of consecutive dry months) and intensity (amount of rainfall) on elephant population structure in Amboseli National Park, Kenya. The result corroborates findings from past studies that calves (<2years) are more susceptible to drought caused mortality and the risk of dying decreased with age. A new finding in this study reveals that the effect of drought induced mortality for the adult elephant is sex and age dependent, with males older than 25 years being less likely to die as compared to females of the same age. This new result is because of the resolution of analysis in this study which focused on the length and severity of drought as opposed to past studies that restricted their analysis to seasonal and inter-annual differences in rainfall pattern. As they grow older and sexually mature, the foraging range of male elephant increase and they begin to take more risks and disperse to unfamiliar habitats to seek for quality forage and mates. Generally, foraging strategies between sexes in many species are more pronounced during periods of food scarcity, and the driving force in the differences appears to be driven by energy need requirements, reproductive status of an individual, body sizes and the social context, all of which differ between sexes. In the next study, I investigated the spatial pattern of elephant mortality in relation to drought occurrences in Tsavo National Park using MaxEnt. The results shows that elephant carcasses were aggregated and elephant mortality was negatively correlated with four months cumulative precipitation prior to death, forage availability and distance to water, while local elephant density showed a positive correlation. This finding rules out dehydration as the cause of elephant mortality in Tsavo as the river where the carcasses were aggregated is perennial. Furthermore, forage availability was low close to water sources and did not show a significant difference close to or further away from the river despite high elephant density around the river. Hence, these elephant mortalities may have occurred as a result of starvation. I went further to focus on two main limiting resources for elephants, namely forage and water, and their effect on elephant-habitat utilization in semi-arid savannas. I first investigated how water source distribution affect elephants’ seasonal movement patterns. Results indicate that male elephant moved maximally 20 km away from the nearest water source in the dry season while the female elephant foraged to a maximum of about 10 km and only moved further than this distances in the wet season. The strong directionality of elephant movement from a distance of 15km towards water sources (rho > 0.5) as they re-visited their watering source in the dry season suggest that elephant have information on location of the water sources. Next, I investigated the factors that determine selection of a foraging site for elephant with a focus on forage nutrients or biomass. Because of their large body size, it is thought that elephant can survive on a less nutritious but high biomass of forage. The results from this study shows that elephant selected foraging site based on forage biomass in dry seasons, whereas they selected areas with higher nutrients in the wet season. Moreover, females selected sites with a higher forage biomass as compared to males. This result may be explained by the difference in social organisation and foraging strategies between the sexes. In the previous studies on human-elephant conflict, for instance, male elephant raided crops more than the mixed herd, perhaps to seek for high quality forage. Together, the four studies in this thesis strongly suggest that elephant starve to death in prolonged drought contrary to the past studies that reported that adult elephant are less affected by drought. Even though prolonged droughts usually result in higher elephant mortalities, the resilience of semi-arid savannas may perhaps be as a result of these deaths that release the system from high browsing pressure and give it a window to regenerate. If that is the case, then drought induced elephant mortality may be a better way to regulate elephant numbers than culling. This finding strongly suggests that semi-arid savannas may in fact be a non-equilibrium system sustained by growth and crashes of herbivore populations. Maintaining the system as natural as possible may therefore keep elephant populations in savannas sustained for posterity. The modern day park managers have daunting challenges such as mass elephant deaths in drought, increased human-wildlife conflicts or changes in wildlife use of the landscape which may all be symptoms of wrong management interventions taken in the past or negative impacts of anthropogenic activities that have tipped the natural functioning of a non-equilibrium system. Therefore, park managers should undergo regular trainings on new conservation techniques and they should apply evidence-based science to make informed long term decision.