Abstract. At a landscape scale, the combined influence of biotic and abiotic factors may determine the distribution patterns of large herbivores in African savanna ecosystems. Herbivores foraging in these ecosystems may become nutritionally stressed during an annual dry season when both forage quality and quantity are reduced. Additionally, the locations of water sources may impose a landscape-scale constraint on dry-season herbivore distributions. We used logistic regression to analyze 13 years of aerial census data collected in the Kruger National Park (KNP), South Africa, and evaluated hypotheses regarding the relative influences that surface water, forage quality, and forage quantity exert on the dryseason, landscape-scale distribution patterns of eight herbivore species. Hypotheses regarding the degree of correlation between species' distributions and distance to water were developed using previous observations of species' relative water dependence. We also developed hypotheses regarding species' responses to the trade-off that may occur between surface-water constraints and nutritional requirements when either forage quality or quantity is reduced. In general, we expect an increase in species' mean distance to water as a result of individuals mitigating limitations in nutritional requirements (i.e., intake quality or quantity) by foraging farther from water. Our analyses suggest that the trade-off between nutritional requirements and surface-water constraints that species face varies according to the species' water dependence, size, and gut morphology. Of the four grazers considered in our analyses, waterbuck distributions appear to be constrained primarily by surfacewater availability. Distributions of buffalo, a large ruminant grazer, suggest that individuals face a trade-off between nutritional requirements and surface-water constraints when forage quantity is reduced. Alternatively, distributions of wildebeest, a smaller ruminant grazer, suggest that individuals face this trade-off when access to high-quality forage is limited. In comparison to buffalo and wildebeest, the strength of this trade-off is moderate for zebra, a nonruminant similar in size to wildebeest, when either forage quality or quantity is reduced. Distribution patterns for browsers are characterized by a weak relationship with distance to water, as expected for these relatively water-independent species. Population densities relative to forage quality confound exploration of this trade-off for mixed feeders.