Long‐term vegetation dynamics mediated by herbivores, weather and fire in a Juniperus‐Quercus savanna

Abstract. Long‐term (45‐yr) basal area dynamics of dominant graminoid species were analyzed across three grazing intensity treatments (heavily grazed, moderately grazed and ungrazed) at the Texas A&M University Agricultural Research Station on the Edwards Plateau, Texas. Grazing intensity was identified as the primary influence on long‐term variations in species composition. Periodic weather events, including a severe drought (1951–1956), had little direct influence on composition dynamics. However, the drought interacted with grazing intensity in the heavily grazed treatment to exacerbate directional changes caused by grazing intensity. Species response to grazing was individualistic and noisy. Three response groups were identified. Taller, more productive mid‐grasses were most abundant under moderate or no grazing. Short grasses were most abundant under heavy grazing. Intermediate species were most abundant under moderate grazing and opportunistic to weather patterns. Graminoid diversity increased with the removal or reduction of grazing intensity. The moderately and ungrazed treatments appeared most resistant to short‐term weather fluctuations, while the heavily grazed treatment demonstrated significant resilience when grazing intensity was reduced after over 110 yr of overgrazing. Identification of a ‘climax’ state is difficult. Significant directional change, which took nearly 20 yr, appears to continue in the ungrazed treatment after 45 yr of succession. The observed, relatively linear patterns of perennial grass composition within the herbaceous patches of this savanna were generally explained by traditional Clementsian succession. However, when dynamics of the herbaceous community are combined with the woody component of this savanna, the frequency and intensity of fire becomes more important. Across the landscape, successional changes follow several pathways. When vegetation change is influenced by several factors, a multi‐scale model is necessary to demonstrate interactions and feedbacks and accurately describe successional patterns. Absence of fires, with or without grazing, leads ultimately to a Juniperus/Quercus woodland with grazing intensity primarily influencing the fuel load and hence fire intensity.