Widespread woody encroachment is a prominent concern for savanna systems as it is often accompanied by losses in productivity and biodiversity. Extensive ecosystem-level work has advanced our understanding of its causes and consequences. However, there is still debate over whether local management can override regional and global drivers of woody encroachment, and it remains largely unknown how encroachment influences woody community assemblages. Here, we examined species-level changes in woody plant distributions and size structure from the late 1980s to the late 2000s based on spatially intensive ground-based surveys across Kruger National Park, South Africa. This study region spans broad gradients in rainfall, soil texture, fire frequency, elephant density, and other topographic variables. Species-level changes in frequency of occurrence and size class proportion reflected widespread woody encroachment primarily by Dichrostachys cinerea and Combretum apiculatum, and a loss of large trees mostly of Sclerocarya birrea and Acacia nigrescens. Environmental variables determining woody species distributions across Kruger varied among species but did not change substantially between two sampling times, indicating that woody encroachers were thickening within their existing ranges. Overall, more areas across Kruger were found to have an increased number of common woody species through time, which indicated an increase in stem density. These areas were generally associated with decreasing fire frequency and rainfall but increasing elephant density. Our results suggest that woody encroachment is a widespread but highly variable trend across landscapes in Kruger National Park and potentially reflects an erosion of local heterogeneity in woody community assemblages. Many savanna managers, including in Kruger, aim to manage for heterogeneity in order to promote biodiversity, where homogenization of vegetation structure counters this specific goal. Increasing fire frequency has some potential as a local intervention. However, many common species increased in commonness even under near-constant disturbance conditions, which likely limits the potential for managing woody encroachment in the face of drivers beyond the scope of local control. Regular field sampling coupled with targeted fire management will enable more accurate monitoring of the rate of encroachment intensification.

This data was collected from Kruger National Park (22° 20' to 25° 30' S; 31° 10' to 32° 00' E). Woody species distribution data for 1989 was collected using a rapid Braun-Blanquet method. Briefly, 1794 sites were selected across the park. At each site, woody vegetation cover and species were examined in a 20 m × 20 m plot according to three height classes: small (0.75–2.5 m), medium (2.5–5 m), and large (> 5 m). These records were stored as presence and absence for each woody species in each size class at each site. Woody species distribution data for 2008 were collected from 342 sites across the park. At each site, a 50 m × 60 m plot was established and four 50-m transects running at 0, 20, 40, and 60 m along the length of the plot were laid out. Woody plant data was collected from eight sampling points located along the transects: 30 m and 50 m along the first and third transects, and 20 m and 40 m along the second and fourth transects. At each sampling point, woody plants were sampled according to their height classes (≤ 1 m, > 1 and ≤ 3 m, and > 3 m) within three corresponding circles with a radius of 1 m, 2 m, and 5 m, respectively.  In other words, woody plants with a height of 1 m or less that were found within a 1-m radius of each sampling point were counted and identified to species; woody plants within a 2-m radius with a height of > 1 m up to 3 m were measured in height and identified to species; and woody plants within a 5-m radius and taller than 3 m were measured in height and identified to species.

This data is comprised of four components: (1) data and code for species occurrence and size distribution analysis; (2) data and code for species distribution modeling; (3) data and code for linear regression modeling; and (4) predicted maps of species distribution. Please refer to the paper for more details.