At high stocking rates, cattle do not functionally replace wild herbivores in shaping understory community composition

The KLEE plots, established in 1995, use barriers to control access to 200 × 200 m (4-ha) treatment plots by three herbivore guilds – wild megaherbivores (‘M’, elephant and giraffe), wild mesoherbivores (‘W’, 50-1000 kg) and cattle (‘C’) – in different combinations. There are three replicate blocks, each consisting of six treatments (18 plots in total): 1) ‘MWC’ (accessed by megaherbivores, wild mesoherbivores and cattle), 2) ‘MW’ (accessed by megaherbivores and wild mesoherbivores), 3) ‘WC’ (accessed by wild mesoherbivores and cattle), 4) ‘W’ (accessed by wild mesoherbivores only), 5) ‘C’ (accessed by cattle only), 6) ‘O’ (excludes cattle, wild mesoherbivores and megaherbivores). The treatment plots accessible to cattle are typically grazed by 100-120 mature Boran cows <i>Bos indicus</i> (sometimes with calves and/or bulls) for 2-3 days (2 hrs day^-1) within a 2-week period, 3-4 times per year. The timing and number of grazing days depends on forage availability and reflects typical grazing regimes of ranches in the region, wherein cattle graze in an area for several days before being moved to allow that area to recover. Each of the treatment plots accessible to cattle (MWC, WC, C) contains a 50 × 50 m (0.25-ha, or 1/16 of the plot) subplot (established in 2008), in which the same cattle herd is grazed for a further 30 mins following the initial 2-hr grazing period in the wider plot, to achieve an approximately five-fold increase in cattle stocking rate compared to the wider plot (Appendix S1: Figs. S2–3). These three additional treatments are named: 1) MWCh, 2) WCh, and 3) Ch, where ‘h’ denotes high cattle stocking rate. We note that ‘grazing’ also involves trampling and nutrient cycling effects (Sitters et al., 2020). Grazing behaviour can also be affected by time of day and the presence of other herbivores (Odadi et al., 2017). Because cattle only access individual plots a few times per year, responses of most large wild herbivores are unlikely to be due to direct interaction with cattle or herders. Fire has not been used as a management tool in this ecosystem for over 50 years and is rarely used by other ranches in the region. Natural-ignition fires have not occurred in decades. See Young et al. (1997) and Young et al. (2018) for further details of the experimental design. To assess understory vegetation, we sampled herbaceous plants during May-August in 2018, 2019, and 2020. We measured aerial cover every 10 m by counting the number of pins of a 10-point pin frame (vertical pins separated by 5 cm) hit by each species (maximum one hit per pin per species). For the main six treatments (O, C, W, WC, MW, MWC), we sampled 10 100-m transects within the central hectare of the 18 4-ha treatment plots, recording pin hits every 20 m for a total of 50 sites. We further subsampled pin hits and leaf verses stem hits for the five dominant species (<i>Brachiaria lachnantha</i>, <i>Themeda triandra</i>, <i>Pennisetum stramineum</i>, <i>P. mezianum</i>, <i>Lintonia nutans</i>) by sampling four of 10 transects (second, fourth, sixth and eighth transects) totalling 20 sites. For the 50 × 50 m high cattle stocking rate subplots (Ch, WCh, MWCh), we ran four 40-m transects (leaving a 10-m buffer along two sides to minimize edge effects of the 4-ha plot), recording pin hits every 10 m for a total of 16 sites. Leaf versus stem pin hits were only recorded at eight sites (second and fourth transects).