Irruptions in plant and animal populations are not uncommon in plants and animals. It has been suggested that these irruptions may be reduced by predators or herbivores, but there is a paucity of controlled experimental evidence. In addition, the factors underlying irruptions are rarely explored quantitatively. Using data from the Kenya Long-term Exclosure Experiment (KLEE), we show that populations of perennial Hibiscus spp. (primarily H. flavifolius) show multiple short-term irruptions a year after rainy periods, increasing in abundance in some cases by more than an order of magnitude before declining in ensuing months and years. We demonstrate that these irruptions are largely limited to experimental plots from which large mammalian herbivores have been excluded, particularly megaherbivores (elephants and giraffes). This represents the only controlled replicated experimental demonstration of top-down regulation of irruptions in either plants or animals. African elephants and giraffes are often at greater risk of local extinction than other large mammals, and their absence appears to destabilize this African savanna ecosystem, providing additional support for their conservation.

This research was carried out from 2007 to 2023 in the Kenya Long-term Exclosure Experiment (KLEE), located on the property of the Mpala Research Centre (0°17’ N, 36°52’ E; 1,800 m asl) in Laikipia, Kenya. Annual rainfall at the site is highly variable interannually, and during the study period averaged 597 + 196 mm/year (mean + SD; range 330–1,009 mm yr-1; Fig. 1c). The rainfall pattern is weakly trimodal, peaking in March–May. The study site is a semi-arid savanna ecosystem underlain with black cotton soils and dominated by the tree Acacia [Vachellia] drepanolobium and five grass species: Pennisetum mezianum Leeke, P. stramineum Peter, Themeda triandra Forssk., Lintonia nutans Stapf., and Brachiaria lachnantha (Hochst.) Stapf. (Young  et al. 1997). Approximately 100 other species of plants occur at lower numbers in the plots, including several Hibiscus species. Mpala is managed for both livestock production and wildlife conservation and supports a diverse variety of herbivores.

The KLEE experiment was established in 1995 and uses semi-permeable barriers to create six 200 × 200 m herbivore exclosure treatments replicated in three blocks, for a total of 18 plots. Treatments and abbreviations are as follows: O = all large (> 20 kg) herbivores excluded, W = wild mesoherbivores present, C = cattle present (Bos indicus L.), WC = cattle and wild mesoherbivores present, MW = megaherbivores and wild mesoherbivores present, MWC = all herbivores present (Young et al. 1997). In this ecosystem the megaherbivores are elephants (Loxodonta africana Blumenbach) and giraffes (Giraffa camelopardalis L.). The mesoherbivores are represented by plains zebras (Equus quagga Boddaert), Grevy's zebras (E. grevyi Oustelet), elands (Tragelaphus oryx Pallas), Grant's gazelles (Nanger [Gazella] granti Brooke), cape buffaloes (Syncerus caffer Sparrman), oryx (*Oryx beisa *Rüppell), and hartebeests (Alcelaphus buselaphus Pallas). Non-excluded mammalian herbivores (< 25 kg) include steinbucks (Raphicerus campestris Thunberg), bush duikers (Sylvicapra grimmia L.), hares (Lepus spp.), and several rodent species. 

Hibiscus spp. surveys were carried out in June of each year from 2007 to 2023 (approximately 2–3 months after peak rainfall). Two species are represented in this data set: H. flavifolius Ulbr. and H. aponeurus Sprague & Hutch, distinguishable only in flower, and are lumped here. Both are single- to few-stemmed, semi-woody perennial herbs that can attain heights of 2–3 m. Based on proportions of these two species flowering in the area (pers. observ.), the great majority of sampled plants were H. flavifolius. In each of the 18 KLEE plots, we counted the number and measured the heights of all these Hibiscus spp. plants in 50 1 × 1 m quadrats, located every 20 m along 10 parallel transects.