Trophic rewilding revives biotic resistance to shrub invasion

Trophic rewilding seeks to rehabilitate degraded ecosystems by repopulating them with large animals, thereby reestablishing strong top-down interactions. Yet there are vanishingly few tests of whether such initiatives can restore ecosystem structure and functions, and on what timescales. Here we show that war-induced collapse of large-mammal populations in Mozambique’s Gorongosa National Park exacerbated woody encroachment by the invasive shrub Mimosa pigra—one of the world’s ‘100 worst’ invasive species—and that one decade of concerted trophic rewilding restored this invasion to pre-war baseline levels. Mimosa occurrence increased between 1972 and 2015, a period encompassing the near-extirpation of large herbivores during the Mozambican Civil War. From 2015–2019, mimosa abundance declined as ungulate biomass recovered. DNA metabarcoding revealed that ruminant herbivores fed heavily on mimosa, and experimental exclosures confirmed the causal role of mammalian herbivory in containing shrub encroachment. Our results provide mechanistic evidence that trophic rewilding has rapidly revived biotic resistance to a notorious woody invader, underscoring the potential for restoring ecosystem structure and functions in degraded African protected areas.

Methods
This study was conducted in Gorongosa National Park, Mozambique.

Post-war field data on Mimosa pigra were obtained from plant surveys (2011-2019) and exclosure experiments (2015-2018) conducted in the Urema Lake floodplain. Pre-war data were extracted from Tinley (1977) "Framework of the Gorongosa ecosystem, Mozambique"

Raw diet data from 2013 to 2018 have been obtained by fecal DNA metabarcoding and high-throughput sequencing. The P6 loop of the chloroplast trnL (UAA) intron was amplified using the universal primers “g” (5’-GGGCAATCCTGAGCCAA-3’) and “h” (5’-CCATTGAGTCTCTGCACCTATC-3’) (Taberlet et al. 2007). Amplicons were sequences on Illumina HiSeq 2500 platform (170 bp single-end sequencing in 2013 and 2 x 150 bp paired-end sequencing in 2015-1018). Initial processing steps were performed using the OBITOOLS software (http://metabarcoding.org/obitools) as follows: (i) Direct and reverse reads corresponding to the same sequence were aligned and merged thanks to the ‘illuminapairedend’ program. Only merged sequences with a high alignment quality score were retained (>=40). (ii) Each merged sequence was assigned to its original sample using the tags information previously added to primers using the ‘ngsfilter’ program. Only sequences containing both primers (with a maximum of 2 mismatches per primer) and exact tag sequences were selected. (iii) Strictly identical sequences were merged together while keeping information about the origin of sequences.

Users of the raw dietary data are requested to coordinate with the corresponding author ([e-mail]) to help ensure accurate interpretation and avoid duplication with other research in progress.