Study area and vegetation
The study site is located in the Bou Hedma National Park (348°390 N, 9°480 E, southern Tunisia). The park covers an area of 5115 ha. Climate is Mediterranean lower arid in the nomenclature of Emberger (Le Houérou 1959). Mean annual rainfall is 180 mm, with the lowest probability of rain during summer and winter seasons, but with important inter-annual variations. Mean annual temperature is 17.3°C, with a minimum of 4.0°C in January and a maximum of 36.3°C in August. The soil of the site is composed of quaternary sandy deposits of alluvial origin on a very flat slope. The vegetation is a savanna with Acacia tortilis subsp. raddiana as single tree species. The park is grazed by introduced wild herbivores (Noumi et al. 2015). The dominant vegetation type is a very open pseudo-savanna of A. tortilis, with several shrub species between trees (Retama raetam, Lycium shawii) and a sparse cover of grasses (Cenchrus ciliaris, Digitaria commutata). The vegetation of the Park is lightly grazed (stocking density of approximately 1 animal per 40 ha) by large introduced herbivores such as Saharan antilopes (Addax nasomaculatus and Oryx leucoryx), dorcas gazelle (Gazella dorcas), mhor gazelle (Gazella dama mhor), Barbary sheep (Ammotragus lervia) and some ostriches (Struthio camelus) (Le Houérou 2005).
Experimental design
The experiments were conducted from October 2017 to October 2019. In order to study plant interactions occurring during the recruitment of A. tortilis, we chose a pioneer shrub, Lycium shawii, as the nurse in this study. L. shawii is a species of thorny shrub adapted to desert environments, and can be found throughout the Arabian peninsula and in Africa. The thin leaved, rigid bush grows to 3 meters high, with many branches and alternating spines along the branches and on their tips that vary in size. The leaves narrow towards their base. It produces small whitish-pink or purple flowers from September until April, and red pea-sized seedy berries that are edible by large introduced herbivores of the Park plants often growing nearby include Acacia tortilis and Prosopis cineraria.
In order to assess the relative importance of the canopy and soil effects of L. shawii on herbaceous understorey species, we randomly selected 8 experimental plots (50 m x 50 m). Half of the plots was randomly chosen and fenced to exclude large herbivores using 2 m high fences (grazing treatment) with a mesh size of 50 × 50 mm. To assess biotic interactions with both the removal and observational methods, we created a patch treatment within each plot by selecting ten individuals of Lycium shrub and five naturally open areas. For five shrub plants the above-ground parts were cut at ground level. The basic design consisted of planting A. tortilis in three conditions of the patch treatment (Open, Lycium and Lycium removed).
Species sampling and environmental variables
During October 2018, 12 months after shrub removal, all plants growing in each plot were collected and identified to the species level. Plant material was dried at 70°C for 48 h and weighed. Aboveground biomass was recorded for each species in each quadrat and total aboveground biomass and species richness were calculated per quadrat (50 × 50 cm). Three soil variables were analyzed: oxidizable soil organic matter, which was determined by the Walkley-Black procedure (Nelson and Sommers 1982); extractable phosphate and total nitrogen, which were determined by Olsen’s bicarbonate extraction (Olsen and Sommers 1982) and Kjeldahl’s method, respectively. Soil moisture (volumetric soil water content) was measured at 30-cm depth with a time domain reflectometry probe (ThetaProbe ML2x; Delta T, Cambridge, UK) in all plots, with five replicates. Measurements were done 1, 7 and 14 days after a significant rain event (40 mm)
Plant − plant interaction indices
To investigate the effects of L. shawii on A. tortilis performance in the transplanted plots, we used the relative interaction index (RII) of Armas et al. (2004):
RII = (Xwith neighbour − Xwithout neighbour)/
(Xwith neighbour + Xwithout neighbour)
where X corresponds to the performances (survival) of A. tortilis in the transplanted plots. This index varies between − 1 and 1. Negative values indicate a negative effect (competition), positive values a positive effect (facilitation), and 0 corresponds to no significant interaction.
Michalet et al. (2015) and Noumi et al. (2016) proposed two different indices for disentangling STE and LTE of neighbours in experiments associating both observational and removal procedures in the same community. STE (canopy effects in Michalet et al. 2015) were quantified using the removal method (with neighbours vs. removed-neighbours conditions) and LTE (soil effects in Michalet et al. 2015) comparing target responses in removed-neighbours vs. open conditions (Noumi et al. 2016).
- STE; relative difference in target performance between shrub-control and shrub-removed plots.
- LTE; relative difference in target performance between shrub-removed and open plots.
2.5. Statistical analyses
The effects of our factors on survival of transplants, biomass, richness, density and environmental variables (soil water content and soil nutrient OM, TN and Extractable P) were assessed with a two-way ANOVA model. All univariate analyses were done using JMP software 10.0 (SAS Institute, Cary, N.C.). Tukey’s HSD tests were used to determine the significant differences between treatment means.