Shaded blocks were nestled amongst Acacia acuminata andMelaleuca sp. trees. Each block contained three plots: one with ambient precipitation, one with a rainout shelter that reduced approximately 50% of incoming precipitation, and one with water manually added after rainfall events (Fig. 1, Figure S1). Each plot was 2.35 × 2.5 m and divided into 72 15 × 15 cm subplots to create eight subplots per species.

We collected seeds for the nine focal species across West Perenjori Nature Reserve in September and October 2019. The seeds were after-ripened in the field over summer to maintain conditions similar to those experienced naturally. We secured the seeds in mesh bags and, to minimise potential after-ripening effects (Dwyer & Erickson 2016), distributed the cages across two open and two shaded locations within the reserve. Seeds were pooled and well mixed before sowing in February 2020.

Subplots were randomly assigned to each species and neighbour treatment. We sowed fifteen seeds into shallow holes (approximately 0.5 cm deep) in the centre of each subplot and covered them with a thin layer of topsoil and one teaspoon of water to adhere the seeds to the soil. Toothpicks were used to mark the top of the sown areas to help distinguish emergence of sown seeds from those emerging from the seedbank. In February, one week of unseasonal heavy rainfall cued emergence for predominantly the two exotic species (A. calendula with 136 plants from 44 subplots and P. airoides with 82 plants from 18 subplots) and some native species (G. rosea with 12 plants from 10 subplots and T. cyanopetala with one plant). We counted the number of plants that emerged and added 5-10 extra seeds to subplots that had more than five plants emerge. While we did not track the fate of these summer plants, conditions returned to typically hot and dry within 10 days and we are confident that they all died before reproducing.

At the start of winter in July, we counted emergence of the sown seeds for each subplot. We then thinned these emerged plants down to a single focal individual closest to the centre of the subplot. To avoid disturbing the soil, we carefully thinned plants one at a time by snipping the main stem close to the soil.

To assess species’ responses to the environment in the presence and absence of neighbours, we randomly assigned three subplots per species per plot to a neighbour removal treatment, and three subplots to remain with naturally occurring assemblages. For the thinned subplots, we removed all neighbouring plants rooting within a 7.5 cm radius of the focal individual as per protocol described in Mayfield and Stouffer (2017), appropriate for capturing most direct plant-plant interactions in this system (Martyn 2020). We recorded the abundance and identity of all neighbouring plants within the interaction neighbourhood during peak flowering in September.