Data analysis
For each site we built one pollinator and one seed dispersal network per season, resulting in a total of six plant-pollinator and six plant-seed disperser networks (i.e. each site-year combination has a corresponding network). The plant-seed disperser network in 2017-2018 in one of the sites without the keystone interaction was excluded from the analysis because several plant species produced no fruits.
Our analysis focused on network “motifs”, which may be regarded as the building blocks of networks, and consist in sub-networks composed of a small number of species exhibiting particular patterns of interactions (Milo et al. 2002). Motifs are valuable tools to assess the structure and ecological importance of species in networks because they have two structural levels of organization. In one structural level, a motif represents a unique pattern of interactions among a subset of species within a community (Simmons et al.2018), while in the other level a motif is composed by two or more unique positions that can be occupied by different species simultaneously (Fig. S1). Each of these positions represents a different ecological pattern with direct and indirect interactions (Stoufferet al. 2012; Baker et al. 2015; Cirtwill & Stouffer 2015; Simmons et al. 2019). Therefore, the frequency of positions that a particular species occupies defines its ecological importance in the community (Simmons et al. 2019). Even if different species occupy the same position, the motif still conserves its ecological function. For example, in the motif integrated by two plant species and a pollinator species, plant species could be “A and B” or “C and D” but the motif would still indicate competition or facilitation between the two plant species. The advantage of motifs is that they are significantly more sensitive to changes in network structure than the network indices commonly used (i.e. degree distribution, nestedness) (Simmons et al. 2019). In addition, the meso-scale level motif analysis incorporates indirect interactions undetected by macro-scale network indices (i.e. nestedness, connectance) and lost at species-level indices such as species strength (Simmons et al. 2018).