Site description: Rucamanque forest (38°39´ S; 72°35´ W) is a relict of native forest located in the surroundings of Temuco town with 435.1 Ha. This site has native forest with Aetoxicon punctatum Ruiz et Pav., Nothofagus obliqua (mirb) Oerst., Eucryphia cordifolia Cav., Laurelia sempervirens Ruiz et Pav, Persea lingue Ruiz et Pav., Lauereliopsis philipiana (Looser) Schode and Weinmannia trichosperma Cav [21][39]. This relict has numerous mountain paths with small ephemeral stream that is present only in rain season (April-August) with a difficult access in winter (Jule-August). The studied site was a small stream located inside the park after walking a long mountain paths[21].
The site was visited in April 2022, that corresponded to southern autumn, when the stream is present. because it has low flow in summer due dry season. The stream is in a mountain slope with native forest. Benthic samples were taken using 50 x 50 cm Surber net randomly of 500 µm2 mesh size, 20 samples were taken. Collected specimens were fixed in absolute ethanol, quantified and identified in according to literature descriptions [40].
Data analysis: Variance/mean ratios were calculated to determine if the spatial distribution pattern of the studied populations was associated, uniform or random [7][8][9][10]. First, we registered the number of individuals for each sample, and then determined the variance and mean of each sample as a way to determine the spatial pattern for both species. So, if the variance-mean ratio value is 1, the distribution is random; whereas if the variance mean ratio is lower than 1, the distribution is uniform; and finally if the variance mean ratio is greater than 1, the spatial distribution is aggregated [9][10]. After this, data were examined using the Poisson, binomial or negative binomial distributions as appropriate probabilistic models of the spatial distribution patterns results obtained by Variance/mean ratio [9][10]. If the first analysis denoted associated, uniform and random spatial distribution, a second step analysis was applied with the negative binomial, positive binomial, and Poisson probability distributions, respectively [9][10], all these statistical analysis were done using Xlstat software [9][10].
A community is structured by competition when the C-score is significantly larger than expected by chance [13][14][41]. The row and column sums of the matrix are preserved in this model. Each random community thus contains the same number of species as the original community (fixed column), and each species occurs with the same frequency as in the original community (fixed row). In a fixed-equiprobable algorithm, only the sum of rows are fixed, and the columns are treated as equiprobable. This null model treats all the samples (columns) as equally suitable for all species [13][14]. The null model analyses were performed using the software R [42] and the package EcosimR [43][44].
For the niche overlap analysis, an individual matrix was built in which rows and columns represented species and sites, respectively. This matrix was used to test if the niche overlap significantly differed from the corresponding value under the null hypothesis (random assemblage). These analyses were applied to data from the second field collection and were based on Pianka and Czekanowski index. The models show the probability of niche sharing compared to the niche overlap of the theoretically simulated community [43][44]. The niche amplitude can be retained or reshuffled. It preserves the specialization of each species when the niche amplitude is retained. By contrast, the algorithm of niche amplitude uses a wide utilization gradient of specialisation when it is reshuffled. Furthermore, zero participation in the observed matrix can be maintained or omitted. We used the RA3 algorithm, which retains the amplitude and reshuffles the zero conditions [44]. This null model analysis was performed using the software R [42] and the package EcosimR [43][44].