Effect of sampling intensity on understanding species co-occurrence pattern of fish community using null model analysis

Species co-occurrence pattern is fundamental in community ecology. Co-occurrence analysis identifies mechanisms for species coexistence in fish community. Different sampling intensity may result in different patterns at certain scales. In this study we conducted a simulation study to evaluate differences in the three species co-occurrence indices under different levels of sampling intensity. We used the fixed–fixed (FF) and fixed-equiprobable (FE) algorithm models to generate new matrices with different sampling intensities for a simple random sampling (SRS) and a systematic sampling (SS) survey. The estimated indices tended to approach the ‘true’ values of the co-occurrence indices of fish communities when sampling intensity increased. Sampling intensity showed a relatively low influence on the estimates of co-occurrence indices when the null hypothesis was accepted and high influence when the null hypothesis was rejected. As for checkerboard score (C-score) and Checker, when the index under both FF and FE algorithm models simultaneously indicated that the community structure was non-random, the overall error rate of FF algorithm model was lower than that of FE algorithm model for both indices. The C-score was a commonly used index of species segregation. We generated new occurrence matrix to investigate the relationships between C-score and the slope of species-area curves of fish communities for exploring the effect of species composition on species co-occurrence analysis. The C-score showed decreasing trends when the slopes of the species-area curves increased for the simulated fish communities and when the sampling intensity increased from 5 to 35 sampling grids in four seasons. The relatively low sampling intensity was needed for fish communities with more widespread species, and relatively high sampling intensity was required for communities with more rare species to ensure the representative data for species co-occurrence analysis. We conclude that sampling intensity and fish community attributes could influence the results of species co-occurrence analysis using null model analysis.