Abstract:

We determined the best-fit model for, and explored the mechanisms shaping species-abundance distributions (SADs) by fitting five widely-used SAD distribution models at several scales. We used data collected in 2005 from a 24-ha dynamic plot in an evergreen broad-leaved forest in the Gutianshan National Nature Reserve. We estimated SAD at different sampling scales from the mean value of SADs taken from 100 randomly-selected subplots within the 600 m×400 m Gutianshan plot. We subsequently used the SADs to test the fit of different models, including the broken stick, lognormal distribution, niche preemption, Zipf, Zipf- Mandelbrot, and neutral models. We employed AIC and χ²values to test goodness-of-fit for these models. All computations were conducted using the Vegan package in R 2.7.1. At smaller scales (10 m×10 m and 20 m×20 m), the broken stick, lognormal distribution, niche preemption, Zipf, and Zipf-Mandelbrot models all fit well to the observed species-abundance distribution. The Zipf-Mandelbrot was the best model at the 20 m×20 m scale. The Lognormal was the best-fit model at the 40 m×40 m scale, and the Zipf-Mandelbrot model was the only suitable one in explaining the observed SAD at scales of 60 m×60 m and 80 m×80 m. None of these models performed well at a scale of 100 m×100 m, but the neutral model was better at explaining patterns of SADs at larger scales (40 m×40 m to 100 m×100 m) than smaller scales and it is suitable in explaining patterns of SADs at all scales. Patterns in SAD were scale-dependent, suggesting that SADs at different scales are likely structured by different ecological processes.

Key words: species-abundance distribution, sampling scale, subtropical ever-green broad-leaved forest, neutral theory, different ecological processes