Linking species richness curves from non-contiguous sampling to contiguous-nested SAR: An empirical study.

The species–area relationship (SAR) is one of the few generalizations in ecology. However, many different relationships are denoted as SARs. Here, we empirically evaluated the differences between SARs derived from nested-contiguous and non-contiguous sampling designs, using plants, birds and butterflies datasets from Great Britain, Greece, Massachusetts, New York and San Diego. The shape of SAR depends on the sampling scheme, but there is little empirical documentation on the magnitude of the deviation between different types of SARs and the factors affecting it. We implemented a strictly nested sampling design to construct nested-contiguous SAR (SA C R), and systematic nested but non-contiguous, and random designs to construct non-contiguous species richness curves (SA S Rs for systematic and SACs for random designs) per dataset. The SA C R lay below any SA S R and most of the SACs. The deviation between them was related to the exponent f of the power law relationship between sampled area and extent. The lower the exponent f , the higher was the deviation between the curves. We linked SA C R to SA S R and SAC through the concept of “effective” area ( A e ), i.e. the nested-contiguous area containing equal number of species with the accumulated sampled area ( A S ) of a non-contiguous sampling. The relationship between effective and sampled area was modeled as log( A e ) = k log( A S ). A Generalized Linear Model was used to estimate the values of k from sampling design and dataset properties. The parameter k increased with the average distance between samples and with beta diversity, while k decreased with f . For both systematic and random sampling, the model performed well in predicting effective area in both the training set and in the test set which was totally independent from the training one. Through effective area, we can link different types of species richness curves based on sampling design properties, sampling effort, spatial scale and beta diversity patterns. [ABSTRACT FROM AUTHOR]