Most of accumulation curves tend to underestimate species richness, as they do not consider spatial heterogeneity in species distribution, or are structured to provide lower bound estimates and limited extrapolations. The total-species (T-S) curve allows extrapolations over large areas while taking into account spatial heterogeneity, making this estimator more prone to attempt upper bound estimates of regional species richness. However, the T-S curve may overestimate species richness due to (1) the mismatch among the spatial units used in the accumulation model and the actual units of variation in beta-diversity across the region, (2) small-scale patchiness, and/or (3) patterns of rarity of species. We propose a new framework allowing the T-S curve to limit overestimation and give an application to a large dataset of marine mollusks spanning over 11 km(2) of subtidal bottom (W Mediterranean). As accumulation patterns are closely related across the taxonomic hierarchy up to family level, improvements of the T-S curve leading to more realistic estimates of family richness, that is, not exceeding the maximum number of known families potentially present in the area, can be considered as conducive to more realistic estimates of species richness. Results on real data showed that improvements of the T-S curve to accounts for true variations in beta-diversity within the sampled areas, small-scale patchiness, and rarity of families led to the most plausible richness when all aspects were considered in the model. Data on simulated communities indicated that in the presence of high heterogeneity, and when the proportion of rare species was not excessive (>2/3), the procedure led to almost unbiased estimates. Our findings highlighted the central role of variations in beta-diversity within the region when attempting to estimate species richness, providing a general framework exploiting the properties of the T-S curve and known family richness to estimate plausible upper bounds in gamma-diversity.
