Any attempt to assess species abundances must employ a sampling design that balances collection of accurate information for many species with a reasonable sampling effort. To assess the accuracy of commonly used withln-site sampling designs for sessile species, we gathered cover data at 2 rocky intertidal locations in Southern California using a high-density point-contact method that maintained the spatial relationships among all points. Different sampling approaches were compared using

more »

... d sampling. Different sampling units (single points, line transects, and quadrats) were modeled at high and low sampling efforts. Sampling units were either distnbuted randomly or with stratified random methods. Sampling accuracy was assessed by comparing cover and species richness estlrnated by the sampling simulations to the actual field data. Randomly placed single point-contacts provlded the best estimates of cover but are usually not logistically feasible in the rocky intertidal, so ecologists typically use quadrats or line transects. With quadrats, some form of stratified random sampling usually gave estimates that were closer to known values than slmple random placement. In nearly all stratified cases, optimum allocation of sample units, where quadrats are allocated among strata according to the amount of variability within each stratum, yielded the most accurate estimates. With 1 exception, line transects placed perpendicular to the elevational contours ('vertical transects') approached or exceeded the accuracy of the best stratified quadrat efforts. The estimates for rare species were consistently poor since sampling units often missed such species altogether, suggesting a systematic blas. Species richness was substantially underestimated by all sampllng approaches tested, whereas these same approaches accurately estimated diversity (H'). These results illustrate the difficulty of obtaining accurate cover estimates in rocky intertidal communities.