Species Accumulation Curve

There is no universal threshold. The curve becomes reliable when it shows a clear asymptotic trend—the rate of new species discovery has slowed substantially. A rough rule of thumb is to sample until you are finding new species at less than 5% of your total species count per new sample. Statistically, EstimateS and similar tools provide confidence intervals that reflect uncertainty.

Rarefaction downsamples your observed data to a uniform effort level, allowing fair comparison of curves. Extrapolation extends the curve beyond your observed data using a fitted model to estimate richness at higher effort. Rarefaction is reliable; extrapolation is speculative and should be reported with wide confidence intervals.

Most real species accumulation curves are asymptotic (approach a limit slowly) rather than flat. Continued increase suggests either that sampling effort is still revealing new species or that a true asymptote exists beyond your current sample size. If you have sampled extensively and the curve is still rising steeply, the habitat likely harbors substantially more undescribed species than you have found.

Yes. For high-throughput sequencing, rarefaction addresses differences in sequencing depth across samples. For eDNA, the approach is similar, but you must carefully define what constitutes a 'species' (e.g., operational taxonomic unit or amplicon sequence variant) and interpret results cautiously because molecular detection does not always equal ecological presence.

Different estimators have different strengths. Chao estimators work well when rare species are present. Jackknife is generally conservative. Bootstrap is flexible but computationally intensive. Compare estimates from multiple methods; if they converge, you have more confidence. EstimateS and R's vegan package compute all three and allow comparison.