Dear Brian, I think you need to provide a bit more information. You say that the sites appear to have similar richness according to the sample-based curve, but that can mean either that they are reaching similar asymptotic values (even if the shapes of curves before the asymptotes are quite different), or that they are not close to asymptotic, but the initial rises are similar across habitat sites. The same goes for comparing the individual-based curves, although in that case you seem to be referring to the initial slopes.
In general, though, the advantage of sample-based rarefaction (and the reason it's usually the 'preferred' method for comparing richnesses) is that it incorporates between-sample heterogeneity, which generally translates to small-scale spatial heterogeneity. The difference between the sample- and individual-based curves reflects this (the greater the heterogeneity, the flatter the sample-based curve relative to the individual-based). If samples are randomly distributed within a habitat type, more samples means not just more of the area within a habitat sampled, but more of the local heterogeneity. To rarefy by sample, then, is to compare sites by taking equal-sized 'maximum heterogeneity' subsamples. This is probably what most people actually want to do, even if they haven't expressed it directly, because a more heterogenous landscape is likely to mean a greater asymptotic richness. Individual-based rarefaction misses that. But ultimately it comes down to want you mean to achieve by standardization, and how you define the spatial scales of your landscape, which are topics in need of much exploration! Shameless plug: you can also do rarefaction and richness estimation online at www.eco-tools.net Gareth Russell NJIT/Rutgers On Thu, 10 Aug 2006 18:37:39 -0400, Brian D. Campbell <[EMAIL PROTECTED]> wrote: >A colleague and I are collaborating on study of plant diversity within >agro-ecosystems of the Eastern Ghats, India. One of our main study >objectives is compare species richness of several habitat types which >include both forested and non-forested habitats. Obviously in addition to >mean richness per sampling unit (in our case, intercept along 10-m line >transects) we are interested in both a) how many new species are found with >additional samples, and b) controlling for sample size (no. of transects), >what is the expected number of species (rarefaction). Here is where the >confusion arises. When comparing richness of forested and non-forested >habitats by sample based accumulation curves, the two habitat types are >quite similar. Yet, when comparing the habitat types by individual-based >curves, clearly the rate of new species sampled in forests is much greater. > Hence, an initial interpretation of these data is that near equivalent >richness between these habitat types is a largely a function of abundance >(with greater abundance being in the non-forested habitat). A similar >example would be comparing richness in three forest categories (say for >example, young, mid-age, and old) with equivalent sample sizes in each. But >lets say the experimenter decided to spend twice as long in one of the three >treatments, hence over biasing the number of individuals sampled (although >sample sizes are the same). To make valid comparisons, would one not then >be in situation of needing to use individual-based rather than sample-based >rarefaction? > >Any thoughts are much appreciated. > >Brian D. Campbell >Department of Biology >Queens University >Kingston, Ontario. >Canada. >=========================================================== ==============
