Hi,
I would like to build a species accumulation curve using specaccum function in R vegan package.
The entry table should be a classic community table species by sites. However, instead of sites, I would use subsequent surveys in a given study area (conducted with the same methodology) and incident data (presence/absence). Is this supported? And, more relevant, do you think this has a theoretical support? References welcome.
Trying to answer this question, I came across iNEXT package that allows incident data. I have still the same doubt about sites/surveys.
What is the most flexible and performing of the two functions?
Hello Giancomo; Another way of looking at the question is ..."How many species are in the sampled area? Or, have I sampled enough?" You would use incidence data for the calculation. I habitually use the Chao 2 estimator with which I can estimate total species richness in the habitat, Sest = Sobs + (Q1)2 / 2Q2 where Sest = the estimated number of species in the community, Sobs = the observed number of species, Q1 = the number of species represented by a single individual, and Q2 = the number of species represented by two individuals. In this expression the rarest species in the sample are used to estimate the number of unsampled rare species that occur in the community. The observed species richness approximates the estimated richness. In spite of the above attempt to account for very rare species, samples are ordinarily underestimates and are biased against rare species (Dornelas et al, 2012, Gotelli and Colwell, 2001, p 388; Mao and Colwell, 2005; Tista and Fiedler, 2011).
This commonly used method seems parallel to your question and so I suppose that your approach is valid. Cheers, Jim Des Lauriers
Yes, as long as the different surveys were conducted using the same methods. Otherwise you will have to use a heterogeneity variance estimator
As far as the mathematics of rarefaction (species accumulation) is concerned, sites are no different to surveys. The key issue is whether your surveys are equivalent units of sampling effort. It sounds to me that they are. As for references, I think the following will be suitable:
Chiarucci, A., Bacaro, G., Rocchini, D. & Fattorini, L. (2008). Discovering and rediscovering the sample-based rarefaction formula in the ecological literature. Commun. Ecol. 9: 121–123.
In which case, you can use the specaccum function in vegan. Just treat your surveys as if they were sites in the data. If you want a curve that adds surveys in temporal order, then use method="collector" (this assumes that the surveys are in temporal order in the data). If you are fine with surveys in random order, use method="exact". I'm pretty sure only the latter option will give you an estimate of variance around the curve.
Dear all, thank you for your very nice answers.
David Nipperess what do you think about the effect of phenology on the community?
In other words: do you think that the species accumulation (or rarefaction) curve assumes that in all the surveys (conducted with the same method) you have the same probability to spot the same species assemblages? This is problematic whether you carry on, let's say, 8 surveys in a site (2 winter, 2 spring, 2 summer, 2 autumn), but a e.g. butterfly species is found only in spring (because of its biological cycle) and then collected only twice.
Thanks!
Giacomo Assandri : with a random accumulation (rarefaction) curve, the probability a species is counted (for x number of surveys) is a function of the frequency (number of surveys) with which they are encountered. So, in other words, the phenology (temporal sequence) is not taken into account. In your example, the butterfly occurs less frequently and thus has a lower probability of contributing to the curve at any point below the total number of surveys.
It might be interesting to plot both a collector's curve (observed temporal sequence) and a rarefaction curve (random accumulation) on the same graph.
e vedi anche questo e altri software qui
https://www.usgs.gov/software/specrich2
Grazie Giovanni Boano mi sto orientando appunto sulle tecniche sviluppate da Chao e colleghi. Proprio loro hanno recentemente prodotto un pacchetto di R che mi piace molto: iNEXT. Article iNEXT: An R package for rarefaction and extrapolation of spe...
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R