i made Species accumulation curves of Deep-sea shrimps resulting from different species richness estimators of the North eastern Indian ocean. Observed richness (Sobs), mean estimated richness by Chao 1and Chao 2 first (Jack1) and second (Jack2) order Jacknife estimators Species accumulation curves of Deep-sea shrimps resulting from different species richness estimators of the North eastern Indian ocean. Observed richness (Sobs), mean estimated richness by Chao 1and Chao 2 first (Jack1) and second (Jack2) order Jacknife estimators. i am little confused how to explain this Species accumulation curves(Chao 2 behaves differently), kindly help me for making a solution. I am attached my plot for the same.
The overall trend is the same for all estimators – they begin and end at approximately the same place. There may be some environmental factor working at c. 300 m, but your species counts are low, so it may be simply be due to the stochastic nature of interactions between environment and organisms :-)
Dear Sileesh,
I have 3 questions regarding your figure: Did you calculate the estimators for each sample? And if so, which criteria follows the ordering of samples on the x-axis?
1,2 and 3: 576-650m depth
7,8,9,and 10: 330-507m depth
4,5 and 6: large range between stats (>2); intermediate depth (mixing; variable?)
Hi,
I agree with Andrew. It seems stochasticity from samples, but there could be something special at 300m. It is weird than you have in some estimators more species at that point that in the tail of the curves. Maybe some species just appeared there and in low quantities? (so the resampling is not detecting them in some models).
Anyway, it would be easier to interpret if you add the 95% confidence intervals obtained form resampling. This would show whether it is just randomness or there is some problem.
I can see only two species accumulation curves: MM and UGE. But in this case there should be the number of stations on the x-axes instead of specific stations with a specific depth. The other estimators and in particular Sobs give only one value per station (may be based on several samples per station) but no "curve". So I am puzzled what is really shown in the figure. That´s why I asked what MM and UGE means, whether the estimators were calculated per station, and which system underlies the ordering of the stations on the x-axes (I cannot see any pattern regarding the water depth, may be a transect?).
Hi Sileesh. I agree with Roland; you need to have number of stations (or some other measure of sampling effort) on the x-axis. Rarefaction will give smooth curves. And Rabi's suggestion of CI's will aid interpretation.
Kind regards, Tom
I agree with the previous answers: Stochasticity may play a role. Also please note that it may be difficult to interpret the curves without a little more info.
While most other indices may also work on incidence data, Chao-1 has been calculated and this is for abundance data. (I have not used UGE but as it is probability-based I think it also ultimately relies in abundances [Roland, I presume in this context MM means Michaelis-Menten UGE is the Ugland-Gray-Elligsen SAC of 2003]).
The thing is that this type of relatively wild walk is more easily seen the lower the abundance numbers--therefore I wonder if the data were just binary (incidence) data (presence-absence) treated as abundances? If that's the case then the curves will have a high implicit stochasticity that could be smoothed out by either using the abundance data or increasing the number of samples.
About the X-axis ordination I also agree: for some reorder-based SACs e.g. UGE it may not matter, but for other indices it would be informative to follow some sort of natural order--e.g. distance along a transect or depth.
Thank you, Arturo, for clarifying the issue concerning MM and UGE. You are probably right.
At site 1, the species richness is lesser compared to all other sites. Also, the site 5 has highest species richness that gives the clarity of highest species occurring in the site and this site is undisturbed with other activities/ interference
Hi,
As I understand it, species accumulation curves are not used to Analyse the species richness / diversity of one sample: it is primarily a tool to see, whether all/most of the species living in a specific Environment have been sampled or not.
Irene -- Well, yes and no -- I think that while a SAC's primary purpose may certainly be to ensure minimal sampling for richness or diversity (see e.g. attached paper), they could also be used to validate rarefaction curves to fixed sample sizes and bootstrapped or jackknifed total richness.
In fact, I'd argue the SAC probability function (roughly, the general geometry and breadth of the area covered by all possible curves after randomly selecting the samples' entry order) can be thought of as the model underlying all three techniques, even though the rarefaction itself is strictly deterministic and based on the species abundances' distribution, exclusively. Even Chao's 1 and 2 probabilistic indices based on singletons and doubletons could be interpreted in terms of the probability of encountering new species further right along a randomly-built SAC.
Regards
Chapter Optimal Sampling for Complexity in Soil Ecosystems
@Arturo,
Thank you for your explanations! It is some time since I used species accumulation curves and I´m no expert. My point is: all these curves and indices are not ment to tell you anything about individual samples. They are ment to describe the speices richness / diversity in the specific habitat characterized by the samples.
In @Sileeshs case, this would mean (based on the graph he provided), that with 10 samples you can get a fair share of the species present in his part of the ocean at depth from approx. 300 to 700m.
If @Sileesh wants to know, why sample 1 and 10 (both from more than 500m depth) are so different in species number, he has to use other information.
@Arturo: would it be appropriate to use SAC or other related techniques to find out, which samples are significantly different in species number? I´m looking forward to reading your paper, btw!
kind regards
Irene
Irene, I think you're right in that SACs won't tell you anything about individual samples. In the case we examine the problem is that we don't have enough information to interpret the curves. A SAC should be built by pooling samples together sequentially. The curve may or may not flatten out; if it does then were at or near the total richness. However, the order in which each sample enters the pooled sample often matters especially for low-richness samples (or too small samples). Thence the recombination techniques that aim at getting richness spectra over sample size. If, however, what we see here are the individual richness of each sample then we are NOT looking at a SAC at all but at a transect. Then the ordination of the samples matters and judicious choice of ordination (e.g. by depth, distance, transect position...) will possibly give the key.
The question of whether there's something special about particular samples, as you say, cannot be solved by merely looking at the SAC, raw or processed. The information has been collapsed and many explanations could be possible. Other respondents have pointed out that "something happens" at 300 m -- but I think that's circumstantial; it's just the fifth sample to enter the SAC (if this was built as expected) and there are two other at about the same depth apparently showing nothing in particular. Reordering samples may then move the high-variance point and if this sticks to that particular depth then we may perhaps say something. Again, that question should probably best solved by common ordination analysis IF all samples are guaranteed to be minimal.
About your question on using SACs to estimate significant differences between individual samples: I don't really know. I haven't worked enough with SACs to give you a clear-cut answer to that. When dealing with samples composed of multiple subsamples one could use statistical inference--but along a SAC curve made up from individually-pooled samples the only variance you may get is a pseudovariance coming from the bootstrapping or reordination of samples, as in the paper I sent earlier. I've done two or three further attempts at that (and a paper using SACs for sample size in frog population genomics by one of my PhD students is due to appear soon, hopefully) but I find the issue rather finnicky and bobby-trapped. That said, I'd look at papers by e.g. Colwell, May, Chao, Krebs, Jost, Tuomisto, Ferrier... which have extensively worked out with diversity components and comparisons.
You may find some references (which may or may not be too helpful for that question) at the end of the attached presentation, an introductory lecture on beta diversity and GDM.
Regards
Conference Paper Beta diversity
I fully agree with Arturo and thank Irene for the helpful questions. My interpretation of the graphs is: What we see in the figure is a mixture of (a) species accumulation curves (SACs) based on 10 samples/stations (MM, UGE) and (b) data rows consisting of single values per station (Sobs, Chao1, Chao2, ..., Bootstrap), which probably are based on several samples per station. The lines between these data points per station are misleading and suggest a "curve" where no curve is, at least no "accumulation" curve. So @Sileesh, I recommend to separate the two issues (draw two figures) for a meaningful interpretation of the species richness (a) over all stations/along a transect (SACs) and (b) of each station. If there are several samples for each station than you could additionally calculate a SAC or a rarefaction curve for each station separately.
Kind regards
Roland
