l have invertebrate survey data from distinct habitats with varying sizes. The sampling procedures (number of traps and sampling effort) were the same for all habitat types but each habitat differs in terms of size.
Is it necessary to account for the habitat sizes when comparing species richness of these habitats and how could this be done?
The relationship between habitat size and species diversity is that the smaller the habitat, the less diversity of species there will be. But since your sampling is the same in each habitat, the results themselves will take this into account!
If you are worrying about the possible bias of having equal sampling effort to systems that might differ in diversity (due to size for example), you can try standardizing your indices by completeness via rarefaction approaches. See papers of Anne Chao and Lou Jost, eg, Chao and Jost 2012 Ecology 93 2533-2547. This approach has an R implementation via iNext package (I think).
Andrew Paul McKenzie Pegman, Elia Guariento , Appreciated . l thought as much. Thanks for confirming that the results themselves shall take the differences in sizes into account.
Elia Guariento, l have 3 sites (a natural forest and a plantation forest with almost the same sizes - 128 ha and an arable field with size of about 30 ha).
Ambrocio Melvin Agulan Matias, thanks very much for your inputs. l will try standardizing the indices using the rarefaction approaches to minimize any possible bias. Thanks for the literature's, they seem nice and significant literature. I will go through it.
As discussed verbally, since the distribution in the large plot captured all possible variabilities there should be no worries.
I would think you would also need to account for not just habitat size - you could look at some of the classical literature on Island Biogeography for whyscholar.google.com/scholar_url?url=http://plantecology.syr.edu/conbio/readings/Simberloff%26Abele1976_science.pdf&hl=en&sa=X&scisig=AAGBfm0U7swltzLKdB3QKPU0Pd8hmLLj6A&nossl=1&oi=scholarr , but also habitat quality and diversity. You mention disparate habitat types, and while rarefaction is important when comparing the asymptotes of species richness for a particular habitat type, it would be important to further think about how those curves would compare across different habitat types, and maybe factor those in as covariates to your models. I've attached Chao et al. 2009, Woodcock et al. 2010, and Hortal et al. 2009. I hope they help!
Eric Yensen , standardization via rarefaction would be essential in this case. Thanks very much for your input.
Andrea Ayala, rarefaction will account for the desparate habitat sizes which will then be used as a basis to compare the habitat quality and diversity of the various systems. Thanks for highlighting the most important issues and recommending these literatures.
Measure per unit area. Density of frogs per meter of shoreline. Density of plankton per liter of water. Etc.
Measure per unit area will be misleading because diversity does not increase linearly with sample size. Diversity follows the species-area curve, or the species-sample curve. You could standardize on the smallest area sampled, but rarifaction would be preferable.
Habitat quality and habitat type do not always correlate to species diversity and therefore rarefaction, especially when you add in fragmentation (different plot sizes and the potential inflation of edge effects). For example, a bottomland hardwood forest and an upland white oak forest in Georgia (USA) should have two different rarefaction curves. You could plot them on the same graph to see which species overlap, but just doing rarefaction and not adding in the covariates may undersimplify your analysis.
Some specifics on the rarefaction: You can sum all samples from one habitat into a single large sample and rarefy that. If you add confidence limits to the rarefaction curve, then you can plot a second rarefaction curve from the sum of samples from a different habitat. If that curve falls outside of the confidence limits of the other curve then you can say infer that the two locations are significantly different. Figure 4 in this article illustrates the point: Article Sampling Biological Communities
It also shows how to vertically compare the curves at the same abundance level and may be of value to the others who teach this approach. Figure 5 shows the collector's curve approach which is one way to ask if you took enough samples to capture most diversity.
Software (free version and commercial version) to do the rarefaction and calculate confidence levels at each abundance level is cited in this article: Article Null Models for Everyone: A Two-Step Approach to Teaching Nu...
Sounds like a wonderful project and you seem to have consensus on rarefaction!
- Badges
- Science topic
- Similar topics
- Statistics
- Survey