I have data on quantitative food webs from different sites over three years and three different isolation levels (10 sites for each isolation level). The idea was to look at spatial-temporal changes in the food-web structure; however, the food webs from each site considered individually in each year are very small and do not give reliable results.
I have two alternatives: either pool the data from the three years together for each site and look for the spatial pattern only, or pool the data over the different isolation levels for each year. Both methods are critical as they do not take into account variability either in time or in space. I am be very interested to have your opinion about which approach could be more appropriate. Thank you very much
It depends what you wish to look at... or to show.... and what type of data you have.
During my PhD I also collected data monthly during 3 yrs at 3 sites to study the food web of a coastal lagoon. I also looked at spatial and temporal patterns, but as I had 3 sites and 56 spp., I used multivariate analyses. I had myself many doubts on how to treat my data and received many critics both during the PhD studies and when tried to publish it. What some colleagues argued it wasn’t enough data for a paper ended up in 4 papers in SCI-indexed journals, and I still have some more unpublished data in there. The same raw data could have been analyzed differently… but then I wanted to move on to other projects, as our Post-doc performance independently from our PhD supervisors is rated when applying for research grants.
Later, I also used multivariate analyses to study coral reefs, fish communities, and bird communities.
I am not sure if it will inspire you in any way but feel free to have a look at my papers here or at http://ualg.academia.edu/VandaMendon%C3%A7a.
I also suggest that you have a look at recent papers on food webs similar to yours and get inspiration from there.
Thank you very much!
Konstans, It seems me very attractive to model interaction frequencies, but I should admit, that I did not understand the slide show from the link you indicate. Can I ask you for some more details about what you exactly mean by "modelling interaction frequency"? Maybe some paper which uses this method?
Dear Vanda, I am really impressed by the field work and paper writting you achieved during your PhD and very surprised that you get so much critics. My data are much smaller than yours and were only sample once a year, so that the temporal scale is quite small anyway. But you give me hope by showing that we can go for publications despite critics and many doubts about the usefulness of our statistics :).
As you say, both methods will likely be critisized, I will try to find a good justification for the one I'll go to.
Any further comments are welcome!
If you mean pooling food web data gathered by others, that is, using food web data pool (open source) for your own analyses, you can access that from the Net. Also, you may visit PeaceLab: www.foodwebs.org/
If you are trying to gather your OWN data, I think you must take possible spatiotemporal variation into account" as Konstans suggests. Then for each spatial and temporal "snapshot" community, construct a web matrix, with predators in columns and prey species in rows. Calculate the number of links, fractions of Basal, Intermediate and Top species, links of trophic omnivory, and the chain length.
It all depends on what you want to study with the food web - quantify interaction strength? examine scale dependence of certain structural properties? determine the dynamics of trophic links? correlate web architectural complexity with ecological functions? For each type of study, there is a large body of publications. I would recommend papers of Neo Martinez and Loui-Felix Bersier for great ideas.
A crucial point in constructing high-resolution webs is to separate different life history stages of each species, if the diet changes with life history of a species. Thus, you identify several ontogenetically distinguished organisms of the species ("ontospecies") if relevant diet shifts occur. You may download my food web papers from my ResearchGate profile.
Thank you very much. I already collected all the data. My main concern is that taken individually, each food web is quite small and renders quite difficult the calculation of common indices such as connectance, vulnerability, generality and link diversity. As I collected the data from each site in three years, I thought about pooling the data of each site over the years to get larger webs. However there may be variation in species occurrence between the years, such that I don't know if I can just add the number of interactions and pretending I have a single web for each site. On the other hand, pooling the web together allows to best represent the range of species that actually occur in the site but may be missed with a single year data because of natural variability in species occurrence. I will have a look at your papers
I'm not really sure if this would be useful, but have you considered some kind of Monte Carlo approach? I mean generate many artificial food webs which have the same characteristics as yours (for each location and time) and then apply the original analysis to them. You can generate food webs with a variety of models: niche, groups model or minimum potential niche (Allesina), trophic hierarchy...
Hope it helps
Pooling will be useful IF the overall species composition at all three sites over the years is the same. Conversely, if sp. X is never found at site A, but frequently found at site B, then merging of all food webs from different sites will be statistically flawed. If all species are found at least once in the temporal food webs at each site, then pooling is certainly valid. You can create randomly generated food webs from the total pool of species - as Francisco de Castro has suggested (Monte Carlo approach). In my own publications, I pooled all spatial and temporal webs. My approach has been (1) examine all the food webs from different sites and different time periods, and then (2) pool all individual "snapshot" food webs, and create thousands of non-randomly assorted webs (with biologically feasible links between each random pair of species) from the pooled repertoire of species.
To me, link diversity and connectance are fairly easy to calculate, though vulnerability is more difficult. However, there are some programs to calculate all these parameters. (I wrote my own programs in QBASIC and PASCAL.) You may contact Peacelab for assistance. ECOWEB community will also be helpful.
Just in case it's useful, I have a MATLAB function to plot nice 3D foodwebs. You can download it from: http://www.mathworks.com/matlabcentral/fileexchange/32419-plot-food-web
Thank you. I applied Monte Carlo simulation to calculate nestedness and as you suggest it may be useful for other food web properties. I usually refrain from using this approach, because I am very not used in applying simulation methods, but I really should become more confident with that, as it is becoming a very used method now. I am saddly not a MATLAB user, but the application looks great.
I think you can do both and compare between them. It will be interesting. I have conducted a research like yours in rice field ecosystem in Indonesia.
Hi Valerie. I think you can do both and compare between them. It will be interesting. I have conducted a research like yours in rice field ecosystem in Indonesia.
Hello Tien! Yes I will definitely do both and think about what is the more meaningful in regard to my hypotheses and the literature. Did you wrote a paper on your worK?
- Badges
- Science topic
- Similar topics
- Papers