Dear community. I'm new to mailing lists and to this list in particular so this is my first post. I'm sending this message, because I'd like some advice on an ecological topic and I would appreciate any input from this community. Many thanks in advance.
I want to work on antagonistic interaction networks between herbivorous insects and their woody host plants. My approach is that I want to sample 15 different plots in three different habitat types (forest fragments, forest edges and forest centers), so 45 plots at the whole. In each plot (20 m x 50 m), I'm going to collect insects from each tree individual by hitting the branches. Afterwards I want to use No Choice Feeding Assays to find out, which insects are actually herbivorous. The herbivores will then be identified by morphospecies and used as a basis for bipartite networks. For this I want to use the "bipartite" package provided for the programme "R", so I hope some people out here might already have some practical experience with this package. My aim is to compare indices the package can calculate between the three different habitat types. So I should end up with 15 repetitions for each value and for each habitat type. My questions now are: Some of the Indices the package can calculate are strongly dependent on sample size. That means, the value moves into an asymptotic region for increasingly bigger networks. So if a network for a single plot is too small to return "asymptotic" values, would you try to fuse together several networks, albeit risking to loose variation in your repetitions? And if so, how would you try to compare the then returned indices between habitats? My approach would have been to then use reiterations like in species-area curves. My second problem is somehow connected to this. It might turn out, that 15 plots per habitat are slightly too much to sample in the time I have. Therefore I have two possibilities: Either reduce plot size and thus lose tree diversity or reduce plotnumbers/repetitions and thus lose variation in network-index variation. What's your opinion on that? My last question is about how to translate feeding herbivores as interactions into a bipartite matrix. I think an example would be the best illustration: Let's imagine two tree individuals A and B of the same species. We observe the same herbivor species on both trees in different abundances, let's say 15 beetles on tree A and 50 beetles on tree B. How many interactions would you then count for this tree species? Either two (the particular beetle species was observed on two tree individuals) or 65 (the whole abundance of this beetle species on the trees). The first approach would lay more weight on the tree abundance in the matrix, the latter more weight on the insect abundance. What do you think? I'm very sorry for this mass of text, I tried to get it as fitting as possible. As I said, I would appreciate any input whatsoever. Many thanks in advance and I am curious on what you think. Regards, Kevin Bähner
