amchiapella
commented
3 years ago This is one of those fun instances where I figured out a good answer (for my case at least) within minutes of posting this issue... maybe putting it down on paper helped me think through my options better. Posting the solution here in case it's helpful for others. Out of my three fish species that each had three size classes, exactly one of those size classes of one of those species had a specialized diet (ie, only ate one thing, as indicated by stomach contents). In this case, small alewife exclusively ate zooplankton. So I calculated the TDF between zooplankton and small alewife, and was even able to account for tissue turnover by taking the mean d13c and d15n of zooplankton for the first three months of the study and subtracting those from the mean d13c and d15n of small alewife from the latter 3 months. I decided these TDFs were more accurate to extrapolate to my other fish species (which were a similar size & fed on roughly similar prey) than some random lab study (or compilation of lab TDFs). In case it's helpful: TDFs were 2.47 (SD 1.15) for d13C and 2.3 (SD 1.08) for d15N. System is a large temperate oligotrophic lake.
ChrisHarrod
The more I get to know my data, run test models, and explore reviews of TDFs, the more apparent the incredible importance of proper TDFs in mixing models becomes. The range of possible TDFs is quite large for my consumer group (fish), and no lab studies exist for my species, or even a similar species (let alone their natural diets). How do folks reconcile this? Just take the mean of all relevant literature TDFs I can find and have a large SD in my discrimination file? Try to use C:N ratios of my prey to somehow adjust lit TDFs accordingly? Even with pretty robust prior diet information (hundred of stomach samples over the course of 6 months) and good model diagnostics, my MixSiar results just really don't make sense, and it's clear to me this is due to the TDFs. Any insight welcome! Thanks