simulate many processes that are known to be part of the lake’s
biogeochemistry, including complex carbon substrate utilization
profiles. We aimed to create a model that captured the broad
spatial patterns and temporal dynamics with the minimum
number of processes possible, thus balancing the model’s completeness
with its simplicity. For example, the observed oxygen minimum
at the thermocline is unusual but not rare for dimictic lakes. There
are many competing explanations for this type of minimum (for
example, changes in temperature, predator abundance or horizontal
mixing19) that are all beyond the scope of the model and so it fails to
predict that minimum. Third, the model was designed to model the
lake’s general seasonal dynamics rather than its behaviour in the
specific season when we conducted the survey. For the purposes of
this study, understanding the lake’s general organizing processes
was more important than understanding the dynamics in a particular
year. Although these discrepancies limit the interpretability
of hypotheses about OTUs’ function that are generated by the biogeochemical
model, we showed that additional bioinformatic and
experimental evidence can together provide a more complete picture.