We use inverse modelling to infer the distribution and drivers of
community-integrated zooplankton grazing dynamics based on the skill
with which different grazing formulations recreate the
satellite-observed seasonal cycle in phytoplankton biomass. We find that
oligotrophic and eutrophic biomes require more and less efficient
grazing dynamics, respectively. This is characteristic of micro- and
mesozooplankton, respectively, and leads to a strong sigmoidal
relationship between observed mean-annual phytoplankton biomass and the
optimal grazing parameterization required to simulate its seasonal
cycle. Globally, we find Type III rather than Type II functional
response curves consistently exhibit higher skill. These new
observationally-based distributions can help constrain, validate and
develop next-generation biogeochemical models.