Abstract Biogeochemical models used to estimate carbon export fluxes are highly sensitive to their parameterization of zooplankton grazing dynamics. Zooplankton grazing dynamics have been shown to vary spatially, however it is less clear how they are responding to underlying environmental drivers. In this study, we use machine learning regression models to quantify the relationship between spatial variability in observationally inferred zooplankton grazing dynamics and three environmental drivers. The majority (R2 ${R}^{2}$ = 0.80) of the variance in zooplankton grazing dynamics can be explained by sea surface temperature, mixed layer depth and chlorophyll. Globally, chlorophyll is the best predictor of zooplankton grazing dynamics, with slower grazing plankton communities associated with higher chlorophyll concentrations and deeper mixed layer depths. The findings show that zooplankton community grazing and composition is well constrained by biological drivers, which could provide a pathway to parameterize more diverse grazing dynamics in models and in turn simulate more realistic carbon export estimates.