Pesticides used in agriculture can end up in nearby streams and can have a negative impact on nontarget organisms such as aquatic invertebrates. During registration, bioaccumulation potential is often investigated using laboratory tests only. Recent studies showed that the magnitude of bioaccumulation in the field substantially differs from laboratory conditions. To investigate this discrepancy, we conducted a field bioaccumulation study in a stream known to receive pollutant loadings from agriculture. Our work incorporates measurements of stream pesticide concentrations at high temporal resolution (every 20 min), as well as sediment, leaves, and caged gammarid analyses (every 2-24 h) over several weeks. Of 49 investigated pesticides, 14 were detected in gammarids with highly variable concentrations of up to 140 ± 28 ng/gww. Toxicokinetic modeling using laboratory-derived uptake and depuration rate constants for azoxystrobin, cyprodinil, and fluopyram showed that despite the highly resolved water concentrations measured, the pesticide burden on gammarids remains underestimated by a factor of 1.9 ± 0.1 to 31 ± 3.0, with the highest underestimations occurring after rain events. Including dietary uptake from polluted detritus leaves and sediment in the model explained this underestimation only to a minor proportion. However, suspended solids analyzed during rain events had high pesticide concentrations, and uptake from them could partially explain the underestimation after rain events. Additional comparison between the measured and modeled data showed that the pesticide depuration in gammarids is slower in the field. This observation suggests that several unknown mechanisms may play a role, including lowered enzyme expression and mixture effects. Thus, it is important to conduct such retrospective risk assessments based on field investigations and adapt the registration accordingly.