Assessments of ecological risk require accurate predictions of contaminant dynamics in natural populations. However, simple deterministic models that assume constant uptake rates and elimination fractions may compromise both their ecological realism and their general application to animals with variable metabolisms or diets. In particular, the temperature—dependent metabolic rates characteristic of ectotherms may lead to significant differences between observed and predicted contaminant kinetics. We examined the influence of a seasonally variable thermal environment on predicting the uptake and annual cycling of contaminants by ectotherms, using a temperature—dependent model of137Cs kinetics in free—living yellow—bellied turtles, Trachemys scripta. We compared predictions from this model with those of deterministic negative exponential and flexibly shaped Richards sigmoidal models. Concentrations of137Cs in a population of this species in Pond B, a radionuclide—contaminated nuclear reactor cooling reservoir, and137Cs uptake by uncontaminated turtles held captive in Pond B for 4 yr confirmed both the pattern of uptake and the equilibrium concentrations predicted by the temperature—dependent model. Almost 90% of the variance in the predicted time—integrated137Cs concentration was explainable by linear relationships with model parameters. The model was also relatively insensitive to uncertainties in the estimates of ambient temperature, suggesting that adequate estimates of temperature—dependent ingestion and elimination may require relatively few measurements of ambient conditions at sites of interest. Analyses of Richards sigmoidal models of137Cs uptake indicated significant differences from a negative exponential trajectory in the 1st yr after the turtles' release into Pond B. We also observed significant annual cycling of137Cs concentrations, apparently due to temperature—dependent metabolism and its influence on ingestion and elimination rates. However, equilibrium concentrations of the radionuclide in the wild population were predictable from negative exponential models based on average annual temperature and its effects on intake and elimination rates, also suggesting that predicting ectotherm responses to long—lived contaminants (such as137Cs) may be possible without complex ecophysiological modeling.