This study assesses the ecological risks (ERA) of pesticides to aquatic organisms in the River Madre de Dios (RMD), which receives surface runoff water from banana, pineapple, and rice plantations on the Caribbean coast of Costa Rica. Water samples collected over 2 years at five sites in the RMD revealed a total of 26 pesticides. Their toxicity risk to aquatic organisms was assessed using three recent ERA models. (1) The PERPEST model showed a high probability (>50 %) of clear toxic effects of pesticide mixtures on algae, macrophytes, zooplankton, macroinvertebrates, and community metabolism and a low probability (<50 %) of clear effects on fish. (2) Species sensitivity distributions (SSD) showed a moderate to high risk of three herbicides: ametryn, bromacil, diuron and four insecticides: carbaryl, diazinon, ethoprophos, terbufos. (3) The multi-substance potentially affected fraction (msPAF) model showed results consistent with PERPEST: high risk to algae (maximum msPAF: 73 %), aquatic plants (61 %), and arthropods (25 %) and low risk to fish (0.2 %) from pesticide mixtures. The pesticides posing the highest risks according to msPAF and that should be substituted with less toxic substances were the herbicides ametryn, diuron, the insecticides carbaryl, chlorpyrifos, diazinon, ethoprophos, and the fungicide difenoconazole. Ecological risks were highest near the plantations and decreased progressively further downstream. The risk to fish was found to be relatively low in these models, but water samples were not collected during fish kill events and some highly toxic pesticides known to be used were not analyzed for in this study. Further sampling and analysis of water samples is needed to determine toxicity risks to fish during peaks of pesticide mixture concentrations. The msPAF model, which estimates the ecological risks of mixtures based on their toxic modes of action, was found to be the most suitable model to assess toxicity risks to aquatic organisms in the RMD. The PERPEST model was found to be a strong tool for screening risk assessments. The SSD approach is useful in deriving water quality criteria for specific pesticides. This study, through the application of three ERA models, clearly shows that pesticides used in plantations within the RMD watershed are expected to have severe adverse effects on most groups of aquatic organisms and that actions are urgently needed to reduce pesticide pollution in this high biodiversity ecosystem.