Environmental monitoring indicators guide the decision-making process of public policies. Bioassays using seeds are a potential tool to evaluate water quality indicators, since their results can be expressed in a phytotoxicity scale, the tests are quite simple, relatively fast, and low cost. In this context, the objectives of this study were to evaluate the phytotoxicity of Rio Grande reservoir waters and its relationship with environmental variables. In Brazil, this reservoir is an important water source for the public supply of the Metropolitan Region of São Paulo. In 2018, subsurface and bottom water were collected in four sampling campaigns. The environmental variables were measured both in field and at the laboratory. In field, it was used a multiprobe of ammonium (N-NH4+), electrical conductivity (EC), nitrate (N-NO3-), pH, dissolved oxygen (DO) and water temperature (T); and, at the laboratory, of total nitrogen (TN) and total phosphorus (TP). Bioassays using seeds of cucumber (Cucumis sativus), garden cress (Lepidium sativum), and mustard (Sinapis alba) were performed (n = 6). The bioassays consisted of incubating the seeds and the water samples on Petri dishes at (20º ± 2 ºC) in the absence of light for 3 days, in case of S. alba, and for 4 days, in case of C. sativus and L. sativum. ISO water was used as a control. Afterward, the number of germinated seeds and the roots' length were determined to calculate the Germination Index (GI). Statistical analyses were performed, such as the Shapiro-Wilk test, the analysis of variance (ANOVA) and the Tukey test, to verify the differences between the seeds' responses to the environmental variables. The Principal Component Analysis (PCA) was applied to the environmental variables and the GI to identify the most representative variables. This study indicated that, in May of 2018, the subsurface water was moderately phytotoxic to L. sativum (GI = 61.3 ± 8.2%); and the bottom water was very phytotoxic to L. sativum (GI = 23.6 ± 14.5%) and S. alba (26.2 ± 12.0%). In all other sampling campaigns, the water samples potentiated seed germination, which can be related to nutrients concentrations The L. sativum proved to be more sensitive, due to the lower degrees of data dispersion and the higher degrees of toxicity when compared to the other analysed seeds.